2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
58 #include <linux/sysctl.h>
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
65 #include <net/net_namespace.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
91 /* Set to 3 to get tracing... */
95 #define ADBG(x) printk x
100 #define INFINITY_LIFE_TIME 0xFFFFFFFF
102 static inline u32 cstamp_delta(unsigned long cstamp)
104 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
107 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
108 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
109 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
112 static void addrconf_sysctl_register(struct inet6_dev *idev);
113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
124 #ifdef CONFIG_IPV6_PRIVACY
125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
127 static void ipv6_regen_rndid(unsigned long data);
130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
131 static int ipv6_count_addresses(struct inet6_dev *idev);
134 * Configured unicast address hash table
136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
137 static DEFINE_SPINLOCK(addrconf_hash_lock);
139 static void addrconf_verify(unsigned long);
141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
142 static DEFINE_SPINLOCK(addrconf_verify_lock);
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_type_change(struct net_device *dev,
148 unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
152 static void addrconf_dad_timer(unsigned long data);
153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
154 static void addrconf_dad_run(struct inet6_dev *idev);
155 static void addrconf_rs_timer(unsigned long data);
156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
160 struct prefix_info *pinfo);
161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
162 struct net_device *dev);
164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
169 .mtu6 = IPV6_MIN_MTU,
171 .accept_redirects = 1,
173 .force_mld_version = 0,
175 .rtr_solicits = MAX_RTR_SOLICITATIONS,
176 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
177 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
178 #ifdef CONFIG_IPV6_PRIVACY
180 .temp_valid_lft = TEMP_VALID_LIFETIME,
181 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
182 .regen_max_retry = REGEN_MAX_RETRY,
183 .max_desync_factor = MAX_DESYNC_FACTOR,
185 .max_addresses = IPV6_MAX_ADDRESSES,
186 .accept_ra_defrtr = 1,
187 .accept_ra_pinfo = 1,
188 #ifdef CONFIG_IPV6_ROUTER_PREF
189 .accept_ra_rtr_pref = 1,
190 .rtr_probe_interval = 60 * HZ,
191 #ifdef CONFIG_IPV6_ROUTE_INFO
192 .accept_ra_rt_info_max_plen = 0,
196 .accept_source_route = 0, /* we do not accept RH0 by default. */
201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
204 .mtu6 = IPV6_MIN_MTU,
206 .accept_redirects = 1,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
212 #ifdef CONFIG_IPV6_PRIVACY
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
219 .max_addresses = IPV6_MAX_ADDRESSES,
220 .accept_ra_defrtr = 1,
221 .accept_ra_pinfo = 1,
222 #ifdef CONFIG_IPV6_ROUTER_PREF
223 .accept_ra_rtr_pref = 1,
224 .rtr_probe_interval = 60 * HZ,
225 #ifdef CONFIG_IPV6_ROUTE_INFO
226 .accept_ra_rt_info_max_plen = 0,
230 .accept_source_route = 0, /* we do not accept RH0 by default. */
235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
241 /* Check if a valid qdisc is available */
242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
244 return !qdisc_tx_is_noop(dev);
247 /* Check if a route is valid prefix route */
248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
250 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
255 if (del_timer(&ifp->timer))
259 enum addrconf_timer_t {
265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
266 enum addrconf_timer_t what,
269 if (!del_timer(&ifp->timer))
274 ifp->timer.function = addrconf_dad_timer;
277 ifp->timer.function = addrconf_rs_timer;
282 ifp->timer.expires = jiffies + when;
283 add_timer(&ifp->timer);
286 static int snmp6_alloc_dev(struct inet6_dev *idev)
288 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
289 sizeof(struct ipstats_mib),
290 __alignof__(struct ipstats_mib)) < 0)
292 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
293 sizeof(struct icmpv6_mib),
294 __alignof__(struct icmpv6_mib)) < 0)
296 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
297 sizeof(struct icmpv6msg_mib),
298 __alignof__(struct icmpv6msg_mib)) < 0)
304 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
306 snmp_mib_free((void __percpu **)idev->stats.ipv6);
311 static void snmp6_free_dev(struct inet6_dev *idev)
313 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
314 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
315 snmp_mib_free((void __percpu **)idev->stats.ipv6);
318 /* Nobody refers to this device, we may destroy it. */
320 void in6_dev_finish_destroy(struct inet6_dev *idev)
322 struct net_device *dev = idev->dev;
324 WARN_ON(!list_empty(&idev->addr_list));
325 WARN_ON(idev->mc_list != NULL);
327 #ifdef NET_REFCNT_DEBUG
328 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
332 pr_warning("Freeing alive inet6 device %p\n", idev);
335 snmp6_free_dev(idev);
336 kfree_rcu(idev, rcu);
339 EXPORT_SYMBOL(in6_dev_finish_destroy);
341 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
343 struct inet6_dev *ndev;
347 if (dev->mtu < IPV6_MIN_MTU)
350 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
355 rwlock_init(&ndev->lock);
357 INIT_LIST_HEAD(&ndev->addr_list);
359 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
360 ndev->cnf.mtu6 = dev->mtu;
361 ndev->cnf.sysctl = NULL;
362 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
363 if (ndev->nd_parms == NULL) {
367 if (ndev->cnf.forwarding)
368 dev_disable_lro(dev);
369 /* We refer to the device */
372 if (snmp6_alloc_dev(ndev) < 0) {
374 "%s(): cannot allocate memory for statistics; dev=%s.\n",
375 __func__, dev->name));
376 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 in6_dev_finish_destroy(ndev);
382 if (snmp6_register_dev(ndev) < 0) {
384 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
385 __func__, dev->name));
386 neigh_parms_release(&nd_tbl, ndev->nd_parms);
388 in6_dev_finish_destroy(ndev);
392 /* One reference from device. We must do this before
393 * we invoke __ipv6_regen_rndid().
397 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
398 ndev->cnf.accept_dad = -1;
400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
401 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
403 "%s: Disabled Multicast RS\n",
405 ndev->cnf.rtr_solicits = 0;
409 #ifdef CONFIG_IPV6_PRIVACY
410 INIT_LIST_HEAD(&ndev->tempaddr_list);
411 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
412 if ((dev->flags&IFF_LOOPBACK) ||
413 dev->type == ARPHRD_TUNNEL ||
414 dev->type == ARPHRD_TUNNEL6 ||
415 dev->type == ARPHRD_SIT ||
416 dev->type == ARPHRD_NONE) {
417 ndev->cnf.use_tempaddr = -1;
420 ipv6_regen_rndid((unsigned long) ndev);
424 if (netif_running(dev) && addrconf_qdisc_ok(dev))
425 ndev->if_flags |= IF_READY;
427 ipv6_mc_init_dev(ndev);
428 ndev->tstamp = jiffies;
429 addrconf_sysctl_register(ndev);
430 /* protected by rtnl_lock */
431 rcu_assign_pointer(dev->ip6_ptr, ndev);
433 /* Join all-node multicast group */
434 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
441 struct inet6_dev *idev;
445 idev = __in6_dev_get(dev);
447 idev = ipv6_add_dev(dev);
452 if (dev->flags&IFF_UP)
458 static void dev_forward_change(struct inet6_dev *idev)
460 struct net_device *dev;
461 struct inet6_ifaddr *ifa;
466 if (idev->cnf.forwarding)
467 dev_disable_lro(dev);
468 if (dev && (dev->flags & IFF_MULTICAST)) {
469 if (idev->cnf.forwarding)
470 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
472 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
475 list_for_each_entry(ifa, &idev->addr_list, if_list) {
476 if (ifa->flags&IFA_F_TENTATIVE)
478 if (idev->cnf.forwarding)
479 addrconf_join_anycast(ifa);
481 addrconf_leave_anycast(ifa);
486 static void addrconf_forward_change(struct net *net, __s32 newf)
488 struct net_device *dev;
489 struct inet6_dev *idev;
492 for_each_netdev_rcu(net, dev) {
493 idev = __in6_dev_get(dev);
495 int changed = (!idev->cnf.forwarding) ^ (!newf);
496 idev->cnf.forwarding = newf;
498 dev_forward_change(idev);
504 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
508 net = (struct net *)table->extra2;
509 if (p == &net->ipv6.devconf_dflt->forwarding)
512 if (!rtnl_trylock()) {
513 /* Restore the original values before restarting */
515 return restart_syscall();
518 if (p == &net->ipv6.devconf_all->forwarding) {
519 __s32 newf = net->ipv6.devconf_all->forwarding;
520 net->ipv6.devconf_dflt->forwarding = newf;
521 addrconf_forward_change(net, newf);
522 } else if ((!*p) ^ (!old))
523 dev_forward_change((struct inet6_dev *)table->extra1);
527 rt6_purge_dflt_routers(net);
532 /* Nobody refers to this ifaddr, destroy it */
533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
535 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
537 #ifdef NET_REFCNT_DEBUG
538 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
541 in6_dev_put(ifp->idev);
543 if (del_timer(&ifp->timer))
544 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
546 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
547 pr_warning("Freeing alive inet6 address %p\n", ifp);
550 dst_release(&ifp->rt->dst);
556 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
559 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
562 * Each device address list is sorted in order of scope -
563 * global before linklocal.
565 list_for_each(p, &idev->addr_list) {
566 struct inet6_ifaddr *ifa
567 = list_entry(p, struct inet6_ifaddr, if_list);
568 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
572 list_add_tail(&ifp->if_list, p);
575 static u32 ipv6_addr_hash(const struct in6_addr *addr)
578 * We perform the hash function over the last 64 bits of the address
579 * This will include the IEEE address token on links that support it.
581 return jhash_2words((__force u32)addr->s6_addr32[2],
582 (__force u32)addr->s6_addr32[3], 0)
583 & (IN6_ADDR_HSIZE - 1);
586 /* On success it returns ifp with increased reference count */
588 static struct inet6_ifaddr *
589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
590 int scope, u32 flags)
592 struct inet6_ifaddr *ifa = NULL;
596 int addr_type = ipv6_addr_type(addr);
598 if (addr_type == IPV6_ADDR_ANY ||
599 addr_type & IPV6_ADDR_MULTICAST ||
600 (!(idev->dev->flags & IFF_LOOPBACK) &&
601 addr_type & IPV6_ADDR_LOOPBACK))
602 return ERR_PTR(-EADDRNOTAVAIL);
606 err = -ENODEV; /*XXX*/
610 if (idev->cnf.disable_ipv6) {
615 spin_lock(&addrconf_hash_lock);
617 /* Ignore adding duplicate addresses on an interface */
618 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
619 ADBG(("ipv6_add_addr: already assigned\n"));
624 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
627 ADBG(("ipv6_add_addr: malloc failed\n"));
632 rt = addrconf_dst_alloc(idev, addr, 0);
638 ipv6_addr_copy(&ifa->addr, addr);
640 spin_lock_init(&ifa->lock);
641 spin_lock_init(&ifa->state_lock);
642 init_timer(&ifa->timer);
643 INIT_HLIST_NODE(&ifa->addr_lst);
644 ifa->timer.data = (unsigned long) ifa;
646 ifa->prefix_len = pfxlen;
647 ifa->flags = flags | IFA_F_TENTATIVE;
648 ifa->cstamp = ifa->tstamp = jiffies;
653 * part one of RFC 4429, section 3.3
654 * We should not configure an address as
655 * optimistic if we do not yet know the link
656 * layer address of our nexhop router
659 if (rt->rt6i_nexthop == NULL)
660 ifa->flags &= ~IFA_F_OPTIMISTIC;
667 /* Add to big hash table */
668 hash = ipv6_addr_hash(addr);
670 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
671 spin_unlock(&addrconf_hash_lock);
673 write_lock(&idev->lock);
674 /* Add to inet6_dev unicast addr list. */
675 ipv6_link_dev_addr(idev, ifa);
677 #ifdef CONFIG_IPV6_PRIVACY
678 if (ifa->flags&IFA_F_TEMPORARY) {
679 list_add(&ifa->tmp_list, &idev->tempaddr_list);
685 write_unlock(&idev->lock);
687 rcu_read_unlock_bh();
689 if (likely(err == 0))
690 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
698 spin_unlock(&addrconf_hash_lock);
702 /* This function wants to get referenced ifp and releases it before return */
704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
706 struct inet6_ifaddr *ifa, *ifn;
707 struct inet6_dev *idev = ifp->idev;
709 int deleted = 0, onlink = 0;
710 unsigned long expires = jiffies;
712 spin_lock_bh(&ifp->state_lock);
714 ifp->state = INET6_IFADDR_STATE_DEAD;
715 spin_unlock_bh(&ifp->state_lock);
717 if (state == INET6_IFADDR_STATE_DEAD)
720 spin_lock_bh(&addrconf_hash_lock);
721 hlist_del_init_rcu(&ifp->addr_lst);
722 spin_unlock_bh(&addrconf_hash_lock);
724 write_lock_bh(&idev->lock);
725 #ifdef CONFIG_IPV6_PRIVACY
726 if (ifp->flags&IFA_F_TEMPORARY) {
727 list_del(&ifp->tmp_list);
729 in6_ifa_put(ifp->ifpub);
736 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
738 list_del_init(&ifp->if_list);
741 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
745 } else if (ifp->flags & IFA_F_PERMANENT) {
746 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
748 if (ifa->flags & IFA_F_PERMANENT) {
753 unsigned long lifetime;
758 spin_lock(&ifa->lock);
760 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
762 * Note: Because this address is
763 * not permanent, lifetime <
764 * LONG_MAX / HZ here.
766 if (time_before(expires,
767 ifa->tstamp + lifetime * HZ))
768 expires = ifa->tstamp + lifetime * HZ;
769 spin_unlock(&ifa->lock);
774 write_unlock_bh(&idev->lock);
776 addrconf_del_timer(ifp);
778 ipv6_ifa_notify(RTM_DELADDR, ifp);
780 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
783 * Purge or update corresponding prefix
785 * 1) we don't purge prefix here if address was not permanent.
786 * prefix is managed by its own lifetime.
787 * 2) if there're no addresses, delete prefix.
788 * 3) if there're still other permanent address(es),
789 * corresponding prefix is still permanent.
790 * 4) otherwise, update prefix lifetime to the
791 * longest valid lifetime among the corresponding
792 * addresses on the device.
793 * Note: subsequent RA will update lifetime.
797 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
798 struct in6_addr prefix;
800 struct net *net = dev_net(ifp->idev->dev);
801 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
802 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
804 if (rt && addrconf_is_prefix_route(rt)) {
808 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
809 rt->rt6i_expires = expires;
810 rt->rt6i_flags |= RTF_EXPIRES;
813 dst_release(&rt->dst);
820 #ifdef CONFIG_IPV6_PRIVACY
821 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
823 struct inet6_dev *idev = ifp->idev;
824 struct in6_addr addr, *tmpaddr;
825 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
826 unsigned long regen_advance;
832 write_lock(&idev->lock);
834 spin_lock_bh(&ift->lock);
835 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
836 spin_unlock_bh(&ift->lock);
843 if (idev->cnf.use_tempaddr <= 0) {
844 write_unlock(&idev->lock);
846 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
851 spin_lock_bh(&ifp->lock);
852 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
853 idev->cnf.use_tempaddr = -1; /*XXX*/
854 spin_unlock_bh(&ifp->lock);
855 write_unlock(&idev->lock);
857 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
863 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
864 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
865 spin_unlock_bh(&ifp->lock);
866 write_unlock(&idev->lock);
868 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
874 memcpy(&addr.s6_addr[8], idev->rndid, 8);
875 age = (jiffies - ifp->tstamp) / HZ;
876 tmp_valid_lft = min_t(__u32,
878 idev->cnf.temp_valid_lft + age);
879 tmp_prefered_lft = min_t(__u32,
881 idev->cnf.temp_prefered_lft + age -
882 idev->cnf.max_desync_factor);
883 tmp_plen = ifp->prefix_len;
884 max_addresses = idev->cnf.max_addresses;
885 tmp_cstamp = ifp->cstamp;
886 tmp_tstamp = ifp->tstamp;
887 spin_unlock_bh(&ifp->lock);
889 regen_advance = idev->cnf.regen_max_retry *
890 idev->cnf.dad_transmits *
891 idev->nd_parms->retrans_time / HZ;
892 write_unlock(&idev->lock);
894 /* A temporary address is created only if this calculated Preferred
895 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
896 * an implementation must not create a temporary address with a zero
897 * Preferred Lifetime.
899 if (tmp_prefered_lft <= regen_advance) {
906 addr_flags = IFA_F_TEMPORARY;
907 /* set in addrconf_prefix_rcv() */
908 if (ifp->flags & IFA_F_OPTIMISTIC)
909 addr_flags |= IFA_F_OPTIMISTIC;
911 ift = !max_addresses ||
912 ipv6_count_addresses(idev) < max_addresses ?
913 ipv6_add_addr(idev, &addr, tmp_plen,
914 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
916 if (!ift || IS_ERR(ift)) {
920 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
922 write_lock(&idev->lock);
926 spin_lock_bh(&ift->lock);
928 ift->valid_lft = tmp_valid_lft;
929 ift->prefered_lft = tmp_prefered_lft;
930 ift->cstamp = tmp_cstamp;
931 ift->tstamp = tmp_tstamp;
932 spin_unlock_bh(&ift->lock);
934 addrconf_dad_start(ift, 0);
943 * Choose an appropriate source address (RFC3484)
946 IPV6_SADDR_RULE_INIT = 0,
947 IPV6_SADDR_RULE_LOCAL,
948 IPV6_SADDR_RULE_SCOPE,
949 IPV6_SADDR_RULE_PREFERRED,
950 #ifdef CONFIG_IPV6_MIP6
954 IPV6_SADDR_RULE_LABEL,
955 #ifdef CONFIG_IPV6_PRIVACY
956 IPV6_SADDR_RULE_PRIVACY,
958 IPV6_SADDR_RULE_ORCHID,
959 IPV6_SADDR_RULE_PREFIX,
963 struct ipv6_saddr_score {
966 struct inet6_ifaddr *ifa;
967 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
972 struct ipv6_saddr_dst {
973 const struct in6_addr *addr;
980 static inline int ipv6_saddr_preferred(int type)
982 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
987 static int ipv6_get_saddr_eval(struct net *net,
988 struct ipv6_saddr_score *score,
989 struct ipv6_saddr_dst *dst,
994 if (i <= score->rule) {
996 case IPV6_SADDR_RULE_SCOPE:
997 ret = score->scopedist;
999 case IPV6_SADDR_RULE_PREFIX:
1000 ret = score->matchlen;
1003 ret = !!test_bit(i, score->scorebits);
1009 case IPV6_SADDR_RULE_INIT:
1010 /* Rule 0: remember if hiscore is not ready yet */
1013 case IPV6_SADDR_RULE_LOCAL:
1014 /* Rule 1: Prefer same address */
1015 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1017 case IPV6_SADDR_RULE_SCOPE:
1018 /* Rule 2: Prefer appropriate scope
1023 * ---+--+-+---> scope
1025 * | d is scope of the destination.
1027 * | \ <- smaller scope is better if
1028 * B-15 | \ if scope is enough for destinaion.
1029 * | ret = B - scope (-1 <= scope >= d <= 15).
1031 * |/ <- greater is better
1032 * -C / if scope is not enough for destination.
1033 * /| ret = scope - C (-1 <= d < scope <= 15).
1035 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1036 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1037 * Assume B = 0 and we get C > 29.
1039 ret = __ipv6_addr_src_scope(score->addr_type);
1040 if (ret >= dst->scope)
1043 ret -= 128; /* 30 is enough */
1044 score->scopedist = ret;
1046 case IPV6_SADDR_RULE_PREFERRED:
1047 /* Rule 3: Avoid deprecated and optimistic addresses */
1048 ret = ipv6_saddr_preferred(score->addr_type) ||
1049 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1051 #ifdef CONFIG_IPV6_MIP6
1052 case IPV6_SADDR_RULE_HOA:
1054 /* Rule 4: Prefer home address */
1055 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1056 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1060 case IPV6_SADDR_RULE_OIF:
1061 /* Rule 5: Prefer outgoing interface */
1062 ret = (!dst->ifindex ||
1063 dst->ifindex == score->ifa->idev->dev->ifindex);
1065 case IPV6_SADDR_RULE_LABEL:
1066 /* Rule 6: Prefer matching label */
1067 ret = ipv6_addr_label(net,
1068 &score->ifa->addr, score->addr_type,
1069 score->ifa->idev->dev->ifindex) == dst->label;
1071 #ifdef CONFIG_IPV6_PRIVACY
1072 case IPV6_SADDR_RULE_PRIVACY:
1074 /* Rule 7: Prefer public address
1075 * Note: prefer temporary address if use_tempaddr >= 2
1077 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1078 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1079 score->ifa->idev->cnf.use_tempaddr >= 2;
1080 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1084 case IPV6_SADDR_RULE_ORCHID:
1085 /* Rule 8-: Prefer ORCHID vs ORCHID or
1086 * non-ORCHID vs non-ORCHID
1088 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1089 ipv6_addr_orchid(dst->addr));
1091 case IPV6_SADDR_RULE_PREFIX:
1092 /* Rule 8: Use longest matching prefix */
1093 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1101 __set_bit(i, score->scorebits);
1107 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1108 const struct in6_addr *daddr, unsigned int prefs,
1109 struct in6_addr *saddr)
1111 struct ipv6_saddr_score scores[2],
1112 *score = &scores[0], *hiscore = &scores[1];
1113 struct ipv6_saddr_dst dst;
1114 struct net_device *dev;
1117 dst_type = __ipv6_addr_type(daddr);
1119 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1120 dst.scope = __ipv6_addr_src_scope(dst_type);
1121 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1125 hiscore->ifa = NULL;
1129 for_each_netdev_rcu(net, dev) {
1130 struct inet6_dev *idev;
1132 /* Candidate Source Address (section 4)
1133 * - multicast and link-local destination address,
1134 * the set of candidate source address MUST only
1135 * include addresses assigned to interfaces
1136 * belonging to the same link as the outgoing
1138 * (- For site-local destination addresses, the
1139 * set of candidate source addresses MUST only
1140 * include addresses assigned to interfaces
1141 * belonging to the same site as the outgoing
1144 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1145 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1146 dst.ifindex && dev->ifindex != dst.ifindex)
1149 idev = __in6_dev_get(dev);
1153 read_lock_bh(&idev->lock);
1154 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1158 * - Tentative Address (RFC2462 section 5.4)
1159 * - A tentative address is not considered
1160 * "assigned to an interface" in the traditional
1161 * sense, unless it is also flagged as optimistic.
1162 * - Candidate Source Address (section 4)
1163 * - In any case, anycast addresses, multicast
1164 * addresses, and the unspecified address MUST
1165 * NOT be included in a candidate set.
1167 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1168 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1173 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1174 score->addr_type & IPV6_ADDR_MULTICAST)) {
1175 LIMIT_NETDEBUG(KERN_DEBUG
1176 "ADDRCONF: unspecified / multicast address "
1177 "assigned as unicast address on %s",
1183 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1185 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1186 int minihiscore, miniscore;
1188 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1189 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1191 if (minihiscore > miniscore) {
1192 if (i == IPV6_SADDR_RULE_SCOPE &&
1193 score->scopedist > 0) {
1196 * each remaining entry
1197 * has too small (not enough)
1198 * scope, because ifa entries
1199 * are sorted by their scope
1205 } else if (minihiscore < miniscore) {
1207 in6_ifa_put(hiscore->ifa);
1209 in6_ifa_hold(score->ifa);
1211 swap(hiscore, score);
1213 /* restore our iterator */
1214 score->ifa = hiscore->ifa;
1221 read_unlock_bh(&idev->lock);
1226 return -EADDRNOTAVAIL;
1228 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1229 in6_ifa_put(hiscore->ifa);
1232 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1234 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1235 unsigned char banned_flags)
1237 struct inet6_dev *idev;
1238 int err = -EADDRNOTAVAIL;
1241 idev = __in6_dev_get(dev);
1243 struct inet6_ifaddr *ifp;
1245 read_lock_bh(&idev->lock);
1246 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1247 if (ifp->scope == IFA_LINK &&
1248 !(ifp->flags & banned_flags)) {
1249 ipv6_addr_copy(addr, &ifp->addr);
1254 read_unlock_bh(&idev->lock);
1260 static int ipv6_count_addresses(struct inet6_dev *idev)
1263 struct inet6_ifaddr *ifp;
1265 read_lock_bh(&idev->lock);
1266 list_for_each_entry(ifp, &idev->addr_list, if_list)
1268 read_unlock_bh(&idev->lock);
1272 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1273 struct net_device *dev, int strict)
1275 struct inet6_ifaddr *ifp;
1276 struct hlist_node *node;
1277 unsigned int hash = ipv6_addr_hash(addr);
1280 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1281 if (!net_eq(dev_net(ifp->idev->dev), net))
1283 if (ipv6_addr_equal(&ifp->addr, addr) &&
1284 !(ifp->flags&IFA_F_TENTATIVE) &&
1285 (dev == NULL || ifp->idev->dev == dev ||
1286 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1287 rcu_read_unlock_bh();
1292 rcu_read_unlock_bh();
1295 EXPORT_SYMBOL(ipv6_chk_addr);
1297 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1298 struct net_device *dev)
1300 unsigned int hash = ipv6_addr_hash(addr);
1301 struct inet6_ifaddr *ifp;
1302 struct hlist_node *node;
1304 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1305 if (!net_eq(dev_net(ifp->idev->dev), net))
1307 if (ipv6_addr_equal(&ifp->addr, addr)) {
1308 if (dev == NULL || ifp->idev->dev == dev)
1315 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1317 struct inet6_dev *idev;
1318 struct inet6_ifaddr *ifa;
1323 idev = __in6_dev_get(dev);
1325 read_lock_bh(&idev->lock);
1326 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1327 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1332 read_unlock_bh(&idev->lock);
1338 EXPORT_SYMBOL(ipv6_chk_prefix);
1340 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1341 struct net_device *dev, int strict)
1343 struct inet6_ifaddr *ifp, *result = NULL;
1344 unsigned int hash = ipv6_addr_hash(addr);
1345 struct hlist_node *node;
1348 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1349 if (!net_eq(dev_net(ifp->idev->dev), net))
1351 if (ipv6_addr_equal(&ifp->addr, addr)) {
1352 if (dev == NULL || ifp->idev->dev == dev ||
1353 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1360 rcu_read_unlock_bh();
1365 /* Gets referenced address, destroys ifaddr */
1367 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1369 if (ifp->flags&IFA_F_PERMANENT) {
1370 spin_lock_bh(&ifp->lock);
1371 addrconf_del_timer(ifp);
1372 ifp->flags |= IFA_F_TENTATIVE;
1374 ifp->flags |= IFA_F_DADFAILED;
1375 spin_unlock_bh(&ifp->lock);
1377 ipv6_ifa_notify(0, ifp);
1379 #ifdef CONFIG_IPV6_PRIVACY
1380 } else if (ifp->flags&IFA_F_TEMPORARY) {
1381 struct inet6_ifaddr *ifpub;
1382 spin_lock_bh(&ifp->lock);
1385 in6_ifa_hold(ifpub);
1386 spin_unlock_bh(&ifp->lock);
1387 ipv6_create_tempaddr(ifpub, ifp);
1390 spin_unlock_bh(&ifp->lock);
1398 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1402 spin_lock(&ifp->state_lock);
1403 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1404 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1407 spin_unlock(&ifp->state_lock);
1412 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1414 struct inet6_dev *idev = ifp->idev;
1416 if (addrconf_dad_end(ifp)) {
1421 if (net_ratelimit())
1422 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1423 ifp->idev->dev->name, &ifp->addr);
1425 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1426 struct in6_addr addr;
1428 addr.s6_addr32[0] = htonl(0xfe800000);
1429 addr.s6_addr32[1] = 0;
1431 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1432 ipv6_addr_equal(&ifp->addr, &addr)) {
1433 /* DAD failed for link-local based on MAC address */
1434 idev->cnf.disable_ipv6 = 1;
1436 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1437 ifp->idev->dev->name);
1441 addrconf_dad_stop(ifp, 1);
1444 /* Join to solicited addr multicast group. */
1446 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1448 struct in6_addr maddr;
1450 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453 addrconf_addr_solict_mult(addr, &maddr);
1454 ipv6_dev_mc_inc(dev, &maddr);
1457 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1459 struct in6_addr maddr;
1461 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464 addrconf_addr_solict_mult(addr, &maddr);
1465 __ipv6_dev_mc_dec(idev, &maddr);
1468 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1470 struct in6_addr addr;
1471 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1472 if (ipv6_addr_any(&addr))
1474 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1477 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1479 struct in6_addr addr;
1480 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1481 if (ipv6_addr_any(&addr))
1483 __ipv6_dev_ac_dec(ifp->idev, &addr);
1486 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1488 if (dev->addr_len != ETH_ALEN)
1490 memcpy(eui, dev->dev_addr, 3);
1491 memcpy(eui + 5, dev->dev_addr + 3, 3);
1494 * The zSeries OSA network cards can be shared among various
1495 * OS instances, but the OSA cards have only one MAC address.
1496 * This leads to duplicate address conflicts in conjunction
1497 * with IPv6 if more than one instance uses the same card.
1499 * The driver for these cards can deliver a unique 16-bit
1500 * identifier for each instance sharing the same card. It is
1501 * placed instead of 0xFFFE in the interface identifier. The
1502 * "u" bit of the interface identifier is not inverted in this
1503 * case. Hence the resulting interface identifier has local
1504 * scope according to RFC2373.
1507 eui[3] = (dev->dev_id >> 8) & 0xFF;
1508 eui[4] = dev->dev_id & 0xFF;
1517 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1519 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1520 if (dev->addr_len != ARCNET_ALEN)
1523 eui[7] = *(u8*)dev->dev_addr;
1527 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1529 if (dev->addr_len != INFINIBAND_ALEN)
1531 memcpy(eui, dev->dev_addr + 12, 8);
1536 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1540 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1541 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1542 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1543 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1544 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1545 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1549 memcpy(eui + 4, &addr, 4);
1553 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1555 if (dev->priv_flags & IFF_ISATAP)
1556 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1560 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1562 switch (dev->type) {
1565 case ARPHRD_IEEE802_TR:
1566 return addrconf_ifid_eui48(eui, dev);
1568 return addrconf_ifid_arcnet(eui, dev);
1569 case ARPHRD_INFINIBAND:
1570 return addrconf_ifid_infiniband(eui, dev);
1572 return addrconf_ifid_sit(eui, dev);
1577 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1580 struct inet6_ifaddr *ifp;
1582 read_lock_bh(&idev->lock);
1583 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1584 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1585 memcpy(eui, ifp->addr.s6_addr+8, 8);
1590 read_unlock_bh(&idev->lock);
1594 #ifdef CONFIG_IPV6_PRIVACY
1595 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1596 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1599 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1600 idev->rndid[0] &= ~0x02;
1603 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1604 * check if generated address is not inappropriate
1606 * - Reserved subnet anycast (RFC 2526)
1607 * 11111101 11....11 1xxxxxxx
1608 * - ISATAP (RFC4214) 6.1
1609 * 00-00-5E-FE-xx-xx-xx-xx
1611 * - XXX: already assigned to an address on the device
1613 if (idev->rndid[0] == 0xfd &&
1614 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1615 (idev->rndid[7]&0x80))
1617 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1618 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1620 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627 static void ipv6_regen_rndid(unsigned long data)
1629 struct inet6_dev *idev = (struct inet6_dev *) data;
1630 unsigned long expires;
1633 write_lock_bh(&idev->lock);
1638 if (__ipv6_regen_rndid(idev) < 0)
1642 idev->cnf.temp_prefered_lft * HZ -
1643 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1644 idev->cnf.max_desync_factor * HZ;
1645 if (time_before(expires, jiffies)) {
1647 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1652 if (!mod_timer(&idev->regen_timer, expires))
1656 write_unlock_bh(&idev->lock);
1657 rcu_read_unlock_bh();
1661 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1664 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1665 ret = __ipv6_regen_rndid(idev);
1675 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1676 unsigned long expires, u32 flags)
1678 struct fib6_config cfg = {
1679 .fc_table = RT6_TABLE_PREFIX,
1680 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1681 .fc_ifindex = dev->ifindex,
1682 .fc_expires = expires,
1684 .fc_flags = RTF_UP | flags,
1685 .fc_nlinfo.nl_net = dev_net(dev),
1686 .fc_protocol = RTPROT_KERNEL,
1689 ipv6_addr_copy(&cfg.fc_dst, pfx);
1691 /* Prevent useless cloning on PtP SIT.
1692 This thing is done here expecting that the whole
1693 class of non-broadcast devices need not cloning.
1695 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1696 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1697 cfg.fc_flags |= RTF_NONEXTHOP;
1700 ip6_route_add(&cfg);
1703 /* Create "default" multicast route to the interface */
1705 static void addrconf_add_mroute(struct net_device *dev)
1707 struct fib6_config cfg = {
1708 .fc_table = RT6_TABLE_LOCAL,
1709 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1710 .fc_ifindex = dev->ifindex,
1713 .fc_nlinfo.nl_net = dev_net(dev),
1716 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1718 ip6_route_add(&cfg);
1721 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1722 static void sit_route_add(struct net_device *dev)
1724 struct fib6_config cfg = {
1725 .fc_table = RT6_TABLE_MAIN,
1726 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1727 .fc_ifindex = dev->ifindex,
1729 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1730 .fc_nlinfo.nl_net = dev_net(dev),
1733 /* prefix length - 96 bits "::d.d.d.d" */
1734 ip6_route_add(&cfg);
1738 static void addrconf_add_lroute(struct net_device *dev)
1740 struct in6_addr addr;
1742 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1743 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1746 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1748 struct inet6_dev *idev;
1752 idev = ipv6_find_idev(dev);
1754 return ERR_PTR(-ENOBUFS);
1756 if (idev->cnf.disable_ipv6)
1757 return ERR_PTR(-EACCES);
1759 /* Add default multicast route */
1760 addrconf_add_mroute(dev);
1762 /* Add link local route */
1763 addrconf_add_lroute(dev);
1767 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1769 struct prefix_info *pinfo;
1773 struct inet6_dev *in6_dev;
1774 struct net *net = dev_net(dev);
1776 pinfo = (struct prefix_info *) opt;
1778 if (len < sizeof(struct prefix_info)) {
1779 ADBG(("addrconf: prefix option too short\n"));
1784 * Validation checks ([ADDRCONF], page 19)
1787 addr_type = ipv6_addr_type(&pinfo->prefix);
1789 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1792 valid_lft = ntohl(pinfo->valid);
1793 prefered_lft = ntohl(pinfo->prefered);
1795 if (prefered_lft > valid_lft) {
1796 if (net_ratelimit())
1797 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1801 in6_dev = in6_dev_get(dev);
1803 if (in6_dev == NULL) {
1804 if (net_ratelimit())
1805 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1810 * Two things going on here:
1811 * 1) Add routes for on-link prefixes
1812 * 2) Configure prefixes with the auto flag set
1815 if (pinfo->onlink) {
1816 struct rt6_info *rt;
1817 unsigned long rt_expires;
1819 /* Avoid arithmetic overflow. Really, we could
1820 * save rt_expires in seconds, likely valid_lft,
1821 * but it would require division in fib gc, that it
1825 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1827 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1829 if (addrconf_finite_timeout(rt_expires))
1832 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1835 if (rt && addrconf_is_prefix_route(rt)) {
1836 /* Autoconf prefix route */
1837 if (valid_lft == 0) {
1840 } else if (addrconf_finite_timeout(rt_expires)) {
1842 rt->rt6i_expires = jiffies + rt_expires;
1843 rt->rt6i_flags |= RTF_EXPIRES;
1845 rt->rt6i_flags &= ~RTF_EXPIRES;
1846 rt->rt6i_expires = 0;
1848 } else if (valid_lft) {
1849 clock_t expires = 0;
1850 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1851 if (addrconf_finite_timeout(rt_expires)) {
1853 flags |= RTF_EXPIRES;
1854 expires = jiffies_to_clock_t(rt_expires);
1856 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1857 dev, expires, flags);
1860 dst_release(&rt->dst);
1863 /* Try to figure out our local address for this prefix */
1865 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1866 struct inet6_ifaddr * ifp;
1867 struct in6_addr addr;
1868 int create = 0, update_lft = 0;
1870 if (pinfo->prefix_len == 64) {
1871 memcpy(&addr, &pinfo->prefix, 8);
1872 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1873 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1874 in6_dev_put(in6_dev);
1879 if (net_ratelimit())
1880 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1882 in6_dev_put(in6_dev);
1887 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1889 if (ifp == NULL && valid_lft) {
1890 int max_addresses = in6_dev->cnf.max_addresses;
1893 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1894 if (in6_dev->cnf.optimistic_dad &&
1895 !net->ipv6.devconf_all->forwarding)
1896 addr_flags = IFA_F_OPTIMISTIC;
1899 /* Do not allow to create too much of autoconfigured
1900 * addresses; this would be too easy way to crash kernel.
1902 if (!max_addresses ||
1903 ipv6_count_addresses(in6_dev) < max_addresses)
1904 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1905 addr_type&IPV6_ADDR_SCOPE_MASK,
1908 if (!ifp || IS_ERR(ifp)) {
1909 in6_dev_put(in6_dev);
1913 update_lft = create = 1;
1914 ifp->cstamp = jiffies;
1915 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1921 #ifdef CONFIG_IPV6_PRIVACY
1922 struct inet6_ifaddr *ift;
1926 /* update lifetime (RFC2462 5.5.3 e) */
1927 spin_lock(&ifp->lock);
1929 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1930 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1933 if (!update_lft && stored_lft) {
1934 if (valid_lft > MIN_VALID_LIFETIME ||
1935 valid_lft > stored_lft)
1937 else if (stored_lft <= MIN_VALID_LIFETIME) {
1938 /* valid_lft <= stored_lft is always true */
1940 * RFC 4862 Section 5.5.3e:
1941 * "Note that the preferred lifetime of
1942 * the corresponding address is always
1943 * reset to the Preferred Lifetime in
1944 * the received Prefix Information
1945 * option, regardless of whether the
1946 * valid lifetime is also reset or
1949 * So if the preferred lifetime in
1950 * this advertisement is different
1951 * than what we have stored, but the
1952 * valid lifetime is invalid, just
1953 * reset prefered_lft.
1955 * We must set the valid lifetime
1956 * to the stored lifetime since we'll
1957 * be updating the timestamp below,
1958 * else we'll set it back to the
1961 if (prefered_lft != ifp->prefered_lft) {
1962 valid_lft = stored_lft;
1966 valid_lft = MIN_VALID_LIFETIME;
1967 if (valid_lft < prefered_lft)
1968 prefered_lft = valid_lft;
1974 ifp->valid_lft = valid_lft;
1975 ifp->prefered_lft = prefered_lft;
1978 ifp->flags &= ~IFA_F_DEPRECATED;
1979 spin_unlock(&ifp->lock);
1981 if (!(flags&IFA_F_TENTATIVE))
1982 ipv6_ifa_notify(0, ifp);
1984 spin_unlock(&ifp->lock);
1986 #ifdef CONFIG_IPV6_PRIVACY
1987 read_lock_bh(&in6_dev->lock);
1988 /* update all temporary addresses in the list */
1989 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1991 * When adjusting the lifetimes of an existing
1992 * temporary address, only lower the lifetimes.
1993 * Implementations must not increase the
1994 * lifetimes of an existing temporary address
1995 * when processing a Prefix Information Option.
1997 if (ifp != ift->ifpub)
2000 spin_lock(&ift->lock);
2002 if (ift->valid_lft > valid_lft &&
2003 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2004 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2005 if (ift->prefered_lft > prefered_lft &&
2006 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2007 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2008 spin_unlock(&ift->lock);
2009 if (!(flags&IFA_F_TENTATIVE))
2010 ipv6_ifa_notify(0, ift);
2013 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2015 * When a new public address is created as described in [ADDRCONF],
2016 * also create a new temporary address. Also create a temporary
2017 * address if it's enabled but no temporary address currently exists.
2019 read_unlock_bh(&in6_dev->lock);
2020 ipv6_create_tempaddr(ifp, NULL);
2022 read_unlock_bh(&in6_dev->lock);
2029 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2030 in6_dev_put(in6_dev);
2034 * Set destination address.
2035 * Special case for SIT interfaces where we create a new "virtual"
2038 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2040 struct in6_ifreq ireq;
2041 struct net_device *dev;
2047 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2050 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2056 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2057 if (dev->type == ARPHRD_SIT) {
2058 const struct net_device_ops *ops = dev->netdev_ops;
2060 struct ip_tunnel_parm p;
2062 err = -EADDRNOTAVAIL;
2063 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2066 memset(&p, 0, sizeof(p));
2067 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2071 p.iph.protocol = IPPROTO_IPV6;
2073 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2075 if (ops->ndo_do_ioctl) {
2076 mm_segment_t oldfs = get_fs();
2079 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2086 dev = __dev_get_by_name(net, p.name);
2089 err = dev_open(dev);
2100 * Manual configuration of address on an interface
2102 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2103 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2106 struct inet6_ifaddr *ifp;
2107 struct inet6_dev *idev;
2108 struct net_device *dev;
2112 unsigned long timeout;
2119 /* check the lifetime */
2120 if (!valid_lft || prefered_lft > valid_lft)
2123 dev = __dev_get_by_index(net, ifindex);
2127 idev = addrconf_add_dev(dev);
2129 return PTR_ERR(idev);
2131 scope = ipv6_addr_scope(pfx);
2133 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2134 if (addrconf_finite_timeout(timeout)) {
2135 expires = jiffies_to_clock_t(timeout * HZ);
2136 valid_lft = timeout;
2137 flags = RTF_EXPIRES;
2141 ifa_flags |= IFA_F_PERMANENT;
2144 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2145 if (addrconf_finite_timeout(timeout)) {
2147 ifa_flags |= IFA_F_DEPRECATED;
2148 prefered_lft = timeout;
2151 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2154 spin_lock_bh(&ifp->lock);
2155 ifp->valid_lft = valid_lft;
2156 ifp->prefered_lft = prefered_lft;
2157 ifp->tstamp = jiffies;
2158 spin_unlock_bh(&ifp->lock);
2160 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2163 * Note that section 3.1 of RFC 4429 indicates
2164 * that the Optimistic flag should not be set for
2165 * manually configured addresses
2167 addrconf_dad_start(ifp, 0);
2173 return PTR_ERR(ifp);
2176 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2179 struct inet6_ifaddr *ifp;
2180 struct inet6_dev *idev;
2181 struct net_device *dev;
2186 dev = __dev_get_by_index(net, ifindex);
2190 if ((idev = __in6_dev_get(dev)) == NULL)
2193 read_lock_bh(&idev->lock);
2194 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2195 if (ifp->prefix_len == plen &&
2196 ipv6_addr_equal(pfx, &ifp->addr)) {
2198 read_unlock_bh(&idev->lock);
2202 /* If the last address is deleted administratively,
2203 disable IPv6 on this interface.
2205 if (list_empty(&idev->addr_list))
2206 addrconf_ifdown(idev->dev, 1);
2210 read_unlock_bh(&idev->lock);
2211 return -EADDRNOTAVAIL;
2215 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2217 struct in6_ifreq ireq;
2220 if (!capable(CAP_NET_ADMIN))
2223 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2227 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2228 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2229 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2234 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2236 struct in6_ifreq ireq;
2239 if (!capable(CAP_NET_ADMIN))
2242 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2246 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2247 ireq.ifr6_prefixlen);
2252 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2253 int plen, int scope)
2255 struct inet6_ifaddr *ifp;
2257 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2259 spin_lock_bh(&ifp->lock);
2260 ifp->flags &= ~IFA_F_TENTATIVE;
2261 spin_unlock_bh(&ifp->lock);
2262 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2267 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2268 static void sit_add_v4_addrs(struct inet6_dev *idev)
2270 struct in6_addr addr;
2271 struct net_device *dev;
2272 struct net *net = dev_net(idev->dev);
2277 memset(&addr, 0, sizeof(struct in6_addr));
2278 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2280 if (idev->dev->flags&IFF_POINTOPOINT) {
2281 addr.s6_addr32[0] = htonl(0xfe800000);
2284 scope = IPV6_ADDR_COMPATv4;
2287 if (addr.s6_addr32[3]) {
2288 add_addr(idev, &addr, 128, scope);
2292 for_each_netdev(net, dev) {
2293 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2294 if (in_dev && (dev->flags & IFF_UP)) {
2295 struct in_ifaddr * ifa;
2299 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2302 addr.s6_addr32[3] = ifa->ifa_local;
2304 if (ifa->ifa_scope == RT_SCOPE_LINK)
2306 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2307 if (idev->dev->flags&IFF_POINTOPOINT)
2311 if (idev->dev->flags&IFF_POINTOPOINT)
2316 add_addr(idev, &addr, plen, flag);
2323 static void init_loopback(struct net_device *dev)
2325 struct inet6_dev *idev;
2331 if ((idev = ipv6_find_idev(dev)) == NULL) {
2332 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2336 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2339 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2341 struct inet6_ifaddr * ifp;
2342 u32 addr_flags = IFA_F_PERMANENT;
2344 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2345 if (idev->cnf.optimistic_dad &&
2346 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2347 addr_flags |= IFA_F_OPTIMISTIC;
2351 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2353 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2354 addrconf_dad_start(ifp, 0);
2359 static void addrconf_dev_config(struct net_device *dev)
2361 struct in6_addr addr;
2362 struct inet6_dev * idev;
2366 if ((dev->type != ARPHRD_ETHER) &&
2367 (dev->type != ARPHRD_FDDI) &&
2368 (dev->type != ARPHRD_IEEE802_TR) &&
2369 (dev->type != ARPHRD_ARCNET) &&
2370 (dev->type != ARPHRD_INFINIBAND)) {
2371 /* Alas, we support only Ethernet autoconfiguration. */
2375 idev = addrconf_add_dev(dev);
2379 memset(&addr, 0, sizeof(struct in6_addr));
2380 addr.s6_addr32[0] = htonl(0xFE800000);
2382 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2383 addrconf_add_linklocal(idev, &addr);
2386 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2387 static void addrconf_sit_config(struct net_device *dev)
2389 struct inet6_dev *idev;
2394 * Configure the tunnel with one of our IPv4
2395 * addresses... we should configure all of
2396 * our v4 addrs in the tunnel
2399 if ((idev = ipv6_find_idev(dev)) == NULL) {
2400 printk(KERN_DEBUG "init sit: add_dev failed\n");
2404 if (dev->priv_flags & IFF_ISATAP) {
2405 struct in6_addr addr;
2407 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2408 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2409 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2410 addrconf_add_linklocal(idev, &addr);
2414 sit_add_v4_addrs(idev);
2416 if (dev->flags&IFF_POINTOPOINT) {
2417 addrconf_add_mroute(dev);
2418 addrconf_add_lroute(dev);
2425 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2427 struct in6_addr lladdr;
2429 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2430 addrconf_add_linklocal(idev, &lladdr);
2436 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2438 struct net_device *link_dev;
2439 struct net *net = dev_net(idev->dev);
2441 /* first try to inherit the link-local address from the link device */
2442 if (idev->dev->iflink &&
2443 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2444 if (!ipv6_inherit_linklocal(idev, link_dev))
2447 /* then try to inherit it from any device */
2448 for_each_netdev(net, link_dev) {
2449 if (!ipv6_inherit_linklocal(idev, link_dev))
2452 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2456 * Autoconfigure tunnel with a link-local address so routing protocols,
2457 * DHCPv6, MLD etc. can be run over the virtual link
2460 static void addrconf_ip6_tnl_config(struct net_device *dev)
2462 struct inet6_dev *idev;
2466 idev = addrconf_add_dev(dev);
2468 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2471 ip6_tnl_add_linklocal(idev);
2474 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2477 struct net_device *dev = (struct net_device *) data;
2478 struct inet6_dev *idev = __in6_dev_get(dev);
2479 int run_pending = 0;
2483 case NETDEV_REGISTER:
2484 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2485 idev = ipv6_add_dev(dev);
2487 return notifier_from_errno(-ENOMEM);
2493 if (dev->flags & IFF_SLAVE)
2496 if (event == NETDEV_UP) {
2497 if (!addrconf_qdisc_ok(dev)) {
2498 /* device is not ready yet. */
2500 "ADDRCONF(NETDEV_UP): %s: "
2501 "link is not ready\n",
2506 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2507 idev = ipv6_add_dev(dev);
2510 idev->if_flags |= IF_READY;
2514 if (!addrconf_qdisc_ok(dev)) {
2515 /* device is still not ready. */
2520 if (idev->if_flags & IF_READY)
2521 /* device is already configured. */
2523 idev->if_flags |= IF_READY;
2527 "ADDRCONF(NETDEV_CHANGE): %s: "
2528 "link becomes ready\n",
2534 switch (dev->type) {
2535 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2537 addrconf_sit_config(dev);
2540 case ARPHRD_TUNNEL6:
2541 addrconf_ip6_tnl_config(dev);
2543 case ARPHRD_LOOPBACK:
2548 addrconf_dev_config(dev);
2554 addrconf_dad_run(idev);
2557 * If the MTU changed during the interface down,
2558 * when the interface up, the changed MTU must be
2559 * reflected in the idev as well as routers.
2561 if (idev->cnf.mtu6 != dev->mtu &&
2562 dev->mtu >= IPV6_MIN_MTU) {
2563 rt6_mtu_change(dev, dev->mtu);
2564 idev->cnf.mtu6 = dev->mtu;
2566 idev->tstamp = jiffies;
2567 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2570 * If the changed mtu during down is lower than
2571 * IPV6_MIN_MTU stop IPv6 on this interface.
2573 if (dev->mtu < IPV6_MIN_MTU)
2574 addrconf_ifdown(dev, 1);
2578 case NETDEV_CHANGEMTU:
2579 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2580 rt6_mtu_change(dev, dev->mtu);
2581 idev->cnf.mtu6 = dev->mtu;
2585 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2586 idev = ipv6_add_dev(dev);
2592 * MTU falled under IPV6_MIN_MTU.
2593 * Stop IPv6 on this interface.
2597 case NETDEV_UNREGISTER:
2599 * Remove all addresses from this interface.
2601 addrconf_ifdown(dev, event != NETDEV_DOWN);
2604 case NETDEV_CHANGENAME:
2606 snmp6_unregister_dev(idev);
2607 addrconf_sysctl_unregister(idev);
2608 addrconf_sysctl_register(idev);
2609 err = snmp6_register_dev(idev);
2611 return notifier_from_errno(err);
2615 case NETDEV_PRE_TYPE_CHANGE:
2616 case NETDEV_POST_TYPE_CHANGE:
2617 addrconf_type_change(dev, event);
2625 * addrconf module should be notified of a device going up
2627 static struct notifier_block ipv6_dev_notf = {
2628 .notifier_call = addrconf_notify,
2631 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2633 struct inet6_dev *idev;
2636 idev = __in6_dev_get(dev);
2638 if (event == NETDEV_POST_TYPE_CHANGE)
2639 ipv6_mc_remap(idev);
2640 else if (event == NETDEV_PRE_TYPE_CHANGE)
2641 ipv6_mc_unmap(idev);
2644 static int addrconf_ifdown(struct net_device *dev, int how)
2646 struct net *net = dev_net(dev);
2647 struct inet6_dev *idev;
2648 struct inet6_ifaddr *ifa;
2653 rt6_ifdown(net, dev);
2654 neigh_ifdown(&nd_tbl, dev);
2656 idev = __in6_dev_get(dev);
2661 * Step 1: remove reference to ipv6 device from parent device.
2667 /* protected by rtnl_lock */
2668 rcu_assign_pointer(dev->ip6_ptr, NULL);
2670 /* Step 1.5: remove snmp6 entry */
2671 snmp6_unregister_dev(idev);
2675 /* Step 2: clear hash table */
2676 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2677 struct hlist_head *h = &inet6_addr_lst[i];
2678 struct hlist_node *n;
2680 spin_lock_bh(&addrconf_hash_lock);
2682 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2683 if (ifa->idev == idev) {
2684 hlist_del_init_rcu(&ifa->addr_lst);
2685 addrconf_del_timer(ifa);
2689 spin_unlock_bh(&addrconf_hash_lock);
2692 write_lock_bh(&idev->lock);
2694 /* Step 2: clear flags for stateless addrconf */
2696 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2698 #ifdef CONFIG_IPV6_PRIVACY
2699 if (how && del_timer(&idev->regen_timer))
2702 /* Step 3: clear tempaddr list */
2703 while (!list_empty(&idev->tempaddr_list)) {
2704 ifa = list_first_entry(&idev->tempaddr_list,
2705 struct inet6_ifaddr, tmp_list);
2706 list_del(&ifa->tmp_list);
2707 write_unlock_bh(&idev->lock);
2708 spin_lock_bh(&ifa->lock);
2711 in6_ifa_put(ifa->ifpub);
2714 spin_unlock_bh(&ifa->lock);
2716 write_lock_bh(&idev->lock);
2720 while (!list_empty(&idev->addr_list)) {
2721 ifa = list_first_entry(&idev->addr_list,
2722 struct inet6_ifaddr, if_list);
2723 addrconf_del_timer(ifa);
2725 list_del(&ifa->if_list);
2727 write_unlock_bh(&idev->lock);
2729 spin_lock_bh(&ifa->state_lock);
2731 ifa->state = INET6_IFADDR_STATE_DEAD;
2732 spin_unlock_bh(&ifa->state_lock);
2734 if (state != INET6_IFADDR_STATE_DEAD) {
2735 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2736 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2740 write_lock_bh(&idev->lock);
2743 write_unlock_bh(&idev->lock);
2745 /* Step 5: Discard multicast list */
2747 ipv6_mc_destroy_dev(idev);
2751 idev->tstamp = jiffies;
2753 /* Last: Shot the device (if unregistered) */
2755 addrconf_sysctl_unregister(idev);
2756 neigh_parms_release(&nd_tbl, idev->nd_parms);
2757 neigh_ifdown(&nd_tbl, dev);
2763 static void addrconf_rs_timer(unsigned long data)
2765 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2766 struct inet6_dev *idev = ifp->idev;
2768 read_lock(&idev->lock);
2769 if (idev->dead || !(idev->if_flags & IF_READY))
2772 if (idev->cnf.forwarding)
2775 /* Announcement received after solicitation was sent */
2776 if (idev->if_flags & IF_RA_RCVD)
2779 spin_lock(&ifp->lock);
2780 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2781 /* The wait after the last probe can be shorter */
2782 addrconf_mod_timer(ifp, AC_RS,
2783 (ifp->probes == idev->cnf.rtr_solicits) ?
2784 idev->cnf.rtr_solicit_delay :
2785 idev->cnf.rtr_solicit_interval);
2786 spin_unlock(&ifp->lock);
2788 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2790 spin_unlock(&ifp->lock);
2792 * Note: we do not support deprecated "all on-link"
2793 * assumption any longer.
2795 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2800 read_unlock(&idev->lock);
2805 * Duplicate Address Detection
2807 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2809 unsigned long rand_num;
2810 struct inet6_dev *idev = ifp->idev;
2812 if (ifp->flags & IFA_F_OPTIMISTIC)
2815 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2817 ifp->probes = idev->cnf.dad_transmits;
2818 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2821 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2823 struct inet6_dev *idev = ifp->idev;
2824 struct net_device *dev = idev->dev;
2826 addrconf_join_solict(dev, &ifp->addr);
2828 net_srandom(ifp->addr.s6_addr32[3]);
2830 read_lock_bh(&idev->lock);
2831 spin_lock(&ifp->lock);
2832 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2835 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2836 idev->cnf.accept_dad < 1 ||
2837 !(ifp->flags&IFA_F_TENTATIVE) ||
2838 ifp->flags & IFA_F_NODAD) {
2839 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2840 spin_unlock(&ifp->lock);
2841 read_unlock_bh(&idev->lock);
2843 addrconf_dad_completed(ifp);
2847 if (!(idev->if_flags & IF_READY)) {
2848 spin_unlock(&ifp->lock);
2849 read_unlock_bh(&idev->lock);
2851 * If the device is not ready:
2852 * - keep it tentative if it is a permanent address.
2853 * - otherwise, kill it.
2856 addrconf_dad_stop(ifp, 0);
2861 * Optimistic nodes can start receiving
2864 if (ifp->flags & IFA_F_OPTIMISTIC)
2865 ip6_ins_rt(ifp->rt);
2867 addrconf_dad_kick(ifp);
2869 spin_unlock(&ifp->lock);
2870 read_unlock_bh(&idev->lock);
2873 static void addrconf_dad_timer(unsigned long data)
2875 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2876 struct inet6_dev *idev = ifp->idev;
2877 struct in6_addr mcaddr;
2879 if (!ifp->probes && addrconf_dad_end(ifp))
2882 read_lock(&idev->lock);
2883 if (idev->dead || !(idev->if_flags & IF_READY)) {
2884 read_unlock(&idev->lock);
2888 spin_lock(&ifp->lock);
2889 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2890 spin_unlock(&ifp->lock);
2891 read_unlock(&idev->lock);
2895 if (ifp->probes == 0) {
2897 * DAD was successful
2900 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2901 spin_unlock(&ifp->lock);
2902 read_unlock(&idev->lock);
2904 addrconf_dad_completed(ifp);
2910 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2911 spin_unlock(&ifp->lock);
2912 read_unlock(&idev->lock);
2914 /* send a neighbour solicitation for our addr */
2915 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2916 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2921 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2923 struct net_device *dev = ifp->idev->dev;
2926 * Configure the address for reception. Now it is valid.
2929 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2931 /* If added prefix is link local and forwarding is off,
2932 start sending router solicitations.
2935 if ((ifp->idev->cnf.forwarding == 0 ||
2936 ifp->idev->cnf.forwarding == 2) &&
2937 ifp->idev->cnf.rtr_solicits > 0 &&
2938 (dev->flags&IFF_LOOPBACK) == 0 &&
2939 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2941 * If a host as already performed a random delay
2942 * [...] as part of DAD [...] there is no need
2943 * to delay again before sending the first RS
2945 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2947 spin_lock_bh(&ifp->lock);
2949 ifp->idev->if_flags |= IF_RS_SENT;
2950 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2951 spin_unlock_bh(&ifp->lock);
2955 static void addrconf_dad_run(struct inet6_dev *idev)
2957 struct inet6_ifaddr *ifp;
2959 read_lock_bh(&idev->lock);
2960 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2961 spin_lock(&ifp->lock);
2962 if (ifp->flags & IFA_F_TENTATIVE &&
2963 ifp->state == INET6_IFADDR_STATE_DAD)
2964 addrconf_dad_kick(ifp);
2965 spin_unlock(&ifp->lock);
2967 read_unlock_bh(&idev->lock);
2970 #ifdef CONFIG_PROC_FS
2971 struct if6_iter_state {
2972 struct seq_net_private p;
2976 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2978 struct inet6_ifaddr *ifa = NULL;
2979 struct if6_iter_state *state = seq->private;
2980 struct net *net = seq_file_net(seq);
2982 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2983 struct hlist_node *n;
2984 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
2986 if (net_eq(dev_net(ifa->idev->dev), net))
2992 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2993 struct inet6_ifaddr *ifa)
2995 struct if6_iter_state *state = seq->private;
2996 struct net *net = seq_file_net(seq);
2997 struct hlist_node *n = &ifa->addr_lst;
2999 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3000 if (net_eq(dev_net(ifa->idev->dev), net))
3003 while (++state->bucket < IN6_ADDR_HSIZE) {
3004 hlist_for_each_entry_rcu_bh(ifa, n,
3005 &inet6_addr_lst[state->bucket], addr_lst) {
3006 if (net_eq(dev_net(ifa->idev->dev), net))
3014 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3016 struct inet6_ifaddr *ifa = if6_get_first(seq);
3019 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3021 return pos ? NULL : ifa;
3024 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3028 return if6_get_idx(seq, *pos);
3031 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3033 struct inet6_ifaddr *ifa;
3035 ifa = if6_get_next(seq, v);
3040 static void if6_seq_stop(struct seq_file *seq, void *v)
3043 rcu_read_unlock_bh();
3046 static int if6_seq_show(struct seq_file *seq, void *v)
3048 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3049 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3051 ifp->idev->dev->ifindex,
3055 ifp->idev->dev->name);
3059 static const struct seq_operations if6_seq_ops = {
3060 .start = if6_seq_start,
3061 .next = if6_seq_next,
3062 .show = if6_seq_show,
3063 .stop = if6_seq_stop,
3066 static int if6_seq_open(struct inode *inode, struct file *file)
3068 return seq_open_net(inode, file, &if6_seq_ops,
3069 sizeof(struct if6_iter_state));
3072 static const struct file_operations if6_fops = {
3073 .owner = THIS_MODULE,
3074 .open = if6_seq_open,
3076 .llseek = seq_lseek,
3077 .release = seq_release_net,
3080 static int __net_init if6_proc_net_init(struct net *net)
3082 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3087 static void __net_exit if6_proc_net_exit(struct net *net)
3089 proc_net_remove(net, "if_inet6");
3092 static struct pernet_operations if6_proc_net_ops = {
3093 .init = if6_proc_net_init,
3094 .exit = if6_proc_net_exit,
3097 int __init if6_proc_init(void)
3099 return register_pernet_subsys(&if6_proc_net_ops);
3102 void if6_proc_exit(void)
3104 unregister_pernet_subsys(&if6_proc_net_ops);
3106 #endif /* CONFIG_PROC_FS */
3108 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3109 /* Check if address is a home address configured on any interface. */
3110 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3113 struct inet6_ifaddr *ifp = NULL;
3114 struct hlist_node *n;
3115 unsigned int hash = ipv6_addr_hash(addr);
3118 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3119 if (!net_eq(dev_net(ifp->idev->dev), net))
3121 if (ipv6_addr_equal(&ifp->addr, addr) &&
3122 (ifp->flags & IFA_F_HOMEADDRESS)) {
3127 rcu_read_unlock_bh();
3133 * Periodic address status verification
3136 static void addrconf_verify(unsigned long foo)
3138 unsigned long now, next, next_sec, next_sched;
3139 struct inet6_ifaddr *ifp;
3140 struct hlist_node *node;
3144 spin_lock(&addrconf_verify_lock);
3146 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3148 del_timer(&addr_chk_timer);
3150 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3152 hlist_for_each_entry_rcu_bh(ifp, node,
3153 &inet6_addr_lst[i], addr_lst) {
3156 if (ifp->flags & IFA_F_PERMANENT)
3159 spin_lock(&ifp->lock);
3160 /* We try to batch several events at once. */
3161 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3163 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3164 age >= ifp->valid_lft) {
3165 spin_unlock(&ifp->lock);
3169 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3170 spin_unlock(&ifp->lock);
3172 } else if (age >= ifp->prefered_lft) {
3173 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3176 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3178 ifp->flags |= IFA_F_DEPRECATED;
3181 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3182 next = ifp->tstamp + ifp->valid_lft * HZ;
3184 spin_unlock(&ifp->lock);
3189 ipv6_ifa_notify(0, ifp);
3193 #ifdef CONFIG_IPV6_PRIVACY
3194 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3195 !(ifp->flags&IFA_F_TENTATIVE)) {
3196 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3197 ifp->idev->cnf.dad_transmits *
3198 ifp->idev->nd_parms->retrans_time / HZ;
3200 if (age >= ifp->prefered_lft - regen_advance) {
3201 struct inet6_ifaddr *ifpub = ifp->ifpub;
3202 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3203 next = ifp->tstamp + ifp->prefered_lft * HZ;
3204 if (!ifp->regen_count && ifpub) {
3207 in6_ifa_hold(ifpub);
3208 spin_unlock(&ifp->lock);
3210 spin_lock(&ifpub->lock);
3211 ifpub->regen_count = 0;
3212 spin_unlock(&ifpub->lock);
3213 ipv6_create_tempaddr(ifpub, ifp);
3218 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3219 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3220 spin_unlock(&ifp->lock);
3223 /* ifp->prefered_lft <= ifp->valid_lft */
3224 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3225 next = ifp->tstamp + ifp->prefered_lft * HZ;
3226 spin_unlock(&ifp->lock);
3231 next_sec = round_jiffies_up(next);
3234 /* If rounded timeout is accurate enough, accept it. */
3235 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3236 next_sched = next_sec;
3238 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3239 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3240 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3242 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3243 now, next, next_sec, next_sched));
3245 addr_chk_timer.expires = next_sched;
3246 add_timer(&addr_chk_timer);
3247 spin_unlock(&addrconf_verify_lock);
3248 rcu_read_unlock_bh();
3251 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3253 struct in6_addr *pfx = NULL;
3256 pfx = nla_data(addr);
3259 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3262 pfx = nla_data(local);
3268 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3269 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3270 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3271 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3275 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3277 struct net *net = sock_net(skb->sk);
3278 struct ifaddrmsg *ifm;
3279 struct nlattr *tb[IFA_MAX+1];
3280 struct in6_addr *pfx;
3283 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3287 ifm = nlmsg_data(nlh);
3288 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3292 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3295 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3296 u32 prefered_lft, u32 valid_lft)
3300 unsigned long timeout;
3302 if (!valid_lft || (prefered_lft > valid_lft))
3305 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3306 if (addrconf_finite_timeout(timeout)) {
3307 expires = jiffies_to_clock_t(timeout * HZ);
3308 valid_lft = timeout;
3309 flags = RTF_EXPIRES;
3313 ifa_flags |= IFA_F_PERMANENT;
3316 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3317 if (addrconf_finite_timeout(timeout)) {
3319 ifa_flags |= IFA_F_DEPRECATED;
3320 prefered_lft = timeout;
3323 spin_lock_bh(&ifp->lock);
3324 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3325 ifp->tstamp = jiffies;
3326 ifp->valid_lft = valid_lft;
3327 ifp->prefered_lft = prefered_lft;
3329 spin_unlock_bh(&ifp->lock);
3330 if (!(ifp->flags&IFA_F_TENTATIVE))
3331 ipv6_ifa_notify(0, ifp);
3333 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3341 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3343 struct net *net = sock_net(skb->sk);
3344 struct ifaddrmsg *ifm;
3345 struct nlattr *tb[IFA_MAX+1];
3346 struct in6_addr *pfx;
3347 struct inet6_ifaddr *ifa;
3348 struct net_device *dev;
3349 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3353 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3357 ifm = nlmsg_data(nlh);
3358 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3362 if (tb[IFA_CACHEINFO]) {
3363 struct ifa_cacheinfo *ci;
3365 ci = nla_data(tb[IFA_CACHEINFO]);
3366 valid_lft = ci->ifa_valid;
3367 preferred_lft = ci->ifa_prefered;
3369 preferred_lft = INFINITY_LIFE_TIME;
3370 valid_lft = INFINITY_LIFE_TIME;
3373 dev = __dev_get_by_index(net, ifm->ifa_index);
3377 /* We ignore other flags so far. */
3378 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3380 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3383 * It would be best to check for !NLM_F_CREATE here but
3384 * userspace alreay relies on not having to provide this.
3386 return inet6_addr_add(net, ifm->ifa_index, pfx,
3387 ifm->ifa_prefixlen, ifa_flags,
3388 preferred_lft, valid_lft);
3391 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3392 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3395 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3402 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3403 u8 scope, int ifindex)
3405 struct ifaddrmsg *ifm;
3407 ifm = nlmsg_data(nlh);
3408 ifm->ifa_family = AF_INET6;
3409 ifm->ifa_prefixlen = prefixlen;
3410 ifm->ifa_flags = flags;
3411 ifm->ifa_scope = scope;
3412 ifm->ifa_index = ifindex;
3415 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3416 unsigned long tstamp, u32 preferred, u32 valid)
3418 struct ifa_cacheinfo ci;
3420 ci.cstamp = cstamp_delta(cstamp);
3421 ci.tstamp = cstamp_delta(tstamp);
3422 ci.ifa_prefered = preferred;
3423 ci.ifa_valid = valid;
3425 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3428 static inline int rt_scope(int ifa_scope)
3430 if (ifa_scope & IFA_HOST)
3431 return RT_SCOPE_HOST;
3432 else if (ifa_scope & IFA_LINK)
3433 return RT_SCOPE_LINK;
3434 else if (ifa_scope & IFA_SITE)
3435 return RT_SCOPE_SITE;
3437 return RT_SCOPE_UNIVERSE;
3440 static inline int inet6_ifaddr_msgsize(void)
3442 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3443 + nla_total_size(16) /* IFA_ADDRESS */
3444 + nla_total_size(sizeof(struct ifa_cacheinfo));
3447 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3448 u32 pid, u32 seq, int event, unsigned int flags)
3450 struct nlmsghdr *nlh;
3451 u32 preferred, valid;
3453 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3457 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3458 ifa->idev->dev->ifindex);
3460 if (!(ifa->flags&IFA_F_PERMANENT)) {
3461 preferred = ifa->prefered_lft;
3462 valid = ifa->valid_lft;
3463 if (preferred != INFINITY_LIFE_TIME) {
3464 long tval = (jiffies - ifa->tstamp)/HZ;
3465 if (preferred > tval)
3469 if (valid != INFINITY_LIFE_TIME) {
3477 preferred = INFINITY_LIFE_TIME;
3478 valid = INFINITY_LIFE_TIME;
3481 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3482 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3483 nlmsg_cancel(skb, nlh);
3487 return nlmsg_end(skb, nlh);
3490 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3491 u32 pid, u32 seq, int event, u16 flags)
3493 struct nlmsghdr *nlh;
3494 u8 scope = RT_SCOPE_UNIVERSE;
3495 int ifindex = ifmca->idev->dev->ifindex;
3497 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3498 scope = RT_SCOPE_SITE;
3500 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3504 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3505 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3506 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3507 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3508 nlmsg_cancel(skb, nlh);
3512 return nlmsg_end(skb, nlh);
3515 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3516 u32 pid, u32 seq, int event, unsigned int flags)
3518 struct nlmsghdr *nlh;
3519 u8 scope = RT_SCOPE_UNIVERSE;
3520 int ifindex = ifaca->aca_idev->dev->ifindex;
3522 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3523 scope = RT_SCOPE_SITE;
3525 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3529 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3530 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3531 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3532 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3533 nlmsg_cancel(skb, nlh);
3537 return nlmsg_end(skb, nlh);
3546 /* called with rcu_read_lock() */
3547 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3548 struct netlink_callback *cb, enum addr_type_t type,
3549 int s_ip_idx, int *p_ip_idx)
3551 struct ifmcaddr6 *ifmca;
3552 struct ifacaddr6 *ifaca;
3554 int ip_idx = *p_ip_idx;
3556 read_lock_bh(&idev->lock);
3558 case UNICAST_ADDR: {
3559 struct inet6_ifaddr *ifa;
3561 /* unicast address incl. temp addr */
3562 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3563 if (++ip_idx < s_ip_idx)
3565 err = inet6_fill_ifaddr(skb, ifa,
3566 NETLINK_CB(cb->skb).pid,
3575 case MULTICAST_ADDR:
3576 /* multicast address */
3577 for (ifmca = idev->mc_list; ifmca;
3578 ifmca = ifmca->next, ip_idx++) {
3579 if (ip_idx < s_ip_idx)
3581 err = inet6_fill_ifmcaddr(skb, ifmca,
3582 NETLINK_CB(cb->skb).pid,
3591 /* anycast address */
3592 for (ifaca = idev->ac_list; ifaca;
3593 ifaca = ifaca->aca_next, ip_idx++) {
3594 if (ip_idx < s_ip_idx)
3596 err = inet6_fill_ifacaddr(skb, ifaca,
3597 NETLINK_CB(cb->skb).pid,
3608 read_unlock_bh(&idev->lock);
3613 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3614 enum addr_type_t type)
3616 struct net *net = sock_net(skb->sk);
3619 int s_idx, s_ip_idx;
3620 struct net_device *dev;
3621 struct inet6_dev *idev;
3622 struct hlist_head *head;
3623 struct hlist_node *node;
3626 s_idx = idx = cb->args[1];
3627 s_ip_idx = ip_idx = cb->args[2];
3630 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3632 head = &net->dev_index_head[h];
3633 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3636 if (h > s_h || idx > s_idx)
3639 idev = __in6_dev_get(dev);
3643 if (in6_dump_addrs(idev, skb, cb, type,
3644 s_ip_idx, &ip_idx) <= 0)
3654 cb->args[2] = ip_idx;
3659 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3661 enum addr_type_t type = UNICAST_ADDR;
3663 return inet6_dump_addr(skb, cb, type);
3666 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3668 enum addr_type_t type = MULTICAST_ADDR;
3670 return inet6_dump_addr(skb, cb, type);
3674 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3676 enum addr_type_t type = ANYCAST_ADDR;
3678 return inet6_dump_addr(skb, cb, type);
3681 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3684 struct net *net = sock_net(in_skb->sk);
3685 struct ifaddrmsg *ifm;
3686 struct nlattr *tb[IFA_MAX+1];
3687 struct in6_addr *addr = NULL;
3688 struct net_device *dev = NULL;
3689 struct inet6_ifaddr *ifa;
3690 struct sk_buff *skb;
3693 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3697 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3703 ifm = nlmsg_data(nlh);
3705 dev = __dev_get_by_index(net, ifm->ifa_index);
3707 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3709 err = -EADDRNOTAVAIL;
3713 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3719 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3720 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3722 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3723 WARN_ON(err == -EMSGSIZE);
3727 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3734 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3736 struct sk_buff *skb;
3737 struct net *net = dev_net(ifa->idev->dev);
3740 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3744 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3746 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3747 WARN_ON(err == -EMSGSIZE);
3751 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3755 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3758 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3759 __s32 *array, int bytes)
3761 BUG_ON(bytes < (DEVCONF_MAX * 4));
3763 memset(array, 0, bytes);
3764 array[DEVCONF_FORWARDING] = cnf->forwarding;
3765 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3766 array[DEVCONF_MTU6] = cnf->mtu6;
3767 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3768 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3769 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3770 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3771 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3772 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3773 jiffies_to_msecs(cnf->rtr_solicit_interval);
3774 array[DEVCONF_RTR_SOLICIT_DELAY] =
3775 jiffies_to_msecs(cnf->rtr_solicit_delay);
3776 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3777 #ifdef CONFIG_IPV6_PRIVACY
3778 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3779 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3780 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3781 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3782 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3784 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3785 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3786 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3787 #ifdef CONFIG_IPV6_ROUTER_PREF
3788 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3789 array[DEVCONF_RTR_PROBE_INTERVAL] =
3790 jiffies_to_msecs(cnf->rtr_probe_interval);
3791 #ifdef CONFIG_IPV6_ROUTE_INFO
3792 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3795 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3796 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3797 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3798 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3800 #ifdef CONFIG_IPV6_MROUTE
3801 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3803 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3804 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3805 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3808 static inline size_t inet6_ifla6_size(void)
3810 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3811 + nla_total_size(sizeof(struct ifla_cacheinfo))
3812 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3813 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3814 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3817 static inline size_t inet6_if_nlmsg_size(void)
3819 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3820 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3821 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3822 + nla_total_size(4) /* IFLA_MTU */
3823 + nla_total_size(4) /* IFLA_LINK */
3824 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3827 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3828 int items, int bytes)
3831 int pad = bytes - sizeof(u64) * items;
3834 /* Use put_unaligned() because stats may not be aligned for u64. */
3835 put_unaligned(items, &stats[0]);
3836 for (i = 1; i < items; i++)
3837 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3839 memset(&stats[items], 0, pad);
3842 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3843 int items, int bytes, size_t syncpoff)
3846 int pad = bytes - sizeof(u64) * items;
3849 /* Use put_unaligned() because stats may not be aligned for u64. */
3850 put_unaligned(items, &stats[0]);
3851 for (i = 1; i < items; i++)
3852 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3854 memset(&stats[items], 0, pad);
3857 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3861 case IFLA_INET6_STATS:
3862 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3863 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3865 case IFLA_INET6_ICMP6STATS:
3866 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3871 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3874 struct ifla_cacheinfo ci;
3876 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3878 ci.max_reasm_len = IPV6_MAXPLEN;
3879 ci.tstamp = cstamp_delta(idev->tstamp);
3880 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3881 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3882 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3884 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3886 goto nla_put_failure;
3887 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3889 /* XXX - MC not implemented */
3891 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3893 goto nla_put_failure;
3894 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3896 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3898 goto nla_put_failure;
3899 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3907 static size_t inet6_get_link_af_size(const struct net_device *dev)
3909 if (!__in6_dev_get(dev))
3912 return inet6_ifla6_size();
3915 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3917 struct inet6_dev *idev = __in6_dev_get(dev);
3922 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3928 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3929 u32 pid, u32 seq, int event, unsigned int flags)
3931 struct net_device *dev = idev->dev;
3932 struct ifinfomsg *hdr;
3933 struct nlmsghdr *nlh;
3936 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3940 hdr = nlmsg_data(nlh);
3941 hdr->ifi_family = AF_INET6;
3943 hdr->ifi_type = dev->type;
3944 hdr->ifi_index = dev->ifindex;
3945 hdr->ifi_flags = dev_get_flags(dev);
3946 hdr->ifi_change = 0;
3948 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3951 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3953 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3954 if (dev->ifindex != dev->iflink)
3955 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3957 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3958 if (protoinfo == NULL)
3959 goto nla_put_failure;
3961 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3962 goto nla_put_failure;
3964 nla_nest_end(skb, protoinfo);
3965 return nlmsg_end(skb, nlh);
3968 nlmsg_cancel(skb, nlh);
3972 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3974 struct net *net = sock_net(skb->sk);
3977 struct net_device *dev;
3978 struct inet6_dev *idev;
3979 struct hlist_head *head;
3980 struct hlist_node *node;
3983 s_idx = cb->args[1];
3986 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3988 head = &net->dev_index_head[h];
3989 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3992 idev = __in6_dev_get(dev);
3995 if (inet6_fill_ifinfo(skb, idev,
3996 NETLINK_CB(cb->skb).pid,
3998 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4012 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4014 struct sk_buff *skb;
4015 struct net *net = dev_net(idev->dev);
4018 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4022 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4024 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4025 WARN_ON(err == -EMSGSIZE);
4029 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4033 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4036 static inline size_t inet6_prefix_nlmsg_size(void)
4038 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4039 + nla_total_size(sizeof(struct in6_addr))
4040 + nla_total_size(sizeof(struct prefix_cacheinfo));
4043 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4044 struct prefix_info *pinfo, u32 pid, u32 seq,
4045 int event, unsigned int flags)
4047 struct prefixmsg *pmsg;
4048 struct nlmsghdr *nlh;
4049 struct prefix_cacheinfo ci;
4051 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4055 pmsg = nlmsg_data(nlh);
4056 pmsg->prefix_family = AF_INET6;
4057 pmsg->prefix_pad1 = 0;
4058 pmsg->prefix_pad2 = 0;
4059 pmsg->prefix_ifindex = idev->dev->ifindex;
4060 pmsg->prefix_len = pinfo->prefix_len;
4061 pmsg->prefix_type = pinfo->type;
4062 pmsg->prefix_pad3 = 0;
4063 pmsg->prefix_flags = 0;
4065 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4066 if (pinfo->autoconf)
4067 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4069 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4071 ci.preferred_time = ntohl(pinfo->prefered);
4072 ci.valid_time = ntohl(pinfo->valid);
4073 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4075 return nlmsg_end(skb, nlh);
4078 nlmsg_cancel(skb, nlh);
4082 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4083 struct prefix_info *pinfo)
4085 struct sk_buff *skb;
4086 struct net *net = dev_net(idev->dev);
4089 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4093 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4095 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4096 WARN_ON(err == -EMSGSIZE);
4100 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4104 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4107 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4109 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4114 * If the address was optimistic
4115 * we inserted the route at the start of
4116 * our DAD process, so we don't need
4119 if (!(ifp->rt->rt6i_node))
4120 ip6_ins_rt(ifp->rt);
4121 if (ifp->idev->cnf.forwarding)
4122 addrconf_join_anycast(ifp);
4125 if (ifp->idev->cnf.forwarding)
4126 addrconf_leave_anycast(ifp);
4127 addrconf_leave_solict(ifp->idev, &ifp->addr);
4128 dst_hold(&ifp->rt->dst);
4130 if (ip6_del_rt(ifp->rt))
4131 dst_free(&ifp->rt->dst);
4136 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4139 if (likely(ifp->idev->dead == 0))
4140 __ipv6_ifa_notify(event, ifp);
4141 rcu_read_unlock_bh();
4144 #ifdef CONFIG_SYSCTL
4147 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4148 void __user *buffer, size_t *lenp, loff_t *ppos)
4150 int *valp = ctl->data;
4155 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4158 ret = addrconf_fixup_forwarding(ctl, valp, val);
4164 static void dev_disable_change(struct inet6_dev *idev)
4166 if (!idev || !idev->dev)
4169 if (idev->cnf.disable_ipv6)
4170 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4172 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4175 static void addrconf_disable_change(struct net *net, __s32 newf)
4177 struct net_device *dev;
4178 struct inet6_dev *idev;
4181 for_each_netdev_rcu(net, dev) {
4182 idev = __in6_dev_get(dev);
4184 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4185 idev->cnf.disable_ipv6 = newf;
4187 dev_disable_change(idev);
4193 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4197 net = (struct net *)table->extra2;
4199 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4202 if (!rtnl_trylock()) {
4203 /* Restore the original values before restarting */
4205 return restart_syscall();
4208 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4209 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4210 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4211 addrconf_disable_change(net, newf);
4212 } else if ((!*p) ^ (!old))
4213 dev_disable_change((struct inet6_dev *)table->extra1);
4220 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4221 void __user *buffer, size_t *lenp, loff_t *ppos)
4223 int *valp = ctl->data;
4228 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4231 ret = addrconf_disable_ipv6(ctl, valp, val);
4237 static struct addrconf_sysctl_table
4239 struct ctl_table_header *sysctl_header;
4240 ctl_table addrconf_vars[DEVCONF_MAX+1];
4242 } addrconf_sysctl __read_mostly = {
4243 .sysctl_header = NULL,
4246 .procname = "forwarding",
4247 .data = &ipv6_devconf.forwarding,
4248 .maxlen = sizeof(int),
4250 .proc_handler = addrconf_sysctl_forward,
4253 .procname = "hop_limit",
4254 .data = &ipv6_devconf.hop_limit,
4255 .maxlen = sizeof(int),
4257 .proc_handler = proc_dointvec,
4261 .data = &ipv6_devconf.mtu6,
4262 .maxlen = sizeof(int),
4264 .proc_handler = proc_dointvec,
4267 .procname = "accept_ra",
4268 .data = &ipv6_devconf.accept_ra,
4269 .maxlen = sizeof(int),
4271 .proc_handler = proc_dointvec,
4274 .procname = "accept_redirects",
4275 .data = &ipv6_devconf.accept_redirects,
4276 .maxlen = sizeof(int),
4278 .proc_handler = proc_dointvec,
4281 .procname = "autoconf",
4282 .data = &ipv6_devconf.autoconf,
4283 .maxlen = sizeof(int),
4285 .proc_handler = proc_dointvec,
4288 .procname = "dad_transmits",
4289 .data = &ipv6_devconf.dad_transmits,
4290 .maxlen = sizeof(int),
4292 .proc_handler = proc_dointvec,
4295 .procname = "router_solicitations",
4296 .data = &ipv6_devconf.rtr_solicits,
4297 .maxlen = sizeof(int),
4299 .proc_handler = proc_dointvec,
4302 .procname = "router_solicitation_interval",
4303 .data = &ipv6_devconf.rtr_solicit_interval,
4304 .maxlen = sizeof(int),
4306 .proc_handler = proc_dointvec_jiffies,
4309 .procname = "router_solicitation_delay",
4310 .data = &ipv6_devconf.rtr_solicit_delay,
4311 .maxlen = sizeof(int),
4313 .proc_handler = proc_dointvec_jiffies,
4316 .procname = "force_mld_version",
4317 .data = &ipv6_devconf.force_mld_version,
4318 .maxlen = sizeof(int),
4320 .proc_handler = proc_dointvec,
4322 #ifdef CONFIG_IPV6_PRIVACY
4324 .procname = "use_tempaddr",
4325 .data = &ipv6_devconf.use_tempaddr,
4326 .maxlen = sizeof(int),
4328 .proc_handler = proc_dointvec,
4331 .procname = "temp_valid_lft",
4332 .data = &ipv6_devconf.temp_valid_lft,
4333 .maxlen = sizeof(int),
4335 .proc_handler = proc_dointvec,
4338 .procname = "temp_prefered_lft",
4339 .data = &ipv6_devconf.temp_prefered_lft,
4340 .maxlen = sizeof(int),
4342 .proc_handler = proc_dointvec,
4345 .procname = "regen_max_retry",
4346 .data = &ipv6_devconf.regen_max_retry,
4347 .maxlen = sizeof(int),
4349 .proc_handler = proc_dointvec,
4352 .procname = "max_desync_factor",
4353 .data = &ipv6_devconf.max_desync_factor,
4354 .maxlen = sizeof(int),
4356 .proc_handler = proc_dointvec,
4360 .procname = "max_addresses",
4361 .data = &ipv6_devconf.max_addresses,
4362 .maxlen = sizeof(int),
4364 .proc_handler = proc_dointvec,
4367 .procname = "accept_ra_defrtr",
4368 .data = &ipv6_devconf.accept_ra_defrtr,
4369 .maxlen = sizeof(int),
4371 .proc_handler = proc_dointvec,
4374 .procname = "accept_ra_pinfo",
4375 .data = &ipv6_devconf.accept_ra_pinfo,
4376 .maxlen = sizeof(int),
4378 .proc_handler = proc_dointvec,
4380 #ifdef CONFIG_IPV6_ROUTER_PREF
4382 .procname = "accept_ra_rtr_pref",
4383 .data = &ipv6_devconf.accept_ra_rtr_pref,
4384 .maxlen = sizeof(int),
4386 .proc_handler = proc_dointvec,
4389 .procname = "router_probe_interval",
4390 .data = &ipv6_devconf.rtr_probe_interval,
4391 .maxlen = sizeof(int),
4393 .proc_handler = proc_dointvec_jiffies,
4395 #ifdef CONFIG_IPV6_ROUTE_INFO
4397 .procname = "accept_ra_rt_info_max_plen",
4398 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4399 .maxlen = sizeof(int),
4401 .proc_handler = proc_dointvec,
4406 .procname = "proxy_ndp",
4407 .data = &ipv6_devconf.proxy_ndp,
4408 .maxlen = sizeof(int),
4410 .proc_handler = proc_dointvec,
4413 .procname = "accept_source_route",
4414 .data = &ipv6_devconf.accept_source_route,
4415 .maxlen = sizeof(int),
4417 .proc_handler = proc_dointvec,
4419 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4421 .procname = "optimistic_dad",
4422 .data = &ipv6_devconf.optimistic_dad,
4423 .maxlen = sizeof(int),
4425 .proc_handler = proc_dointvec,
4429 #ifdef CONFIG_IPV6_MROUTE
4431 .procname = "mc_forwarding",
4432 .data = &ipv6_devconf.mc_forwarding,
4433 .maxlen = sizeof(int),
4435 .proc_handler = proc_dointvec,
4439 .procname = "disable_ipv6",
4440 .data = &ipv6_devconf.disable_ipv6,
4441 .maxlen = sizeof(int),
4443 .proc_handler = addrconf_sysctl_disable,
4446 .procname = "accept_dad",
4447 .data = &ipv6_devconf.accept_dad,
4448 .maxlen = sizeof(int),
4450 .proc_handler = proc_dointvec,
4453 .procname = "force_tllao",
4454 .data = &ipv6_devconf.force_tllao,
4455 .maxlen = sizeof(int),
4457 .proc_handler = proc_dointvec
4465 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4466 struct inet6_dev *idev, struct ipv6_devconf *p)
4469 struct addrconf_sysctl_table *t;
4471 #define ADDRCONF_CTL_PATH_DEV 3
4473 struct ctl_path addrconf_ctl_path[] = {
4474 { .procname = "net", },
4475 { .procname = "ipv6", },
4476 { .procname = "conf", },
4477 { /* to be set */ },
4482 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4486 for (i = 0; t->addrconf_vars[i].data; i++) {
4487 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4488 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4489 t->addrconf_vars[i].extra2 = net;
4493 * Make a copy of dev_name, because '.procname' is regarded as const
4494 * by sysctl and we wouldn't want anyone to change it under our feet
4495 * (see SIOCSIFNAME).
4497 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4501 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4503 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4505 if (t->sysctl_header == NULL)
4519 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4521 struct addrconf_sysctl_table *t;
4523 if (p->sysctl == NULL)
4528 unregister_net_sysctl_table(t->sysctl_header);
4533 static void addrconf_sysctl_register(struct inet6_dev *idev)
4535 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4536 &ndisc_ifinfo_sysctl_change);
4537 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4541 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4543 __addrconf_sysctl_unregister(&idev->cnf);
4544 neigh_sysctl_unregister(idev->nd_parms);
4550 static int __net_init addrconf_init_net(struct net *net)
4553 struct ipv6_devconf *all, *dflt;
4556 all = &ipv6_devconf;
4557 dflt = &ipv6_devconf_dflt;
4559 if (!net_eq(net, &init_net)) {
4560 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4564 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4566 goto err_alloc_dflt;
4568 /* these will be inherited by all namespaces */
4569 dflt->autoconf = ipv6_defaults.autoconf;
4570 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4573 net->ipv6.devconf_all = all;
4574 net->ipv6.devconf_dflt = dflt;
4576 #ifdef CONFIG_SYSCTL
4577 err = __addrconf_sysctl_register(net, "all", NULL, all);
4581 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4587 #ifdef CONFIG_SYSCTL
4589 __addrconf_sysctl_unregister(all);
4599 static void __net_exit addrconf_exit_net(struct net *net)
4601 #ifdef CONFIG_SYSCTL
4602 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4603 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4605 if (!net_eq(net, &init_net)) {
4606 kfree(net->ipv6.devconf_dflt);
4607 kfree(net->ipv6.devconf_all);
4611 static struct pernet_operations addrconf_ops = {
4612 .init = addrconf_init_net,
4613 .exit = addrconf_exit_net,
4620 int register_inet6addr_notifier(struct notifier_block *nb)
4622 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4624 EXPORT_SYMBOL(register_inet6addr_notifier);
4626 int unregister_inet6addr_notifier(struct notifier_block *nb)
4628 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4630 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4632 static struct rtnl_af_ops inet6_ops = {
4634 .fill_link_af = inet6_fill_link_af,
4635 .get_link_af_size = inet6_get_link_af_size,
4639 * Init / cleanup code
4642 int __init addrconf_init(void)
4646 err = ipv6_addr_label_init();
4648 printk(KERN_CRIT "IPv6 Addrconf:"
4649 " cannot initialize default policy table: %d.\n", err);
4653 err = register_pernet_subsys(&addrconf_ops);
4657 /* The addrconf netdev notifier requires that loopback_dev
4658 * has it's ipv6 private information allocated and setup
4659 * before it can bring up and give link-local addresses
4660 * to other devices which are up.
4662 * Unfortunately, loopback_dev is not necessarily the first
4663 * entry in the global dev_base list of net devices. In fact,
4664 * it is likely to be the very last entry on that list.
4665 * So this causes the notifier registry below to try and
4666 * give link-local addresses to all devices besides loopback_dev
4667 * first, then loopback_dev, which cases all the non-loopback_dev
4668 * devices to fail to get a link-local address.
4670 * So, as a temporary fix, allocate the ipv6 structure for
4671 * loopback_dev first by hand.
4672 * Longer term, all of the dependencies ipv6 has upon the loopback
4673 * device and it being up should be removed.
4676 if (!ipv6_add_dev(init_net.loopback_dev))
4682 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4683 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4685 register_netdevice_notifier(&ipv6_dev_notf);
4689 err = rtnl_af_register(&inet6_ops);
4693 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4697 /* Only the first call to __rtnl_register can fail */
4698 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4699 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4700 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4701 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4702 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4704 ipv6_addr_label_rtnl_register();
4708 rtnl_af_unregister(&inet6_ops);
4710 unregister_netdevice_notifier(&ipv6_dev_notf);
4712 unregister_pernet_subsys(&addrconf_ops);
4714 ipv6_addr_label_cleanup();
4719 void addrconf_cleanup(void)
4721 struct net_device *dev;
4724 unregister_netdevice_notifier(&ipv6_dev_notf);
4725 unregister_pernet_subsys(&addrconf_ops);
4726 ipv6_addr_label_cleanup();
4730 __rtnl_af_unregister(&inet6_ops);
4732 /* clean dev list */
4733 for_each_netdev(&init_net, dev) {
4734 if (__in6_dev_get(dev) == NULL)
4736 addrconf_ifdown(dev, 1);
4738 addrconf_ifdown(init_net.loopback_dev, 2);
4743 spin_lock_bh(&addrconf_hash_lock);
4744 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4745 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4746 spin_unlock_bh(&addrconf_hash_lock);
4748 del_timer(&addr_chk_timer);