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
101 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
103 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
104 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
105 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
108 static void addrconf_sysctl_register(struct inet6_dev *idev);
109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
120 #ifdef CONFIG_IPV6_PRIVACY
121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
123 static void ipv6_regen_rndid(unsigned long data);
125 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
129 static int ipv6_count_addresses(struct inet6_dev *idev);
132 * Configured unicast address hash table
134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
135 static DEFINE_SPINLOCK(addrconf_hash_lock);
137 static void addrconf_verify(unsigned long);
139 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
140 static DEFINE_SPINLOCK(addrconf_verify_lock);
142 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
143 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_type_change(struct net_device *dev,
146 unsigned long event);
147 static int addrconf_ifdown(struct net_device *dev, int how);
149 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
150 static void addrconf_dad_timer(unsigned long data);
151 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
152 static void addrconf_dad_run(struct inet6_dev *idev);
153 static void addrconf_rs_timer(unsigned long data);
154 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
155 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
157 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
158 struct prefix_info *pinfo);
159 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
160 struct net_device *dev);
162 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
164 static struct ipv6_devconf ipv6_devconf __read_mostly = {
166 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
167 .mtu6 = IPV6_MIN_MTU,
169 .accept_redirects = 1,
171 .force_mld_version = 0,
173 .rtr_solicits = MAX_RTR_SOLICITATIONS,
174 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
175 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
176 #ifdef CONFIG_IPV6_PRIVACY
178 .temp_valid_lft = TEMP_VALID_LIFETIME,
179 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
180 .regen_max_retry = REGEN_MAX_RETRY,
181 .max_desync_factor = MAX_DESYNC_FACTOR,
183 .max_addresses = IPV6_MAX_ADDRESSES,
184 .accept_ra_defrtr = 1,
185 .accept_ra_pinfo = 1,
186 #ifdef CONFIG_IPV6_ROUTER_PREF
187 .accept_ra_rtr_pref = 1,
188 .rtr_probe_interval = 60 * HZ,
189 #ifdef CONFIG_IPV6_ROUTE_INFO
190 .accept_ra_rt_info_max_plen = 0,
194 .accept_source_route = 0, /* we do not accept RH0 by default. */
199 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
201 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
202 .mtu6 = IPV6_MIN_MTU,
204 .accept_redirects = 1,
207 .rtr_solicits = MAX_RTR_SOLICITATIONS,
208 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
209 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
210 #ifdef CONFIG_IPV6_PRIVACY
212 .temp_valid_lft = TEMP_VALID_LIFETIME,
213 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
214 .regen_max_retry = REGEN_MAX_RETRY,
215 .max_desync_factor = MAX_DESYNC_FACTOR,
217 .max_addresses = IPV6_MAX_ADDRESSES,
218 .accept_ra_defrtr = 1,
219 .accept_ra_pinfo = 1,
220 #ifdef CONFIG_IPV6_ROUTER_PREF
221 .accept_ra_rtr_pref = 1,
222 .rtr_probe_interval = 60 * HZ,
223 #ifdef CONFIG_IPV6_ROUTE_INFO
224 .accept_ra_rt_info_max_plen = 0,
228 .accept_source_route = 0, /* we do not accept RH0 by default. */
233 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
234 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
235 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
236 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
237 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
239 /* Check if a valid qdisc is available */
240 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
242 return !qdisc_tx_is_noop(dev);
245 /* Check if a route is valid prefix route */
246 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
248 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
251 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
253 if (del_timer(&ifp->timer))
257 enum addrconf_timer_t {
263 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
264 enum addrconf_timer_t what,
267 if (!del_timer(&ifp->timer))
272 ifp->timer.function = addrconf_dad_timer;
275 ifp->timer.function = addrconf_rs_timer;
280 ifp->timer.expires = jiffies + when;
281 add_timer(&ifp->timer);
284 static int snmp6_alloc_dev(struct inet6_dev *idev)
286 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
287 sizeof(struct ipstats_mib)) < 0)
289 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
290 sizeof(struct icmpv6_mib)) < 0)
292 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
293 sizeof(struct icmpv6msg_mib)) < 0)
299 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
301 snmp_mib_free((void __percpu **)idev->stats.ipv6);
306 static void snmp6_free_dev(struct inet6_dev *idev)
308 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
309 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
310 snmp_mib_free((void __percpu **)idev->stats.ipv6);
313 /* Nobody refers to this device, we may destroy it. */
315 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
317 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
323 struct net_device *dev = idev->dev;
325 WARN_ON(!list_empty(&idev->addr_list));
326 WARN_ON(idev->mc_list != NULL);
328 #ifdef NET_REFCNT_DEBUG
329 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
333 pr_warning("Freeing alive inet6 device %p\n", idev);
336 snmp6_free_dev(idev);
337 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 struct inet6_dev *ndev;
348 if (dev->mtu < IPV6_MIN_MTU)
351 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
356 rwlock_init(&ndev->lock);
358 INIT_LIST_HEAD(&ndev->addr_list);
360 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (ndev->nd_parms == NULL) {
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
373 if (snmp6_alloc_dev(ndev) < 0) {
375 "%s(): cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name));
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
379 in6_dev_finish_destroy(ndev);
383 if (snmp6_register_dev(ndev) < 0) {
385 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name));
387 neigh_parms_release(&nd_tbl, ndev->nd_parms);
389 in6_dev_finish_destroy(ndev);
393 /* One reference from device. We must do this before
394 * we invoke __ipv6_regen_rndid().
398 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399 ndev->cnf.accept_dad = -1;
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
404 "%s: Disabled Multicast RS\n",
406 ndev->cnf.rtr_solicits = 0;
410 #ifdef CONFIG_IPV6_PRIVACY
411 INIT_LIST_HEAD(&ndev->tempaddr_list);
412 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 if ((dev->flags&IFF_LOOPBACK) ||
414 dev->type == ARPHRD_TUNNEL ||
415 dev->type == ARPHRD_TUNNEL6 ||
416 dev->type == ARPHRD_SIT ||
417 dev->type == ARPHRD_NONE) {
419 "%s: Disabled Privacy Extensions\n",
421 ndev->cnf.use_tempaddr = -1;
424 ipv6_regen_rndid((unsigned long) ndev);
428 if (netif_running(dev) && addrconf_qdisc_ok(dev))
429 ndev->if_flags |= IF_READY;
431 ipv6_mc_init_dev(ndev);
432 ndev->tstamp = jiffies;
433 addrconf_sysctl_register(ndev);
434 /* protected by rtnl_lock */
435 rcu_assign_pointer(dev->ip6_ptr, ndev);
437 /* Join all-node multicast group */
438 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
443 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
445 struct inet6_dev *idev;
449 idev = __in6_dev_get(dev);
451 idev = ipv6_add_dev(dev);
456 if (dev->flags&IFF_UP)
462 static void dev_forward_change(struct inet6_dev *idev)
464 struct net_device *dev;
465 struct inet6_ifaddr *ifa;
470 if (idev->cnf.forwarding)
471 dev_disable_lro(dev);
472 if (dev && (dev->flags & IFF_MULTICAST)) {
473 if (idev->cnf.forwarding)
474 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
476 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
479 list_for_each_entry(ifa, &idev->addr_list, if_list) {
480 if (ifa->flags&IFA_F_TENTATIVE)
482 if (idev->cnf.forwarding)
483 addrconf_join_anycast(ifa);
485 addrconf_leave_anycast(ifa);
490 static void addrconf_forward_change(struct net *net, __s32 newf)
492 struct net_device *dev;
493 struct inet6_dev *idev;
496 for_each_netdev_rcu(net, dev) {
497 idev = __in6_dev_get(dev);
499 int changed = (!idev->cnf.forwarding) ^ (!newf);
500 idev->cnf.forwarding = newf;
502 dev_forward_change(idev);
508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
512 net = (struct net *)table->extra2;
513 if (p == &net->ipv6.devconf_dflt->forwarding)
516 if (!rtnl_trylock()) {
517 /* Restore the original values before restarting */
519 return restart_syscall();
522 if (p == &net->ipv6.devconf_all->forwarding) {
523 __s32 newf = net->ipv6.devconf_all->forwarding;
524 net->ipv6.devconf_dflt->forwarding = newf;
525 addrconf_forward_change(net, newf);
526 } else if ((!*p) ^ (!old))
527 dev_forward_change((struct inet6_dev *)table->extra1);
531 rt6_purge_dflt_routers(net);
536 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
538 struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
542 /* Nobody refers to this ifaddr, destroy it */
543 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
545 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
547 #ifdef NET_REFCNT_DEBUG
548 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
551 in6_dev_put(ifp->idev);
553 if (del_timer(&ifp->timer))
554 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
557 pr_warning("Freeing alive inet6 address %p\n", ifp);
560 dst_release(&ifp->rt->u.dst);
562 call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
566 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
569 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
572 * Each device address list is sorted in order of scope -
573 * global before linklocal.
575 list_for_each(p, &idev->addr_list) {
576 struct inet6_ifaddr *ifa
577 = list_entry(p, struct inet6_ifaddr, if_list);
578 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
582 list_add_tail(&ifp->if_list, p);
585 static u32 ipv6_addr_hash(const struct in6_addr *addr)
588 * We perform the hash function over the last 64 bits of the address
589 * This will include the IEEE address token on links that support it.
591 return jhash_2words(addr->s6_addr32[2], addr->s6_addr32[3], 0)
592 & (IN6_ADDR_HSIZE - 1);
595 /* On success it returns ifp with increased reference count */
597 static struct inet6_ifaddr *
598 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
599 int scope, u32 flags)
601 struct inet6_ifaddr *ifa = NULL;
605 int addr_type = ipv6_addr_type(addr);
607 if (addr_type == IPV6_ADDR_ANY ||
608 addr_type & IPV6_ADDR_MULTICAST ||
609 (!(idev->dev->flags & IFF_LOOPBACK) &&
610 addr_type & IPV6_ADDR_LOOPBACK))
611 return ERR_PTR(-EADDRNOTAVAIL);
615 err = -ENODEV; /*XXX*/
619 if (idev->cnf.disable_ipv6) {
624 spin_lock(&addrconf_hash_lock);
626 /* Ignore adding duplicate addresses on an interface */
627 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
628 ADBG(("ipv6_add_addr: already assigned\n"));
633 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
636 ADBG(("ipv6_add_addr: malloc failed\n"));
641 rt = addrconf_dst_alloc(idev, addr, 0);
647 ipv6_addr_copy(&ifa->addr, addr);
649 spin_lock_init(&ifa->lock);
650 init_timer(&ifa->timer);
651 INIT_HLIST_NODE(&ifa->addr_lst);
652 ifa->timer.data = (unsigned long) ifa;
654 ifa->prefix_len = pfxlen;
655 ifa->flags = flags | IFA_F_TENTATIVE;
656 ifa->cstamp = ifa->tstamp = jiffies;
661 * part one of RFC 4429, section 3.3
662 * We should not configure an address as
663 * optimistic if we do not yet know the link
664 * layer address of our nexhop router
667 if (rt->rt6i_nexthop == NULL)
668 ifa->flags &= ~IFA_F_OPTIMISTIC;
675 /* Add to big hash table */
676 hash = ipv6_addr_hash(addr);
678 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
680 spin_unlock(&addrconf_hash_lock);
682 write_lock(&idev->lock);
683 /* Add to inet6_dev unicast addr list. */
684 ipv6_link_dev_addr(idev, ifa);
686 #ifdef CONFIG_IPV6_PRIVACY
687 if (ifa->flags&IFA_F_TEMPORARY) {
688 list_add(&ifa->tmp_list, &idev->tempaddr_list);
694 write_unlock(&idev->lock);
696 rcu_read_unlock_bh();
698 if (likely(err == 0))
699 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
707 spin_unlock(&addrconf_hash_lock);
711 /* This function wants to get referenced ifp and releases it before return */
713 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
715 struct inet6_ifaddr *ifa, *ifn;
716 struct inet6_dev *idev = ifp->idev;
718 int deleted = 0, onlink = 0;
719 unsigned long expires = jiffies;
721 hash = ipv6_addr_hash(&ifp->addr);
725 spin_lock_bh(&addrconf_hash_lock);
726 hlist_del_init_rcu(&ifp->addr_lst);
728 spin_unlock_bh(&addrconf_hash_lock);
730 write_lock_bh(&idev->lock);
731 #ifdef CONFIG_IPV6_PRIVACY
732 if (ifp->flags&IFA_F_TEMPORARY) {
733 list_del(&ifp->tmp_list);
735 in6_ifa_put(ifp->ifpub);
742 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
744 list_del_init(&ifp->if_list);
747 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
751 } else if (ifp->flags & IFA_F_PERMANENT) {
752 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
754 if (ifa->flags & IFA_F_PERMANENT) {
759 unsigned long lifetime;
764 spin_lock(&ifa->lock);
766 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
768 * Note: Because this address is
769 * not permanent, lifetime <
770 * LONG_MAX / HZ here.
772 if (time_before(expires,
773 ifa->tstamp + lifetime * HZ))
774 expires = ifa->tstamp + lifetime * HZ;
775 spin_unlock(&ifa->lock);
780 write_unlock_bh(&idev->lock);
782 addrconf_del_timer(ifp);
784 ipv6_ifa_notify(RTM_DELADDR, ifp);
786 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
789 * Purge or update corresponding prefix
791 * 1) we don't purge prefix here if address was not permanent.
792 * prefix is managed by its own lifetime.
793 * 2) if there're no addresses, delete prefix.
794 * 3) if there're still other permanent address(es),
795 * corresponding prefix is still permanent.
796 * 4) otherwise, update prefix lifetime to the
797 * longest valid lifetime among the corresponding
798 * addresses on the device.
799 * Note: subsequent RA will update lifetime.
803 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
804 struct in6_addr prefix;
806 struct net *net = dev_net(ifp->idev->dev);
807 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
808 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
810 if (rt && addrconf_is_prefix_route(rt)) {
814 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
815 rt->rt6i_expires = expires;
816 rt->rt6i_flags |= RTF_EXPIRES;
819 dst_release(&rt->u.dst);
825 #ifdef CONFIG_IPV6_PRIVACY
826 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
828 struct inet6_dev *idev = ifp->idev;
829 struct in6_addr addr, *tmpaddr;
830 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
831 unsigned long regen_advance;
837 write_lock(&idev->lock);
839 spin_lock_bh(&ift->lock);
840 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
841 spin_unlock_bh(&ift->lock);
848 if (idev->cnf.use_tempaddr <= 0) {
849 write_unlock(&idev->lock);
851 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
856 spin_lock_bh(&ifp->lock);
857 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
858 idev->cnf.use_tempaddr = -1; /*XXX*/
859 spin_unlock_bh(&ifp->lock);
860 write_unlock(&idev->lock);
862 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
868 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
869 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
870 spin_unlock_bh(&ifp->lock);
871 write_unlock(&idev->lock);
873 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
879 memcpy(&addr.s6_addr[8], idev->rndid, 8);
880 tmp_valid_lft = min_t(__u32,
882 idev->cnf.temp_valid_lft);
883 tmp_prefered_lft = min_t(__u32,
885 idev->cnf.temp_prefered_lft - desync_factor / HZ);
886 tmp_plen = ifp->prefix_len;
887 max_addresses = idev->cnf.max_addresses;
888 tmp_cstamp = ifp->cstamp;
889 tmp_tstamp = ifp->tstamp;
890 spin_unlock_bh(&ifp->lock);
892 regen_advance = idev->cnf.regen_max_retry *
893 idev->cnf.dad_transmits *
894 idev->nd_parms->retrans_time / HZ;
895 write_unlock(&idev->lock);
897 /* A temporary address is created only if this calculated Preferred
898 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
899 * an implementation must not create a temporary address with a zero
900 * Preferred Lifetime.
902 if (tmp_prefered_lft <= regen_advance) {
909 addr_flags = IFA_F_TEMPORARY;
910 /* set in addrconf_prefix_rcv() */
911 if (ifp->flags & IFA_F_OPTIMISTIC)
912 addr_flags |= IFA_F_OPTIMISTIC;
914 ift = !max_addresses ||
915 ipv6_count_addresses(idev) < max_addresses ?
916 ipv6_add_addr(idev, &addr, tmp_plen,
917 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
919 if (!ift || IS_ERR(ift)) {
923 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
925 write_lock(&idev->lock);
929 spin_lock_bh(&ift->lock);
931 ift->valid_lft = tmp_valid_lft;
932 ift->prefered_lft = tmp_prefered_lft;
933 ift->cstamp = tmp_cstamp;
934 ift->tstamp = tmp_tstamp;
935 spin_unlock_bh(&ift->lock);
937 addrconf_dad_start(ift, 0);
946 * Choose an appropriate source address (RFC3484)
949 IPV6_SADDR_RULE_INIT = 0,
950 IPV6_SADDR_RULE_LOCAL,
951 IPV6_SADDR_RULE_SCOPE,
952 IPV6_SADDR_RULE_PREFERRED,
953 #ifdef CONFIG_IPV6_MIP6
957 IPV6_SADDR_RULE_LABEL,
958 #ifdef CONFIG_IPV6_PRIVACY
959 IPV6_SADDR_RULE_PRIVACY,
961 IPV6_SADDR_RULE_ORCHID,
962 IPV6_SADDR_RULE_PREFIX,
966 struct ipv6_saddr_score {
969 struct inet6_ifaddr *ifa;
970 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
975 struct ipv6_saddr_dst {
976 const struct in6_addr *addr;
983 static inline int ipv6_saddr_preferred(int type)
985 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
990 static int ipv6_get_saddr_eval(struct net *net,
991 struct ipv6_saddr_score *score,
992 struct ipv6_saddr_dst *dst,
997 if (i <= score->rule) {
999 case IPV6_SADDR_RULE_SCOPE:
1000 ret = score->scopedist;
1002 case IPV6_SADDR_RULE_PREFIX:
1003 ret = score->matchlen;
1006 ret = !!test_bit(i, score->scorebits);
1012 case IPV6_SADDR_RULE_INIT:
1013 /* Rule 0: remember if hiscore is not ready yet */
1016 case IPV6_SADDR_RULE_LOCAL:
1017 /* Rule 1: Prefer same address */
1018 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1020 case IPV6_SADDR_RULE_SCOPE:
1021 /* Rule 2: Prefer appropriate scope
1026 * ---+--+-+---> scope
1028 * | d is scope of the destination.
1030 * | \ <- smaller scope is better if
1031 * B-15 | \ if scope is enough for destinaion.
1032 * | ret = B - scope (-1 <= scope >= d <= 15).
1034 * |/ <- greater is better
1035 * -C / if scope is not enough for destination.
1036 * /| ret = scope - C (-1 <= d < scope <= 15).
1038 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1039 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1040 * Assume B = 0 and we get C > 29.
1042 ret = __ipv6_addr_src_scope(score->addr_type);
1043 if (ret >= dst->scope)
1046 ret -= 128; /* 30 is enough */
1047 score->scopedist = ret;
1049 case IPV6_SADDR_RULE_PREFERRED:
1050 /* Rule 3: Avoid deprecated and optimistic addresses */
1051 ret = ipv6_saddr_preferred(score->addr_type) ||
1052 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1054 #ifdef CONFIG_IPV6_MIP6
1055 case IPV6_SADDR_RULE_HOA:
1057 /* Rule 4: Prefer home address */
1058 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1059 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1063 case IPV6_SADDR_RULE_OIF:
1064 /* Rule 5: Prefer outgoing interface */
1065 ret = (!dst->ifindex ||
1066 dst->ifindex == score->ifa->idev->dev->ifindex);
1068 case IPV6_SADDR_RULE_LABEL:
1069 /* Rule 6: Prefer matching label */
1070 ret = ipv6_addr_label(net,
1071 &score->ifa->addr, score->addr_type,
1072 score->ifa->idev->dev->ifindex) == dst->label;
1074 #ifdef CONFIG_IPV6_PRIVACY
1075 case IPV6_SADDR_RULE_PRIVACY:
1077 /* Rule 7: Prefer public address
1078 * Note: prefer temprary address if use_tempaddr >= 2
1080 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1081 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1082 score->ifa->idev->cnf.use_tempaddr >= 2;
1083 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1087 case IPV6_SADDR_RULE_ORCHID:
1088 /* Rule 8-: Prefer ORCHID vs ORCHID or
1089 * non-ORCHID vs non-ORCHID
1091 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1092 ipv6_addr_orchid(dst->addr));
1094 case IPV6_SADDR_RULE_PREFIX:
1095 /* Rule 8: Use longest matching prefix */
1096 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1104 __set_bit(i, score->scorebits);
1110 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1111 const struct in6_addr *daddr, unsigned int prefs,
1112 struct in6_addr *saddr)
1114 struct ipv6_saddr_score scores[2],
1115 *score = &scores[0], *hiscore = &scores[1];
1116 struct ipv6_saddr_dst dst;
1117 struct net_device *dev;
1120 dst_type = __ipv6_addr_type(daddr);
1122 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1123 dst.scope = __ipv6_addr_src_scope(dst_type);
1124 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1128 hiscore->ifa = NULL;
1132 for_each_netdev_rcu(net, dev) {
1133 struct inet6_dev *idev;
1135 /* Candidate Source Address (section 4)
1136 * - multicast and link-local destination address,
1137 * the set of candidate source address MUST only
1138 * include addresses assigned to interfaces
1139 * belonging to the same link as the outgoing
1141 * (- For site-local destination addresses, the
1142 * set of candidate source addresses MUST only
1143 * include addresses assigned to interfaces
1144 * belonging to the same site as the outgoing
1147 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1148 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1149 dst.ifindex && dev->ifindex != dst.ifindex)
1152 idev = __in6_dev_get(dev);
1156 read_lock_bh(&idev->lock);
1157 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1161 * - Tentative Address (RFC2462 section 5.4)
1162 * - A tentative address is not considered
1163 * "assigned to an interface" in the traditional
1164 * sense, unless it is also flagged as optimistic.
1165 * - Candidate Source Address (section 4)
1166 * - In any case, anycast addresses, multicast
1167 * addresses, and the unspecified address MUST
1168 * NOT be included in a candidate set.
1170 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1171 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1174 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1176 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1177 score->addr_type & IPV6_ADDR_MULTICAST)) {
1178 LIMIT_NETDEBUG(KERN_DEBUG
1179 "ADDRCONF: unspecified / multicast address "
1180 "assigned as unicast address on %s",
1186 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1188 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1189 int minihiscore, miniscore;
1191 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1192 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1194 if (minihiscore > miniscore) {
1195 if (i == IPV6_SADDR_RULE_SCOPE &&
1196 score->scopedist > 0) {
1199 * each remaining entry
1200 * has too small (not enough)
1201 * scope, because ifa entries
1202 * are sorted by their scope
1208 } else if (minihiscore < miniscore) {
1210 in6_ifa_put(hiscore->ifa);
1212 in6_ifa_hold(score->ifa);
1214 swap(hiscore, score);
1216 /* restore our iterator */
1217 score->ifa = hiscore->ifa;
1224 read_unlock_bh(&idev->lock);
1229 return -EADDRNOTAVAIL;
1231 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1232 in6_ifa_put(hiscore->ifa);
1235 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1237 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1238 unsigned char banned_flags)
1240 struct inet6_dev *idev;
1241 int err = -EADDRNOTAVAIL;
1244 idev = __in6_dev_get(dev);
1246 struct inet6_ifaddr *ifp;
1248 read_lock_bh(&idev->lock);
1249 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1250 if (ifp->scope == IFA_LINK &&
1251 !(ifp->flags & banned_flags)) {
1252 ipv6_addr_copy(addr, &ifp->addr);
1257 read_unlock_bh(&idev->lock);
1263 static int ipv6_count_addresses(struct inet6_dev *idev)
1266 struct inet6_ifaddr *ifp;
1268 read_lock_bh(&idev->lock);
1269 list_for_each_entry(ifp, &idev->addr_list, if_list)
1271 read_unlock_bh(&idev->lock);
1275 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1276 struct net_device *dev, int strict)
1278 struct inet6_ifaddr *ifp = NULL;
1279 struct hlist_node *node;
1280 unsigned int hash = ipv6_addr_hash(addr);
1283 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1284 if (!net_eq(dev_net(ifp->idev->dev), net))
1286 if (ipv6_addr_equal(&ifp->addr, addr) &&
1287 !(ifp->flags&IFA_F_TENTATIVE)) {
1288 if (dev == NULL || ifp->idev->dev == dev ||
1289 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1293 rcu_read_unlock_bh();
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1299 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1300 struct net_device *dev)
1302 unsigned int hash = ipv6_addr_hash(addr);
1303 struct inet6_ifaddr *ifp;
1304 struct hlist_node *node;
1306 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1307 if (!net_eq(dev_net(ifp->idev->dev), net))
1309 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310 if (dev == NULL || ifp->idev->dev == dev)
1317 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1319 struct inet6_dev *idev;
1320 struct inet6_ifaddr *ifa;
1325 idev = __in6_dev_get(dev);
1327 read_lock_bh(&idev->lock);
1328 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1329 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1334 read_unlock_bh(&idev->lock);
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343 struct net_device *dev, int strict)
1345 struct inet6_ifaddr *ifp, *result = NULL;
1346 unsigned int hash = ipv6_addr_hash(addr);
1347 struct hlist_node *node;
1350 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1351 if (!net_eq(dev_net(ifp->idev->dev), net))
1353 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354 if (dev == NULL || ifp->idev->dev == dev ||
1355 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1362 rcu_read_unlock_bh();
1367 /* Gets referenced address, destroys ifaddr */
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1371 if (ifp->flags&IFA_F_PERMANENT) {
1372 spin_lock_bh(&ifp->lock);
1373 addrconf_del_timer(ifp);
1374 ifp->flags |= IFA_F_TENTATIVE;
1376 ifp->flags |= IFA_F_DADFAILED;
1377 spin_unlock_bh(&ifp->lock);
1379 ipv6_ifa_notify(0, ifp);
1381 #ifdef CONFIG_IPV6_PRIVACY
1382 } else if (ifp->flags&IFA_F_TEMPORARY) {
1383 struct inet6_ifaddr *ifpub;
1384 spin_lock_bh(&ifp->lock);
1387 in6_ifa_hold(ifpub);
1388 spin_unlock_bh(&ifp->lock);
1389 ipv6_create_tempaddr(ifpub, ifp);
1392 spin_unlock_bh(&ifp->lock);
1400 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1402 struct inet6_dev *idev = ifp->idev;
1404 if (net_ratelimit())
1405 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1406 ifp->idev->dev->name, &ifp->addr);
1408 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1409 struct in6_addr addr;
1411 addr.s6_addr32[0] = htonl(0xfe800000);
1412 addr.s6_addr32[1] = 0;
1414 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1415 ipv6_addr_equal(&ifp->addr, &addr)) {
1416 /* DAD failed for link-local based on MAC address */
1417 idev->cnf.disable_ipv6 = 1;
1419 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1420 ifp->idev->dev->name);
1424 addrconf_dad_stop(ifp, 1);
1427 /* Join to solicited addr multicast group. */
1429 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1431 struct in6_addr maddr;
1433 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1436 addrconf_addr_solict_mult(addr, &maddr);
1437 ipv6_dev_mc_inc(dev, &maddr);
1440 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1442 struct in6_addr maddr;
1444 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1447 addrconf_addr_solict_mult(addr, &maddr);
1448 __ipv6_dev_mc_dec(idev, &maddr);
1451 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1453 struct in6_addr addr;
1454 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1455 if (ipv6_addr_any(&addr))
1457 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1460 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1462 struct in6_addr addr;
1463 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1464 if (ipv6_addr_any(&addr))
1466 __ipv6_dev_ac_dec(ifp->idev, &addr);
1469 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1471 if (dev->addr_len != ETH_ALEN)
1473 memcpy(eui, dev->dev_addr, 3);
1474 memcpy(eui + 5, dev->dev_addr + 3, 3);
1477 * The zSeries OSA network cards can be shared among various
1478 * OS instances, but the OSA cards have only one MAC address.
1479 * This leads to duplicate address conflicts in conjunction
1480 * with IPv6 if more than one instance uses the same card.
1482 * The driver for these cards can deliver a unique 16-bit
1483 * identifier for each instance sharing the same card. It is
1484 * placed instead of 0xFFFE in the interface identifier. The
1485 * "u" bit of the interface identifier is not inverted in this
1486 * case. Hence the resulting interface identifier has local
1487 * scope according to RFC2373.
1490 eui[3] = (dev->dev_id >> 8) & 0xFF;
1491 eui[4] = dev->dev_id & 0xFF;
1500 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1502 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1503 if (dev->addr_len != ARCNET_ALEN)
1506 eui[7] = *(u8*)dev->dev_addr;
1510 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1512 if (dev->addr_len != INFINIBAND_ALEN)
1514 memcpy(eui, dev->dev_addr + 12, 8);
1519 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1523 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1524 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1525 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1526 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1527 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1528 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1532 memcpy(eui + 4, &addr, 4);
1535 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1537 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1539 if (dev->priv_flags & IFF_ISATAP)
1540 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1544 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1546 switch (dev->type) {
1549 case ARPHRD_IEEE802_TR:
1550 return addrconf_ifid_eui48(eui, dev);
1552 return addrconf_ifid_arcnet(eui, dev);
1553 case ARPHRD_INFINIBAND:
1554 return addrconf_ifid_infiniband(eui, dev);
1556 return addrconf_ifid_sit(eui, dev);
1561 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1564 struct inet6_ifaddr *ifp;
1566 read_lock_bh(&idev->lock);
1567 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1568 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1569 memcpy(eui, ifp->addr.s6_addr+8, 8);
1574 read_unlock_bh(&idev->lock);
1578 #ifdef CONFIG_IPV6_PRIVACY
1579 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1580 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1583 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1584 idev->rndid[0] &= ~0x02;
1587 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1588 * check if generated address is not inappropriate
1590 * - Reserved subnet anycast (RFC 2526)
1591 * 11111101 11....11 1xxxxxxx
1592 * - ISATAP (RFC4214) 6.1
1593 * 00-00-5E-FE-xx-xx-xx-xx
1595 * - XXX: already assigned to an address on the device
1597 if (idev->rndid[0] == 0xfd &&
1598 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1599 (idev->rndid[7]&0x80))
1601 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1602 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1604 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1611 static void ipv6_regen_rndid(unsigned long data)
1613 struct inet6_dev *idev = (struct inet6_dev *) data;
1614 unsigned long expires;
1617 write_lock_bh(&idev->lock);
1622 if (__ipv6_regen_rndid(idev) < 0)
1626 idev->cnf.temp_prefered_lft * HZ -
1627 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1628 if (time_before(expires, jiffies)) {
1630 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1635 if (!mod_timer(&idev->regen_timer, expires))
1639 write_unlock_bh(&idev->lock);
1640 rcu_read_unlock_bh();
1644 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1647 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1648 ret = __ipv6_regen_rndid(idev);
1658 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1659 unsigned long expires, u32 flags)
1661 struct fib6_config cfg = {
1662 .fc_table = RT6_TABLE_PREFIX,
1663 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1664 .fc_ifindex = dev->ifindex,
1665 .fc_expires = expires,
1667 .fc_flags = RTF_UP | flags,
1668 .fc_nlinfo.nl_net = dev_net(dev),
1669 .fc_protocol = RTPROT_KERNEL,
1672 ipv6_addr_copy(&cfg.fc_dst, pfx);
1674 /* Prevent useless cloning on PtP SIT.
1675 This thing is done here expecting that the whole
1676 class of non-broadcast devices need not cloning.
1678 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1679 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1680 cfg.fc_flags |= RTF_NONEXTHOP;
1683 ip6_route_add(&cfg);
1686 /* Create "default" multicast route to the interface */
1688 static void addrconf_add_mroute(struct net_device *dev)
1690 struct fib6_config cfg = {
1691 .fc_table = RT6_TABLE_LOCAL,
1692 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1693 .fc_ifindex = dev->ifindex,
1696 .fc_nlinfo.nl_net = dev_net(dev),
1699 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1701 ip6_route_add(&cfg);
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705 static void sit_route_add(struct net_device *dev)
1707 struct fib6_config cfg = {
1708 .fc_table = RT6_TABLE_MAIN,
1709 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1710 .fc_ifindex = dev->ifindex,
1712 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1713 .fc_nlinfo.nl_net = dev_net(dev),
1716 /* prefix length - 96 bits "::d.d.d.d" */
1717 ip6_route_add(&cfg);
1721 static void addrconf_add_lroute(struct net_device *dev)
1723 struct in6_addr addr;
1725 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1726 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1729 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1731 struct inet6_dev *idev;
1735 idev = ipv6_find_idev(dev);
1739 /* Add default multicast route */
1740 addrconf_add_mroute(dev);
1742 /* Add link local route */
1743 addrconf_add_lroute(dev);
1747 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1749 struct prefix_info *pinfo;
1753 struct inet6_dev *in6_dev;
1754 struct net *net = dev_net(dev);
1756 pinfo = (struct prefix_info *) opt;
1758 if (len < sizeof(struct prefix_info)) {
1759 ADBG(("addrconf: prefix option too short\n"));
1764 * Validation checks ([ADDRCONF], page 19)
1767 addr_type = ipv6_addr_type(&pinfo->prefix);
1769 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1772 valid_lft = ntohl(pinfo->valid);
1773 prefered_lft = ntohl(pinfo->prefered);
1775 if (prefered_lft > valid_lft) {
1776 if (net_ratelimit())
1777 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1781 in6_dev = in6_dev_get(dev);
1783 if (in6_dev == NULL) {
1784 if (net_ratelimit())
1785 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1790 * Two things going on here:
1791 * 1) Add routes for on-link prefixes
1792 * 2) Configure prefixes with the auto flag set
1795 if (pinfo->onlink) {
1796 struct rt6_info *rt;
1797 unsigned long rt_expires;
1799 /* Avoid arithmetic overflow. Really, we could
1800 * save rt_expires in seconds, likely valid_lft,
1801 * but it would require division in fib gc, that it
1805 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1807 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1809 if (addrconf_finite_timeout(rt_expires))
1812 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1815 if (rt && addrconf_is_prefix_route(rt)) {
1816 /* Autoconf prefix route */
1817 if (valid_lft == 0) {
1820 } else if (addrconf_finite_timeout(rt_expires)) {
1822 rt->rt6i_expires = jiffies + rt_expires;
1823 rt->rt6i_flags |= RTF_EXPIRES;
1825 rt->rt6i_flags &= ~RTF_EXPIRES;
1826 rt->rt6i_expires = 0;
1828 } else if (valid_lft) {
1829 clock_t expires = 0;
1830 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1831 if (addrconf_finite_timeout(rt_expires)) {
1833 flags |= RTF_EXPIRES;
1834 expires = jiffies_to_clock_t(rt_expires);
1836 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1837 dev, expires, flags);
1840 dst_release(&rt->u.dst);
1843 /* Try to figure out our local address for this prefix */
1845 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1846 struct inet6_ifaddr * ifp;
1847 struct in6_addr addr;
1848 int create = 0, update_lft = 0;
1850 if (pinfo->prefix_len == 64) {
1851 memcpy(&addr, &pinfo->prefix, 8);
1852 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1853 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1854 in6_dev_put(in6_dev);
1859 if (net_ratelimit())
1860 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1862 in6_dev_put(in6_dev);
1867 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1869 if (ifp == NULL && valid_lft) {
1870 int max_addresses = in6_dev->cnf.max_addresses;
1873 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1874 if (in6_dev->cnf.optimistic_dad &&
1875 !net->ipv6.devconf_all->forwarding)
1876 addr_flags = IFA_F_OPTIMISTIC;
1879 /* Do not allow to create too much of autoconfigured
1880 * addresses; this would be too easy way to crash kernel.
1882 if (!max_addresses ||
1883 ipv6_count_addresses(in6_dev) < max_addresses)
1884 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1885 addr_type&IPV6_ADDR_SCOPE_MASK,
1888 if (!ifp || IS_ERR(ifp)) {
1889 in6_dev_put(in6_dev);
1893 update_lft = create = 1;
1894 ifp->cstamp = jiffies;
1895 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1901 #ifdef CONFIG_IPV6_PRIVACY
1902 struct inet6_ifaddr *ift;
1906 /* update lifetime (RFC2462 5.5.3 e) */
1907 spin_lock(&ifp->lock);
1909 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1910 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1913 if (!update_lft && stored_lft) {
1914 if (valid_lft > MIN_VALID_LIFETIME ||
1915 valid_lft > stored_lft)
1917 else if (stored_lft <= MIN_VALID_LIFETIME) {
1918 /* valid_lft <= stored_lft is always true */
1920 * RFC 4862 Section 5.5.3e:
1921 * "Note that the preferred lifetime of
1922 * the corresponding address is always
1923 * reset to the Preferred Lifetime in
1924 * the received Prefix Information
1925 * option, regardless of whether the
1926 * valid lifetime is also reset or
1929 * So if the preferred lifetime in
1930 * this advertisement is different
1931 * than what we have stored, but the
1932 * valid lifetime is invalid, just
1933 * reset prefered_lft.
1935 * We must set the valid lifetime
1936 * to the stored lifetime since we'll
1937 * be updating the timestamp below,
1938 * else we'll set it back to the
1941 if (prefered_lft != ifp->prefered_lft) {
1942 valid_lft = stored_lft;
1946 valid_lft = MIN_VALID_LIFETIME;
1947 if (valid_lft < prefered_lft)
1948 prefered_lft = valid_lft;
1954 ifp->valid_lft = valid_lft;
1955 ifp->prefered_lft = prefered_lft;
1958 ifp->flags &= ~IFA_F_DEPRECATED;
1959 spin_unlock(&ifp->lock);
1961 if (!(flags&IFA_F_TENTATIVE))
1962 ipv6_ifa_notify(0, ifp);
1964 spin_unlock(&ifp->lock);
1966 #ifdef CONFIG_IPV6_PRIVACY
1967 read_lock_bh(&in6_dev->lock);
1968 /* update all temporary addresses in the list */
1969 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1971 * When adjusting the lifetimes of an existing
1972 * temporary address, only lower the lifetimes.
1973 * Implementations must not increase the
1974 * lifetimes of an existing temporary address
1975 * when processing a Prefix Information Option.
1977 if (ifp != ift->ifpub)
1980 spin_lock(&ift->lock);
1982 if (ift->valid_lft > valid_lft &&
1983 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1984 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1985 if (ift->prefered_lft > prefered_lft &&
1986 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1987 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1988 spin_unlock(&ift->lock);
1989 if (!(flags&IFA_F_TENTATIVE))
1990 ipv6_ifa_notify(0, ift);
1993 if (create && in6_dev->cnf.use_tempaddr > 0) {
1995 * When a new public address is created as described in [ADDRCONF],
1996 * also create a new temporary address.
1998 read_unlock_bh(&in6_dev->lock);
1999 ipv6_create_tempaddr(ifp, NULL);
2001 read_unlock_bh(&in6_dev->lock);
2008 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2009 in6_dev_put(in6_dev);
2013 * Set destination address.
2014 * Special case for SIT interfaces where we create a new "virtual"
2017 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2019 struct in6_ifreq ireq;
2020 struct net_device *dev;
2026 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2029 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2035 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2036 if (dev->type == ARPHRD_SIT) {
2037 const struct net_device_ops *ops = dev->netdev_ops;
2039 struct ip_tunnel_parm p;
2041 err = -EADDRNOTAVAIL;
2042 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2045 memset(&p, 0, sizeof(p));
2046 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2050 p.iph.protocol = IPPROTO_IPV6;
2052 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2054 if (ops->ndo_do_ioctl) {
2055 mm_segment_t oldfs = get_fs();
2058 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2065 dev = __dev_get_by_name(net, p.name);
2068 err = dev_open(dev);
2079 * Manual configuration of address on an interface
2081 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2082 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2085 struct inet6_ifaddr *ifp;
2086 struct inet6_dev *idev;
2087 struct net_device *dev;
2091 unsigned long timeout;
2098 /* check the lifetime */
2099 if (!valid_lft || prefered_lft > valid_lft)
2102 dev = __dev_get_by_index(net, ifindex);
2106 if ((idev = addrconf_add_dev(dev)) == NULL)
2109 scope = ipv6_addr_scope(pfx);
2111 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2112 if (addrconf_finite_timeout(timeout)) {
2113 expires = jiffies_to_clock_t(timeout * HZ);
2114 valid_lft = timeout;
2115 flags = RTF_EXPIRES;
2119 ifa_flags |= IFA_F_PERMANENT;
2122 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2123 if (addrconf_finite_timeout(timeout)) {
2125 ifa_flags |= IFA_F_DEPRECATED;
2126 prefered_lft = timeout;
2129 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2132 spin_lock_bh(&ifp->lock);
2133 ifp->valid_lft = valid_lft;
2134 ifp->prefered_lft = prefered_lft;
2135 ifp->tstamp = jiffies;
2136 spin_unlock_bh(&ifp->lock);
2138 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2141 * Note that section 3.1 of RFC 4429 indicates
2142 * that the Optimistic flag should not be set for
2143 * manually configured addresses
2145 addrconf_dad_start(ifp, 0);
2151 return PTR_ERR(ifp);
2154 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2157 struct inet6_ifaddr *ifp;
2158 struct inet6_dev *idev;
2159 struct net_device *dev;
2164 dev = __dev_get_by_index(net, ifindex);
2168 if ((idev = __in6_dev_get(dev)) == NULL)
2171 read_lock_bh(&idev->lock);
2172 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2173 if (ifp->prefix_len == plen &&
2174 ipv6_addr_equal(pfx, &ifp->addr)) {
2176 read_unlock_bh(&idev->lock);
2180 /* If the last address is deleted administratively,
2181 disable IPv6 on this interface.
2183 if (list_empty(&idev->addr_list))
2184 addrconf_ifdown(idev->dev, 1);
2188 read_unlock_bh(&idev->lock);
2189 return -EADDRNOTAVAIL;
2193 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2195 struct in6_ifreq ireq;
2198 if (!capable(CAP_NET_ADMIN))
2201 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2205 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2206 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2207 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2212 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2214 struct in6_ifreq ireq;
2217 if (!capable(CAP_NET_ADMIN))
2220 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2224 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2225 ireq.ifr6_prefixlen);
2230 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2231 int plen, int scope)
2233 struct inet6_ifaddr *ifp;
2235 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2237 spin_lock_bh(&ifp->lock);
2238 ifp->flags &= ~IFA_F_TENTATIVE;
2239 spin_unlock_bh(&ifp->lock);
2240 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2245 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2246 static void sit_add_v4_addrs(struct inet6_dev *idev)
2248 struct in6_addr addr;
2249 struct net_device *dev;
2250 struct net *net = dev_net(idev->dev);
2255 memset(&addr, 0, sizeof(struct in6_addr));
2256 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2258 if (idev->dev->flags&IFF_POINTOPOINT) {
2259 addr.s6_addr32[0] = htonl(0xfe800000);
2262 scope = IPV6_ADDR_COMPATv4;
2265 if (addr.s6_addr32[3]) {
2266 add_addr(idev, &addr, 128, scope);
2270 for_each_netdev(net, dev) {
2271 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2272 if (in_dev && (dev->flags & IFF_UP)) {
2273 struct in_ifaddr * ifa;
2277 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2280 addr.s6_addr32[3] = ifa->ifa_local;
2282 if (ifa->ifa_scope == RT_SCOPE_LINK)
2284 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2285 if (idev->dev->flags&IFF_POINTOPOINT)
2289 if (idev->dev->flags&IFF_POINTOPOINT)
2294 add_addr(idev, &addr, plen, flag);
2301 static void init_loopback(struct net_device *dev)
2303 struct inet6_dev *idev;
2309 if ((idev = ipv6_find_idev(dev)) == NULL) {
2310 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2314 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2317 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2319 struct inet6_ifaddr * ifp;
2320 u32 addr_flags = IFA_F_PERMANENT;
2322 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2323 if (idev->cnf.optimistic_dad &&
2324 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2325 addr_flags |= IFA_F_OPTIMISTIC;
2329 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2331 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2332 addrconf_dad_start(ifp, 0);
2337 static void addrconf_dev_config(struct net_device *dev)
2339 struct in6_addr addr;
2340 struct inet6_dev * idev;
2344 if ((dev->type != ARPHRD_ETHER) &&
2345 (dev->type != ARPHRD_FDDI) &&
2346 (dev->type != ARPHRD_IEEE802_TR) &&
2347 (dev->type != ARPHRD_ARCNET) &&
2348 (dev->type != ARPHRD_INFINIBAND)) {
2349 /* Alas, we support only Ethernet autoconfiguration. */
2353 idev = addrconf_add_dev(dev);
2357 memset(&addr, 0, sizeof(struct in6_addr));
2358 addr.s6_addr32[0] = htonl(0xFE800000);
2360 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2361 addrconf_add_linklocal(idev, &addr);
2364 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2365 static void addrconf_sit_config(struct net_device *dev)
2367 struct inet6_dev *idev;
2372 * Configure the tunnel with one of our IPv4
2373 * addresses... we should configure all of
2374 * our v4 addrs in the tunnel
2377 if ((idev = ipv6_find_idev(dev)) == NULL) {
2378 printk(KERN_DEBUG "init sit: add_dev failed\n");
2382 if (dev->priv_flags & IFF_ISATAP) {
2383 struct in6_addr addr;
2385 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2386 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2387 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2388 addrconf_add_linklocal(idev, &addr);
2392 sit_add_v4_addrs(idev);
2394 if (dev->flags&IFF_POINTOPOINT) {
2395 addrconf_add_mroute(dev);
2396 addrconf_add_lroute(dev);
2403 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2405 struct in6_addr lladdr;
2407 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2408 addrconf_add_linklocal(idev, &lladdr);
2414 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2416 struct net_device *link_dev;
2417 struct net *net = dev_net(idev->dev);
2419 /* first try to inherit the link-local address from the link device */
2420 if (idev->dev->iflink &&
2421 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2422 if (!ipv6_inherit_linklocal(idev, link_dev))
2425 /* then try to inherit it from any device */
2426 for_each_netdev(net, link_dev) {
2427 if (!ipv6_inherit_linklocal(idev, link_dev))
2430 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2434 * Autoconfigure tunnel with a link-local address so routing protocols,
2435 * DHCPv6, MLD etc. can be run over the virtual link
2438 static void addrconf_ip6_tnl_config(struct net_device *dev)
2440 struct inet6_dev *idev;
2444 idev = addrconf_add_dev(dev);
2446 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2449 ip6_tnl_add_linklocal(idev);
2452 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2455 struct net_device *dev = (struct net_device *) data;
2456 struct inet6_dev *idev = __in6_dev_get(dev);
2457 int run_pending = 0;
2461 case NETDEV_REGISTER:
2462 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2463 idev = ipv6_add_dev(dev);
2465 return notifier_from_errno(-ENOMEM);
2471 if (dev->flags & IFF_SLAVE)
2474 if (event == NETDEV_UP) {
2475 if (!addrconf_qdisc_ok(dev)) {
2476 /* device is not ready yet. */
2478 "ADDRCONF(NETDEV_UP): %s: "
2479 "link is not ready\n",
2484 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2485 idev = ipv6_add_dev(dev);
2488 idev->if_flags |= IF_READY;
2492 if (!addrconf_qdisc_ok(dev)) {
2493 /* device is still not ready. */
2498 if (idev->if_flags & IF_READY)
2499 /* device is already configured. */
2501 idev->if_flags |= IF_READY;
2505 "ADDRCONF(NETDEV_CHANGE): %s: "
2506 "link becomes ready\n",
2512 switch (dev->type) {
2513 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2515 addrconf_sit_config(dev);
2518 case ARPHRD_TUNNEL6:
2519 addrconf_ip6_tnl_config(dev);
2521 case ARPHRD_LOOPBACK:
2526 addrconf_dev_config(dev);
2532 addrconf_dad_run(idev);
2535 * If the MTU changed during the interface down,
2536 * when the interface up, the changed MTU must be
2537 * reflected in the idev as well as routers.
2539 if (idev->cnf.mtu6 != dev->mtu &&
2540 dev->mtu >= IPV6_MIN_MTU) {
2541 rt6_mtu_change(dev, dev->mtu);
2542 idev->cnf.mtu6 = dev->mtu;
2544 idev->tstamp = jiffies;
2545 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2548 * If the changed mtu during down is lower than
2549 * IPV6_MIN_MTU stop IPv6 on this interface.
2551 if (dev->mtu < IPV6_MIN_MTU)
2552 addrconf_ifdown(dev, 1);
2556 case NETDEV_CHANGEMTU:
2557 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2558 rt6_mtu_change(dev, dev->mtu);
2559 idev->cnf.mtu6 = dev->mtu;
2563 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2564 idev = ipv6_add_dev(dev);
2570 * MTU falled under IPV6_MIN_MTU.
2571 * Stop IPv6 on this interface.
2575 case NETDEV_UNREGISTER:
2577 * Remove all addresses from this interface.
2579 addrconf_ifdown(dev, event != NETDEV_DOWN);
2582 case NETDEV_CHANGENAME:
2584 snmp6_unregister_dev(idev);
2585 addrconf_sysctl_unregister(idev);
2586 addrconf_sysctl_register(idev);
2587 err = snmp6_register_dev(idev);
2589 return notifier_from_errno(err);
2593 case NETDEV_PRE_TYPE_CHANGE:
2594 case NETDEV_POST_TYPE_CHANGE:
2595 addrconf_type_change(dev, event);
2603 * addrconf module should be notified of a device going up
2605 static struct notifier_block ipv6_dev_notf = {
2606 .notifier_call = addrconf_notify,
2609 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2611 struct inet6_dev *idev;
2614 idev = __in6_dev_get(dev);
2616 if (event == NETDEV_POST_TYPE_CHANGE)
2617 ipv6_mc_remap(idev);
2618 else if (event == NETDEV_PRE_TYPE_CHANGE)
2619 ipv6_mc_unmap(idev);
2622 static int addrconf_ifdown(struct net_device *dev, int how)
2624 struct net *net = dev_net(dev);
2625 struct inet6_dev *idev;
2626 struct inet6_ifaddr *ifa;
2627 LIST_HEAD(keep_list);
2631 rt6_ifdown(net, dev);
2632 neigh_ifdown(&nd_tbl, dev);
2634 idev = __in6_dev_get(dev);
2639 * Step 1: remove reference to ipv6 device from parent device.
2645 /* protected by rtnl_lock */
2646 rcu_assign_pointer(dev->ip6_ptr, NULL);
2648 /* Step 1.5: remove snmp6 entry */
2649 snmp6_unregister_dev(idev);
2653 write_lock_bh(&idev->lock);
2655 /* Step 2: clear flags for stateless addrconf */
2657 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2659 #ifdef CONFIG_IPV6_PRIVACY
2660 if (how && del_timer(&idev->regen_timer))
2663 /* Step 3: clear tempaddr list */
2664 while (!list_empty(&idev->tempaddr_list)) {
2665 ifa = list_first_entry(&idev->tempaddr_list,
2666 struct inet6_ifaddr, tmp_list);
2667 list_del(&ifa->tmp_list);
2669 write_unlock_bh(&idev->lock);
2670 spin_lock_bh(&ifa->lock);
2673 in6_ifa_put(ifa->ifpub);
2676 spin_unlock_bh(&ifa->lock);
2678 write_lock_bh(&idev->lock);
2682 while (!list_empty(&idev->addr_list)) {
2683 ifa = list_first_entry(&idev->addr_list,
2684 struct inet6_ifaddr, if_list);
2685 addrconf_del_timer(ifa);
2687 /* If just doing link down, and address is permanent
2688 and not link-local, then retain it. */
2690 (ifa->flags&IFA_F_PERMANENT) &&
2691 !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2692 list_move_tail(&ifa->if_list, &keep_list);
2694 /* If not doing DAD on this address, just keep it. */
2695 if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2696 idev->cnf.accept_dad <= 0 ||
2697 (ifa->flags & IFA_F_NODAD))
2700 /* If it was tentative already, no need to notify */
2701 if (ifa->flags & IFA_F_TENTATIVE)
2704 /* Flag it for later restoration when link comes up */
2705 ifa->flags |= IFA_F_TENTATIVE;
2708 list_del(&ifa->if_list);
2711 write_unlock_bh(&idev->lock);
2713 /* clear hash table */
2714 spin_lock_bh(&addrconf_hash_lock);
2715 hlist_del_init_rcu(&ifa->addr_lst);
2717 spin_unlock_bh(&addrconf_hash_lock);
2719 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2720 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2723 write_lock_bh(&idev->lock);
2726 list_splice(&keep_list, &idev->addr_list);
2728 write_unlock_bh(&idev->lock);
2730 /* Step 5: Discard multicast list */
2732 ipv6_mc_destroy_dev(idev);
2736 idev->tstamp = jiffies;
2738 /* Last: Shot the device (if unregistered) */
2740 addrconf_sysctl_unregister(idev);
2741 neigh_parms_release(&nd_tbl, idev->nd_parms);
2742 neigh_ifdown(&nd_tbl, dev);
2748 static void addrconf_rs_timer(unsigned long data)
2750 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2751 struct inet6_dev *idev = ifp->idev;
2753 read_lock(&idev->lock);
2754 if (idev->dead || !(idev->if_flags & IF_READY))
2757 if (idev->cnf.forwarding)
2760 /* Announcement received after solicitation was sent */
2761 if (idev->if_flags & IF_RA_RCVD)
2764 spin_lock(&ifp->lock);
2765 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2766 /* The wait after the last probe can be shorter */
2767 addrconf_mod_timer(ifp, AC_RS,
2768 (ifp->probes == idev->cnf.rtr_solicits) ?
2769 idev->cnf.rtr_solicit_delay :
2770 idev->cnf.rtr_solicit_interval);
2771 spin_unlock(&ifp->lock);
2773 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2775 spin_unlock(&ifp->lock);
2777 * Note: we do not support deprecated "all on-link"
2778 * assumption any longer.
2780 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2785 read_unlock(&idev->lock);
2790 * Duplicate Address Detection
2792 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2794 unsigned long rand_num;
2795 struct inet6_dev *idev = ifp->idev;
2797 if (ifp->flags & IFA_F_OPTIMISTIC)
2800 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2802 ifp->probes = idev->cnf.dad_transmits;
2803 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2806 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2808 struct inet6_dev *idev = ifp->idev;
2809 struct net_device *dev = idev->dev;
2811 addrconf_join_solict(dev, &ifp->addr);
2813 net_srandom(ifp->addr.s6_addr32[3]);
2815 read_lock_bh(&idev->lock);
2819 spin_lock(&ifp->lock);
2820 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2821 idev->cnf.accept_dad < 1 ||
2822 !(ifp->flags&IFA_F_TENTATIVE) ||
2823 ifp->flags & IFA_F_NODAD) {
2824 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2825 spin_unlock(&ifp->lock);
2826 read_unlock_bh(&idev->lock);
2828 addrconf_dad_completed(ifp);
2832 if (!(idev->if_flags & IF_READY)) {
2833 spin_unlock(&ifp->lock);
2834 read_unlock_bh(&idev->lock);
2836 * If the device is not ready:
2837 * - keep it tentative if it is a permanent address.
2838 * - otherwise, kill it.
2841 addrconf_dad_stop(ifp, 0);
2846 * Optimistic nodes can start receiving
2849 if (ifp->flags & IFA_F_OPTIMISTIC)
2850 ip6_ins_rt(ifp->rt);
2852 addrconf_dad_kick(ifp);
2853 spin_unlock(&ifp->lock);
2855 read_unlock_bh(&idev->lock);
2858 static void addrconf_dad_timer(unsigned long data)
2860 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2861 struct inet6_dev *idev = ifp->idev;
2862 struct in6_addr mcaddr;
2864 read_lock(&idev->lock);
2865 if (idev->dead || !(idev->if_flags & IF_READY)) {
2866 read_unlock(&idev->lock);
2870 spin_lock(&ifp->lock);
2871 if (ifp->probes == 0) {
2873 * DAD was successful
2876 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2877 spin_unlock(&ifp->lock);
2878 read_unlock(&idev->lock);
2880 addrconf_dad_completed(ifp);
2886 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2887 spin_unlock(&ifp->lock);
2888 read_unlock(&idev->lock);
2890 /* send a neighbour solicitation for our addr */
2891 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2892 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2897 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2899 struct net_device *dev = ifp->idev->dev;
2902 * Configure the address for reception. Now it is valid.
2905 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2907 /* If added prefix is link local and forwarding is off,
2908 start sending router solicitations.
2911 if (ifp->idev->cnf.forwarding == 0 &&
2912 ifp->idev->cnf.rtr_solicits > 0 &&
2913 (dev->flags&IFF_LOOPBACK) == 0 &&
2914 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2916 * If a host as already performed a random delay
2917 * [...] as part of DAD [...] there is no need
2918 * to delay again before sending the first RS
2920 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2922 spin_lock_bh(&ifp->lock);
2924 ifp->idev->if_flags |= IF_RS_SENT;
2925 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2926 spin_unlock_bh(&ifp->lock);
2930 static void addrconf_dad_run(struct inet6_dev *idev)
2932 struct inet6_ifaddr *ifp;
2934 read_lock_bh(&idev->lock);
2935 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2936 spin_lock(&ifp->lock);
2937 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2938 spin_unlock(&ifp->lock);
2941 spin_unlock(&ifp->lock);
2942 addrconf_dad_kick(ifp);
2944 read_unlock_bh(&idev->lock);
2947 #ifdef CONFIG_PROC_FS
2948 struct if6_iter_state {
2949 struct seq_net_private p;
2953 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2955 struct inet6_ifaddr *ifa = NULL;
2956 struct if6_iter_state *state = seq->private;
2957 struct net *net = seq_file_net(seq);
2959 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2960 struct hlist_node *n;
2961 hlist_for_each_entry_rcu(ifa, n, &inet6_addr_lst[state->bucket],
2963 if (net_eq(dev_net(ifa->idev->dev), net))
2969 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2970 struct inet6_ifaddr *ifa)
2972 struct if6_iter_state *state = seq->private;
2973 struct net *net = seq_file_net(seq);
2974 struct hlist_node *n = &ifa->addr_lst;
2976 hlist_for_each_entry_continue_rcu(ifa, n, addr_lst)
2977 if (net_eq(dev_net(ifa->idev->dev), net))
2980 while (++state->bucket < IN6_ADDR_HSIZE) {
2981 hlist_for_each_entry(ifa, n,
2982 &inet6_addr_lst[state->bucket], addr_lst) {
2983 if (net_eq(dev_net(ifa->idev->dev), net))
2991 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2993 struct inet6_ifaddr *ifa = if6_get_first(seq);
2996 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2998 return pos ? NULL : ifa;
3001 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3005 return if6_get_idx(seq, *pos);
3008 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3010 struct inet6_ifaddr *ifa;
3012 ifa = if6_get_next(seq, v);
3017 static void if6_seq_stop(struct seq_file *seq, void *v)
3020 rcu_read_unlock_bh();
3023 static int if6_seq_show(struct seq_file *seq, void *v)
3025 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3026 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3028 ifp->idev->dev->ifindex,
3032 ifp->idev->dev->name);
3036 static const struct seq_operations if6_seq_ops = {
3037 .start = if6_seq_start,
3038 .next = if6_seq_next,
3039 .show = if6_seq_show,
3040 .stop = if6_seq_stop,
3043 static int if6_seq_open(struct inode *inode, struct file *file)
3045 return seq_open_net(inode, file, &if6_seq_ops,
3046 sizeof(struct if6_iter_state));
3049 static const struct file_operations if6_fops = {
3050 .owner = THIS_MODULE,
3051 .open = if6_seq_open,
3053 .llseek = seq_lseek,
3054 .release = seq_release_net,
3057 static int __net_init if6_proc_net_init(struct net *net)
3059 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3064 static void __net_exit if6_proc_net_exit(struct net *net)
3066 proc_net_remove(net, "if_inet6");
3069 static struct pernet_operations if6_proc_net_ops = {
3070 .init = if6_proc_net_init,
3071 .exit = if6_proc_net_exit,
3074 int __init if6_proc_init(void)
3076 return register_pernet_subsys(&if6_proc_net_ops);
3079 void if6_proc_exit(void)
3081 unregister_pernet_subsys(&if6_proc_net_ops);
3083 #endif /* CONFIG_PROC_FS */
3085 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3086 /* Check if address is a home address configured on any interface. */
3087 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3090 struct inet6_ifaddr *ifp = NULL;
3091 struct hlist_node *n;
3092 unsigned int hash = ipv6_addr_hash(addr);
3095 hlist_for_each_entry_rcu(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3096 if (!net_eq(dev_net(ifp->idev->dev), net))
3098 if (ipv6_addr_equal(&ifp->addr, addr) &&
3099 (ifp->flags & IFA_F_HOMEADDRESS)) {
3104 rcu_read_unlock_bh();
3110 * Periodic address status verification
3113 static void addrconf_verify(unsigned long foo)
3115 unsigned long now, next, next_sec, next_sched;
3116 struct inet6_ifaddr *ifp;
3117 struct hlist_node *node;
3121 spin_lock(&addrconf_verify_lock);
3123 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3125 del_timer(&addr_chk_timer);
3127 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3129 hlist_for_each_entry_rcu(ifp, node,
3130 &inet6_addr_lst[i], addr_lst) {
3133 if (ifp->flags & IFA_F_PERMANENT)
3136 spin_lock(&ifp->lock);
3137 /* We try to batch several events at once. */
3138 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3140 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3141 age >= ifp->valid_lft) {
3142 spin_unlock(&ifp->lock);
3146 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3147 spin_unlock(&ifp->lock);
3149 } else if (age >= ifp->prefered_lft) {
3150 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3153 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3155 ifp->flags |= IFA_F_DEPRECATED;
3158 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3159 next = ifp->tstamp + ifp->valid_lft * HZ;
3161 spin_unlock(&ifp->lock);
3166 ipv6_ifa_notify(0, ifp);
3170 #ifdef CONFIG_IPV6_PRIVACY
3171 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3172 !(ifp->flags&IFA_F_TENTATIVE)) {
3173 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3174 ifp->idev->cnf.dad_transmits *
3175 ifp->idev->nd_parms->retrans_time / HZ;
3177 if (age >= ifp->prefered_lft - regen_advance) {
3178 struct inet6_ifaddr *ifpub = ifp->ifpub;
3179 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3180 next = ifp->tstamp + ifp->prefered_lft * HZ;
3181 if (!ifp->regen_count && ifpub) {
3184 in6_ifa_hold(ifpub);
3185 spin_unlock(&ifp->lock);
3187 spin_lock(&ifpub->lock);
3188 ifpub->regen_count = 0;
3189 spin_unlock(&ifpub->lock);
3190 ipv6_create_tempaddr(ifpub, ifp);
3195 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3196 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3197 spin_unlock(&ifp->lock);
3200 /* ifp->prefered_lft <= ifp->valid_lft */
3201 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3202 next = ifp->tstamp + ifp->prefered_lft * HZ;
3203 spin_unlock(&ifp->lock);
3208 next_sec = round_jiffies_up(next);
3211 /* If rounded timeout is accurate enough, accept it. */
3212 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3213 next_sched = next_sec;
3215 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3216 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3217 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3219 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3220 now, next, next_sec, next_sched));
3222 addr_chk_timer.expires = next_sched;
3223 add_timer(&addr_chk_timer);
3224 spin_unlock(&addrconf_verify_lock);
3225 rcu_read_unlock_bh();
3228 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3230 struct in6_addr *pfx = NULL;
3233 pfx = nla_data(addr);
3236 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3239 pfx = nla_data(local);
3245 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3246 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3247 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3248 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3252 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3254 struct net *net = sock_net(skb->sk);
3255 struct ifaddrmsg *ifm;
3256 struct nlattr *tb[IFA_MAX+1];
3257 struct in6_addr *pfx;
3260 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3264 ifm = nlmsg_data(nlh);
3265 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3269 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3272 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3273 u32 prefered_lft, u32 valid_lft)
3277 unsigned long timeout;
3279 if (!valid_lft || (prefered_lft > valid_lft))
3282 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3283 if (addrconf_finite_timeout(timeout)) {
3284 expires = jiffies_to_clock_t(timeout * HZ);
3285 valid_lft = timeout;
3286 flags = RTF_EXPIRES;
3290 ifa_flags |= IFA_F_PERMANENT;
3293 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3294 if (addrconf_finite_timeout(timeout)) {
3296 ifa_flags |= IFA_F_DEPRECATED;
3297 prefered_lft = timeout;
3300 spin_lock_bh(&ifp->lock);
3301 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3302 ifp->tstamp = jiffies;
3303 ifp->valid_lft = valid_lft;
3304 ifp->prefered_lft = prefered_lft;
3306 spin_unlock_bh(&ifp->lock);
3307 if (!(ifp->flags&IFA_F_TENTATIVE))
3308 ipv6_ifa_notify(0, ifp);
3310 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3318 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3320 struct net *net = sock_net(skb->sk);
3321 struct ifaddrmsg *ifm;
3322 struct nlattr *tb[IFA_MAX+1];
3323 struct in6_addr *pfx;
3324 struct inet6_ifaddr *ifa;
3325 struct net_device *dev;
3326 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3330 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3334 ifm = nlmsg_data(nlh);
3335 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3339 if (tb[IFA_CACHEINFO]) {
3340 struct ifa_cacheinfo *ci;
3342 ci = nla_data(tb[IFA_CACHEINFO]);
3343 valid_lft = ci->ifa_valid;
3344 preferred_lft = ci->ifa_prefered;
3346 preferred_lft = INFINITY_LIFE_TIME;
3347 valid_lft = INFINITY_LIFE_TIME;
3350 dev = __dev_get_by_index(net, ifm->ifa_index);
3354 /* We ignore other flags so far. */
3355 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3357 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3360 * It would be best to check for !NLM_F_CREATE here but
3361 * userspace alreay relies on not having to provide this.
3363 return inet6_addr_add(net, ifm->ifa_index, pfx,
3364 ifm->ifa_prefixlen, ifa_flags,
3365 preferred_lft, valid_lft);
3368 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3369 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3372 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3379 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3380 u8 scope, int ifindex)
3382 struct ifaddrmsg *ifm;
3384 ifm = nlmsg_data(nlh);
3385 ifm->ifa_family = AF_INET6;
3386 ifm->ifa_prefixlen = prefixlen;
3387 ifm->ifa_flags = flags;
3388 ifm->ifa_scope = scope;
3389 ifm->ifa_index = ifindex;
3392 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3393 unsigned long tstamp, u32 preferred, u32 valid)
3395 struct ifa_cacheinfo ci;
3397 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3398 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3399 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3400 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3401 ci.ifa_prefered = preferred;
3402 ci.ifa_valid = valid;
3404 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3407 static inline int rt_scope(int ifa_scope)
3409 if (ifa_scope & IFA_HOST)
3410 return RT_SCOPE_HOST;
3411 else if (ifa_scope & IFA_LINK)
3412 return RT_SCOPE_LINK;
3413 else if (ifa_scope & IFA_SITE)
3414 return RT_SCOPE_SITE;
3416 return RT_SCOPE_UNIVERSE;
3419 static inline int inet6_ifaddr_msgsize(void)
3421 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3422 + nla_total_size(16) /* IFA_ADDRESS */
3423 + nla_total_size(sizeof(struct ifa_cacheinfo));
3426 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3427 u32 pid, u32 seq, int event, unsigned int flags)
3429 struct nlmsghdr *nlh;
3430 u32 preferred, valid;
3432 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3436 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3437 ifa->idev->dev->ifindex);
3439 if (!(ifa->flags&IFA_F_PERMANENT)) {
3440 preferred = ifa->prefered_lft;
3441 valid = ifa->valid_lft;
3442 if (preferred != INFINITY_LIFE_TIME) {
3443 long tval = (jiffies - ifa->tstamp)/HZ;
3444 if (preferred > tval)
3448 if (valid != INFINITY_LIFE_TIME)
3452 preferred = INFINITY_LIFE_TIME;
3453 valid = INFINITY_LIFE_TIME;
3456 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3457 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3458 nlmsg_cancel(skb, nlh);
3462 return nlmsg_end(skb, nlh);
3465 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3466 u32 pid, u32 seq, int event, u16 flags)
3468 struct nlmsghdr *nlh;
3469 u8 scope = RT_SCOPE_UNIVERSE;
3470 int ifindex = ifmca->idev->dev->ifindex;
3472 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3473 scope = RT_SCOPE_SITE;
3475 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3479 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3480 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3481 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3482 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3483 nlmsg_cancel(skb, nlh);
3487 return nlmsg_end(skb, nlh);
3490 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3491 u32 pid, u32 seq, int event, unsigned int flags)
3493 struct nlmsghdr *nlh;
3494 u8 scope = RT_SCOPE_UNIVERSE;
3495 int ifindex = ifaca->aca_idev->dev->ifindex;
3497 if (ipv6_addr_scope(&ifaca->aca_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_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3506 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3507 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3508 nlmsg_cancel(skb, nlh);
3512 return nlmsg_end(skb, nlh);
3521 /* called with rcu_read_lock() */
3522 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3523 struct netlink_callback *cb, enum addr_type_t type,
3524 int s_ip_idx, int *p_ip_idx)
3526 struct ifmcaddr6 *ifmca;
3527 struct ifacaddr6 *ifaca;
3529 int ip_idx = *p_ip_idx;
3531 read_lock_bh(&idev->lock);
3533 case UNICAST_ADDR: {
3534 struct inet6_ifaddr *ifa;
3536 /* unicast address incl. temp addr */
3537 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3538 if (++ip_idx < s_ip_idx)
3540 err = inet6_fill_ifaddr(skb, ifa,
3541 NETLINK_CB(cb->skb).pid,
3550 case MULTICAST_ADDR:
3551 /* multicast address */
3552 for (ifmca = idev->mc_list; ifmca;
3553 ifmca = ifmca->next, ip_idx++) {
3554 if (ip_idx < s_ip_idx)
3556 err = inet6_fill_ifmcaddr(skb, ifmca,
3557 NETLINK_CB(cb->skb).pid,
3566 /* anycast address */
3567 for (ifaca = idev->ac_list; ifaca;
3568 ifaca = ifaca->aca_next, ip_idx++) {
3569 if (ip_idx < s_ip_idx)
3571 err = inet6_fill_ifacaddr(skb, ifaca,
3572 NETLINK_CB(cb->skb).pid,
3583 read_unlock_bh(&idev->lock);
3588 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3589 enum addr_type_t type)
3591 struct net *net = sock_net(skb->sk);
3594 int s_idx, s_ip_idx;
3595 struct net_device *dev;
3596 struct inet6_dev *idev;
3597 struct hlist_head *head;
3598 struct hlist_node *node;
3601 s_idx = idx = cb->args[1];
3602 s_ip_idx = ip_idx = cb->args[2];
3605 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3607 head = &net->dev_index_head[h];
3608 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3611 if (h > s_h || idx > s_idx)
3614 idev = __in6_dev_get(dev);
3618 if (in6_dump_addrs(idev, skb, cb, type,
3619 s_ip_idx, &ip_idx) <= 0)
3629 cb->args[2] = ip_idx;
3634 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3636 enum addr_type_t type = UNICAST_ADDR;
3638 return inet6_dump_addr(skb, cb, type);
3641 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3643 enum addr_type_t type = MULTICAST_ADDR;
3645 return inet6_dump_addr(skb, cb, type);
3649 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3651 enum addr_type_t type = ANYCAST_ADDR;
3653 return inet6_dump_addr(skb, cb, type);
3656 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3659 struct net *net = sock_net(in_skb->sk);
3660 struct ifaddrmsg *ifm;
3661 struct nlattr *tb[IFA_MAX+1];
3662 struct in6_addr *addr = NULL;
3663 struct net_device *dev = NULL;
3664 struct inet6_ifaddr *ifa;
3665 struct sk_buff *skb;
3668 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3672 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3678 ifm = nlmsg_data(nlh);
3680 dev = __dev_get_by_index(net, ifm->ifa_index);
3682 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3684 err = -EADDRNOTAVAIL;
3688 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3694 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3695 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3697 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3698 WARN_ON(err == -EMSGSIZE);
3702 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3709 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3711 struct sk_buff *skb;
3712 struct net *net = dev_net(ifa->idev->dev);
3715 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3719 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3721 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3722 WARN_ON(err == -EMSGSIZE);
3726 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3730 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3733 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3734 __s32 *array, int bytes)
3736 BUG_ON(bytes < (DEVCONF_MAX * 4));
3738 memset(array, 0, bytes);
3739 array[DEVCONF_FORWARDING] = cnf->forwarding;
3740 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3741 array[DEVCONF_MTU6] = cnf->mtu6;
3742 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3743 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3744 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3745 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3746 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3747 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3748 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3749 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3750 #ifdef CONFIG_IPV6_PRIVACY
3751 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3752 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3753 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3754 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3755 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3757 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3758 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3759 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3760 #ifdef CONFIG_IPV6_ROUTER_PREF
3761 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3762 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3763 #ifdef CONFIG_IPV6_ROUTE_INFO
3764 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3767 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3768 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3769 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3770 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3772 #ifdef CONFIG_IPV6_MROUTE
3773 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3775 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3776 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3777 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3780 static inline size_t inet6_if_nlmsg_size(void)
3782 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3783 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3784 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3785 + nla_total_size(4) /* IFLA_MTU */
3786 + nla_total_size(4) /* IFLA_LINK */
3787 + nla_total_size( /* IFLA_PROTINFO */
3788 nla_total_size(4) /* IFLA_INET6_FLAGS */
3789 + nla_total_size(sizeof(struct ifla_cacheinfo))
3790 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3791 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3792 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3796 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3797 int items, int bytes)
3800 int pad = bytes - sizeof(u64) * items;
3803 /* Use put_unaligned() because stats may not be aligned for u64. */
3804 put_unaligned(items, &stats[0]);
3805 for (i = 1; i < items; i++)
3806 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3808 memset(&stats[items], 0, pad);
3811 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3815 case IFLA_INET6_STATS:
3816 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3818 case IFLA_INET6_ICMP6STATS:
3819 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3824 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3825 u32 pid, u32 seq, int event, unsigned int flags)
3827 struct net_device *dev = idev->dev;
3829 struct ifinfomsg *hdr;
3830 struct nlmsghdr *nlh;
3832 struct ifla_cacheinfo ci;
3834 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3838 hdr = nlmsg_data(nlh);
3839 hdr->ifi_family = AF_INET6;
3841 hdr->ifi_type = dev->type;
3842 hdr->ifi_index = dev->ifindex;
3843 hdr->ifi_flags = dev_get_flags(dev);
3844 hdr->ifi_change = 0;
3846 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3849 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3851 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3852 if (dev->ifindex != dev->iflink)
3853 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3855 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3856 if (protoinfo == NULL)
3857 goto nla_put_failure;
3859 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3861 ci.max_reasm_len = IPV6_MAXPLEN;
3862 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3863 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3864 ci.reachable_time = idev->nd_parms->reachable_time;
3865 ci.retrans_time = idev->nd_parms->retrans_time;
3866 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3868 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3870 goto nla_put_failure;
3871 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3873 /* XXX - MC not implemented */
3875 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3877 goto nla_put_failure;
3878 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3880 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3882 goto nla_put_failure;
3883 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3885 nla_nest_end(skb, protoinfo);
3886 return nlmsg_end(skb, nlh);
3889 nlmsg_cancel(skb, nlh);
3893 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3895 struct net *net = sock_net(skb->sk);
3898 struct net_device *dev;
3899 struct inet6_dev *idev;
3900 struct hlist_head *head;
3901 struct hlist_node *node;
3904 s_idx = cb->args[1];
3907 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3909 head = &net->dev_index_head[h];
3910 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3913 idev = __in6_dev_get(dev);
3916 if (inet6_fill_ifinfo(skb, idev,
3917 NETLINK_CB(cb->skb).pid,
3919 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3933 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3935 struct sk_buff *skb;
3936 struct net *net = dev_net(idev->dev);
3939 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3943 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3945 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3946 WARN_ON(err == -EMSGSIZE);
3950 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3954 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3957 static inline size_t inet6_prefix_nlmsg_size(void)
3959 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3960 + nla_total_size(sizeof(struct in6_addr))
3961 + nla_total_size(sizeof(struct prefix_cacheinfo));
3964 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3965 struct prefix_info *pinfo, u32 pid, u32 seq,
3966 int event, unsigned int flags)
3968 struct prefixmsg *pmsg;
3969 struct nlmsghdr *nlh;
3970 struct prefix_cacheinfo ci;
3972 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3976 pmsg = nlmsg_data(nlh);
3977 pmsg->prefix_family = AF_INET6;
3978 pmsg->prefix_pad1 = 0;
3979 pmsg->prefix_pad2 = 0;
3980 pmsg->prefix_ifindex = idev->dev->ifindex;
3981 pmsg->prefix_len = pinfo->prefix_len;
3982 pmsg->prefix_type = pinfo->type;
3983 pmsg->prefix_pad3 = 0;
3984 pmsg->prefix_flags = 0;
3986 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3987 if (pinfo->autoconf)
3988 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3990 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3992 ci.preferred_time = ntohl(pinfo->prefered);
3993 ci.valid_time = ntohl(pinfo->valid);
3994 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3996 return nlmsg_end(skb, nlh);
3999 nlmsg_cancel(skb, nlh);
4003 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4004 struct prefix_info *pinfo)
4006 struct sk_buff *skb;
4007 struct net *net = dev_net(idev->dev);
4010 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4014 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4016 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4017 WARN_ON(err == -EMSGSIZE);
4021 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4025 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4028 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4030 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4035 * If the address was optimistic
4036 * we inserted the route at the start of
4037 * our DAD process, so we don't need
4040 if (!(ifp->rt->rt6i_node))
4041 ip6_ins_rt(ifp->rt);
4042 if (ifp->idev->cnf.forwarding)
4043 addrconf_join_anycast(ifp);
4046 if (ifp->idev->cnf.forwarding)
4047 addrconf_leave_anycast(ifp);
4048 addrconf_leave_solict(ifp->idev, &ifp->addr);
4049 dst_hold(&ifp->rt->u.dst);
4051 if (ifp->dead && ip6_del_rt(ifp->rt))
4052 dst_free(&ifp->rt->u.dst);
4057 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4060 if (likely(ifp->idev->dead == 0))
4061 __ipv6_ifa_notify(event, ifp);
4062 rcu_read_unlock_bh();
4065 #ifdef CONFIG_SYSCTL
4068 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4069 void __user *buffer, size_t *lenp, loff_t *ppos)
4071 int *valp = ctl->data;
4076 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4079 ret = addrconf_fixup_forwarding(ctl, valp, val);
4085 static void dev_disable_change(struct inet6_dev *idev)
4087 if (!idev || !idev->dev)
4090 if (idev->cnf.disable_ipv6)
4091 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4093 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4096 static void addrconf_disable_change(struct net *net, __s32 newf)
4098 struct net_device *dev;
4099 struct inet6_dev *idev;
4102 for_each_netdev_rcu(net, dev) {
4103 idev = __in6_dev_get(dev);
4105 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4106 idev->cnf.disable_ipv6 = newf;
4108 dev_disable_change(idev);
4114 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4118 net = (struct net *)table->extra2;
4120 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4123 if (!rtnl_trylock()) {
4124 /* Restore the original values before restarting */
4126 return restart_syscall();
4129 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4130 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4131 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4132 addrconf_disable_change(net, newf);
4133 } else if ((!*p) ^ (!old))
4134 dev_disable_change((struct inet6_dev *)table->extra1);
4141 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4142 void __user *buffer, size_t *lenp, loff_t *ppos)
4144 int *valp = ctl->data;
4149 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4152 ret = addrconf_disable_ipv6(ctl, valp, val);
4158 static struct addrconf_sysctl_table
4160 struct ctl_table_header *sysctl_header;
4161 ctl_table addrconf_vars[DEVCONF_MAX+1];
4163 } addrconf_sysctl __read_mostly = {
4164 .sysctl_header = NULL,
4167 .procname = "forwarding",
4168 .data = &ipv6_devconf.forwarding,
4169 .maxlen = sizeof(int),
4171 .proc_handler = addrconf_sysctl_forward,
4174 .procname = "hop_limit",
4175 .data = &ipv6_devconf.hop_limit,
4176 .maxlen = sizeof(int),
4178 .proc_handler = proc_dointvec,
4182 .data = &ipv6_devconf.mtu6,
4183 .maxlen = sizeof(int),
4185 .proc_handler = proc_dointvec,
4188 .procname = "accept_ra",
4189 .data = &ipv6_devconf.accept_ra,
4190 .maxlen = sizeof(int),
4192 .proc_handler = proc_dointvec,
4195 .procname = "accept_redirects",
4196 .data = &ipv6_devconf.accept_redirects,
4197 .maxlen = sizeof(int),
4199 .proc_handler = proc_dointvec,
4202 .procname = "autoconf",
4203 .data = &ipv6_devconf.autoconf,
4204 .maxlen = sizeof(int),
4206 .proc_handler = proc_dointvec,
4209 .procname = "dad_transmits",
4210 .data = &ipv6_devconf.dad_transmits,
4211 .maxlen = sizeof(int),
4213 .proc_handler = proc_dointvec,
4216 .procname = "router_solicitations",
4217 .data = &ipv6_devconf.rtr_solicits,
4218 .maxlen = sizeof(int),
4220 .proc_handler = proc_dointvec,
4223 .procname = "router_solicitation_interval",
4224 .data = &ipv6_devconf.rtr_solicit_interval,
4225 .maxlen = sizeof(int),
4227 .proc_handler = proc_dointvec_jiffies,
4230 .procname = "router_solicitation_delay",
4231 .data = &ipv6_devconf.rtr_solicit_delay,
4232 .maxlen = sizeof(int),
4234 .proc_handler = proc_dointvec_jiffies,
4237 .procname = "force_mld_version",
4238 .data = &ipv6_devconf.force_mld_version,
4239 .maxlen = sizeof(int),
4241 .proc_handler = proc_dointvec,
4243 #ifdef CONFIG_IPV6_PRIVACY
4245 .procname = "use_tempaddr",
4246 .data = &ipv6_devconf.use_tempaddr,
4247 .maxlen = sizeof(int),
4249 .proc_handler = proc_dointvec,
4252 .procname = "temp_valid_lft",
4253 .data = &ipv6_devconf.temp_valid_lft,
4254 .maxlen = sizeof(int),
4256 .proc_handler = proc_dointvec,
4259 .procname = "temp_prefered_lft",
4260 .data = &ipv6_devconf.temp_prefered_lft,
4261 .maxlen = sizeof(int),
4263 .proc_handler = proc_dointvec,
4266 .procname = "regen_max_retry",
4267 .data = &ipv6_devconf.regen_max_retry,
4268 .maxlen = sizeof(int),
4270 .proc_handler = proc_dointvec,
4273 .procname = "max_desync_factor",
4274 .data = &ipv6_devconf.max_desync_factor,
4275 .maxlen = sizeof(int),
4277 .proc_handler = proc_dointvec,
4281 .procname = "max_addresses",
4282 .data = &ipv6_devconf.max_addresses,
4283 .maxlen = sizeof(int),
4285 .proc_handler = proc_dointvec,
4288 .procname = "accept_ra_defrtr",
4289 .data = &ipv6_devconf.accept_ra_defrtr,
4290 .maxlen = sizeof(int),
4292 .proc_handler = proc_dointvec,
4295 .procname = "accept_ra_pinfo",
4296 .data = &ipv6_devconf.accept_ra_pinfo,
4297 .maxlen = sizeof(int),
4299 .proc_handler = proc_dointvec,
4301 #ifdef CONFIG_IPV6_ROUTER_PREF
4303 .procname = "accept_ra_rtr_pref",
4304 .data = &ipv6_devconf.accept_ra_rtr_pref,
4305 .maxlen = sizeof(int),
4307 .proc_handler = proc_dointvec,
4310 .procname = "router_probe_interval",
4311 .data = &ipv6_devconf.rtr_probe_interval,
4312 .maxlen = sizeof(int),
4314 .proc_handler = proc_dointvec_jiffies,
4316 #ifdef CONFIG_IPV6_ROUTE_INFO
4318 .procname = "accept_ra_rt_info_max_plen",
4319 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4320 .maxlen = sizeof(int),
4322 .proc_handler = proc_dointvec,
4327 .procname = "proxy_ndp",
4328 .data = &ipv6_devconf.proxy_ndp,
4329 .maxlen = sizeof(int),
4331 .proc_handler = proc_dointvec,
4334 .procname = "accept_source_route",
4335 .data = &ipv6_devconf.accept_source_route,
4336 .maxlen = sizeof(int),
4338 .proc_handler = proc_dointvec,
4340 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4342 .procname = "optimistic_dad",
4343 .data = &ipv6_devconf.optimistic_dad,
4344 .maxlen = sizeof(int),
4346 .proc_handler = proc_dointvec,
4350 #ifdef CONFIG_IPV6_MROUTE
4352 .procname = "mc_forwarding",
4353 .data = &ipv6_devconf.mc_forwarding,
4354 .maxlen = sizeof(int),
4356 .proc_handler = proc_dointvec,
4360 .procname = "disable_ipv6",
4361 .data = &ipv6_devconf.disable_ipv6,
4362 .maxlen = sizeof(int),
4364 .proc_handler = addrconf_sysctl_disable,
4367 .procname = "accept_dad",
4368 .data = &ipv6_devconf.accept_dad,
4369 .maxlen = sizeof(int),
4371 .proc_handler = proc_dointvec,
4374 .procname = "force_tllao",
4375 .data = &ipv6_devconf.force_tllao,
4376 .maxlen = sizeof(int),
4378 .proc_handler = proc_dointvec
4386 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4387 struct inet6_dev *idev, struct ipv6_devconf *p)
4390 struct addrconf_sysctl_table *t;
4392 #define ADDRCONF_CTL_PATH_DEV 3
4394 struct ctl_path addrconf_ctl_path[] = {
4395 { .procname = "net", },
4396 { .procname = "ipv6", },
4397 { .procname = "conf", },
4398 { /* to be set */ },
4403 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4407 for (i = 0; t->addrconf_vars[i].data; i++) {
4408 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4409 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4410 t->addrconf_vars[i].extra2 = net;
4414 * Make a copy of dev_name, because '.procname' is regarded as const
4415 * by sysctl and we wouldn't want anyone to change it under our feet
4416 * (see SIOCSIFNAME).
4418 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4422 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4424 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4426 if (t->sysctl_header == NULL)
4440 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4442 struct addrconf_sysctl_table *t;
4444 if (p->sysctl == NULL)
4449 unregister_sysctl_table(t->sysctl_header);
4454 static void addrconf_sysctl_register(struct inet6_dev *idev)
4456 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4457 &ndisc_ifinfo_sysctl_change);
4458 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4462 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4464 __addrconf_sysctl_unregister(&idev->cnf);
4465 neigh_sysctl_unregister(idev->nd_parms);
4471 static int __net_init addrconf_init_net(struct net *net)
4474 struct ipv6_devconf *all, *dflt;
4477 all = &ipv6_devconf;
4478 dflt = &ipv6_devconf_dflt;
4480 if (!net_eq(net, &init_net)) {
4481 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4485 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4487 goto err_alloc_dflt;
4489 /* these will be inherited by all namespaces */
4490 dflt->autoconf = ipv6_defaults.autoconf;
4491 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4494 net->ipv6.devconf_all = all;
4495 net->ipv6.devconf_dflt = dflt;
4497 #ifdef CONFIG_SYSCTL
4498 err = __addrconf_sysctl_register(net, "all", NULL, all);
4502 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4508 #ifdef CONFIG_SYSCTL
4510 __addrconf_sysctl_unregister(all);
4520 static void __net_exit addrconf_exit_net(struct net *net)
4522 #ifdef CONFIG_SYSCTL
4523 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4524 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4526 if (!net_eq(net, &init_net)) {
4527 kfree(net->ipv6.devconf_dflt);
4528 kfree(net->ipv6.devconf_all);
4532 static struct pernet_operations addrconf_ops = {
4533 .init = addrconf_init_net,
4534 .exit = addrconf_exit_net,
4541 int register_inet6addr_notifier(struct notifier_block *nb)
4543 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4545 EXPORT_SYMBOL(register_inet6addr_notifier);
4547 int unregister_inet6addr_notifier(struct notifier_block *nb)
4549 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4551 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4554 * Init / cleanup code
4557 int __init addrconf_init(void)
4561 err = ipv6_addr_label_init();
4563 printk(KERN_CRIT "IPv6 Addrconf:"
4564 " cannot initialize default policy table: %d.\n", err);
4568 register_pernet_subsys(&addrconf_ops);
4570 /* The addrconf netdev notifier requires that loopback_dev
4571 * has it's ipv6 private information allocated and setup
4572 * before it can bring up and give link-local addresses
4573 * to other devices which are up.
4575 * Unfortunately, loopback_dev is not necessarily the first
4576 * entry in the global dev_base list of net devices. In fact,
4577 * it is likely to be the very last entry on that list.
4578 * So this causes the notifier registry below to try and
4579 * give link-local addresses to all devices besides loopback_dev
4580 * first, then loopback_dev, which cases all the non-loopback_dev
4581 * devices to fail to get a link-local address.
4583 * So, as a temporary fix, allocate the ipv6 structure for
4584 * loopback_dev first by hand.
4585 * Longer term, all of the dependencies ipv6 has upon the loopback
4586 * device and it being up should be removed.
4589 if (!ipv6_add_dev(init_net.loopback_dev))
4595 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4596 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4598 register_netdevice_notifier(&ipv6_dev_notf);
4602 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4606 /* Only the first call to __rtnl_register can fail */
4607 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4608 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4609 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4610 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4611 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4613 ipv6_addr_label_rtnl_register();
4617 unregister_netdevice_notifier(&ipv6_dev_notf);
4619 unregister_pernet_subsys(&addrconf_ops);
4624 void addrconf_cleanup(void)
4626 struct net_device *dev;
4629 unregister_netdevice_notifier(&ipv6_dev_notf);
4630 unregister_pernet_subsys(&addrconf_ops);
4634 /* clean dev list */
4635 for_each_netdev(&init_net, dev) {
4636 if (__in6_dev_get(dev) == NULL)
4638 addrconf_ifdown(dev, 1);
4640 addrconf_ifdown(init_net.loopback_dev, 2);
4645 spin_lock_bh(&addrconf_hash_lock);
4646 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4647 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4648 spin_unlock_bh(&addrconf_hash_lock);
4650 del_timer(&addr_chk_timer);