IPv6: addrconf cleanup addrconf_verify
[linux-2.6/kvm.git] / net / ipv6 / addrconf.c
blob36ebb4ad8a0a1f956f8b3c57bec81643d62a1814
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
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.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
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
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
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 #ifdef CONFIG_SYSCTL
57 #include <linux/sysctl.h>
58 #endif
59 #include <linux/capability.h>
60 #include <linux/delay.h>
61 #include <linux/notifier.h>
62 #include <linux/string.h>
64 #include <net/net_namespace.h>
65 #include <net/sock.h>
66 #include <net/snmp.h>
68 #include <net/ipv6.h>
69 #include <net/protocol.h>
70 #include <net/ndisc.h>
71 #include <net/ip6_route.h>
72 #include <net/addrconf.h>
73 #include <net/tcp.h>
74 #include <net/ip.h>
75 #include <net/netlink.h>
76 #include <net/pkt_sched.h>
77 #include <linux/if_tunnel.h>
78 #include <linux/rtnetlink.h>
80 #ifdef CONFIG_IPV6_PRIVACY
81 #include <linux/random.h>
82 #endif
84 #include <linux/uaccess.h>
85 #include <asm/unaligned.h>
87 #include <linux/proc_fs.h>
88 #include <linux/seq_file.h>
90 /* Set to 3 to get tracing... */
91 #define ACONF_DEBUG 2
93 #if ACONF_DEBUG >= 3
94 #define ADBG(x) printk x
95 #else
96 #define ADBG(x)
97 #endif
99 #define INFINITY_LIFE_TIME 0xFFFFFFFF
100 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
102 #ifdef CONFIG_SYSCTL
103 static void addrconf_sysctl_register(struct inet6_dev *idev);
104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
105 #else
106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
113 #endif
115 #ifdef CONFIG_IPV6_PRIVACY
116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
118 static void ipv6_regen_rndid(unsigned long data);
120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
121 #endif
123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
124 static int ipv6_count_addresses(struct inet6_dev *idev);
127 * Configured unicast address hash table
129 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_SPINLOCK(addrconf_hash_lock);
132 static void addrconf_verify(unsigned long);
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
140 static void addrconf_type_change(struct net_device *dev,
141 unsigned long event);
142 static int addrconf_ifdown(struct net_device *dev, int how);
144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
145 static void addrconf_dad_timer(unsigned long data);
146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
147 static void addrconf_dad_run(struct inet6_dev *idev);
148 static void addrconf_rs_timer(unsigned long data);
149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
153 struct prefix_info *pinfo);
154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
155 struct net_device *dev);
157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
160 .forwarding = 0,
161 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
162 .mtu6 = IPV6_MIN_MTU,
163 .accept_ra = 1,
164 .accept_redirects = 1,
165 .autoconf = 1,
166 .force_mld_version = 0,
167 .dad_transmits = 1,
168 .rtr_solicits = MAX_RTR_SOLICITATIONS,
169 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
170 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
171 #ifdef CONFIG_IPV6_PRIVACY
172 .use_tempaddr = 0,
173 .temp_valid_lft = TEMP_VALID_LIFETIME,
174 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
175 .regen_max_retry = REGEN_MAX_RETRY,
176 .max_desync_factor = MAX_DESYNC_FACTOR,
177 #endif
178 .max_addresses = IPV6_MAX_ADDRESSES,
179 .accept_ra_defrtr = 1,
180 .accept_ra_pinfo = 1,
181 #ifdef CONFIG_IPV6_ROUTER_PREF
182 .accept_ra_rtr_pref = 1,
183 .rtr_probe_interval = 60 * HZ,
184 #ifdef CONFIG_IPV6_ROUTE_INFO
185 .accept_ra_rt_info_max_plen = 0,
186 #endif
187 #endif
188 .proxy_ndp = 0,
189 .accept_source_route = 0, /* we do not accept RH0 by default. */
190 .disable_ipv6 = 0,
191 .accept_dad = 1,
194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
195 .forwarding = 0,
196 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
197 .mtu6 = IPV6_MIN_MTU,
198 .accept_ra = 1,
199 .accept_redirects = 1,
200 .autoconf = 1,
201 .dad_transmits = 1,
202 .rtr_solicits = MAX_RTR_SOLICITATIONS,
203 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
204 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
205 #ifdef CONFIG_IPV6_PRIVACY
206 .use_tempaddr = 0,
207 .temp_valid_lft = TEMP_VALID_LIFETIME,
208 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
209 .regen_max_retry = REGEN_MAX_RETRY,
210 .max_desync_factor = MAX_DESYNC_FACTOR,
211 #endif
212 .max_addresses = IPV6_MAX_ADDRESSES,
213 .accept_ra_defrtr = 1,
214 .accept_ra_pinfo = 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216 .accept_ra_rtr_pref = 1,
217 .rtr_probe_interval = 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219 .accept_ra_rt_info_max_plen = 0,
220 #endif
221 #endif
222 .proxy_ndp = 0,
223 .accept_source_route = 0, /* we do not accept RH0 by default. */
224 .disable_ipv6 = 0,
225 .accept_dad = 1,
228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
234 /* Check if a valid qdisc is available */
235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
237 return !qdisc_tx_is_noop(dev);
240 /* Check if a route is valid prefix route */
241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
243 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
248 if (del_timer(&ifp->timer))
249 __in6_ifa_put(ifp);
252 enum addrconf_timer_t {
253 AC_NONE,
254 AC_DAD,
255 AC_RS,
258 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
259 enum addrconf_timer_t what,
260 unsigned long when)
262 if (!del_timer(&ifp->timer))
263 in6_ifa_hold(ifp);
265 switch (what) {
266 case AC_DAD:
267 ifp->timer.function = addrconf_dad_timer;
268 break;
269 case AC_RS:
270 ifp->timer.function = addrconf_rs_timer;
271 break;
272 default:
273 break;
275 ifp->timer.expires = jiffies + when;
276 add_timer(&ifp->timer);
279 static int snmp6_alloc_dev(struct inet6_dev *idev)
281 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
282 sizeof(struct ipstats_mib)) < 0)
283 goto err_ip;
284 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
285 sizeof(struct icmpv6_mib)) < 0)
286 goto err_icmp;
287 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
288 sizeof(struct icmpv6msg_mib)) < 0)
289 goto err_icmpmsg;
291 return 0;
293 err_icmpmsg:
294 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
295 err_icmp:
296 snmp_mib_free((void __percpu **)idev->stats.ipv6);
297 err_ip:
298 return -ENOMEM;
301 static void snmp6_free_dev(struct inet6_dev *idev)
303 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
304 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
305 snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 /* Nobody refers to this device, we may destroy it. */
310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
312 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
313 kfree(idev);
316 void in6_dev_finish_destroy(struct inet6_dev *idev)
318 struct net_device *dev = idev->dev;
320 WARN_ON(!list_empty(&idev->addr_list));
321 WARN_ON(idev->mc_list != NULL);
323 #ifdef NET_REFCNT_DEBUG
324 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
325 #endif
326 dev_put(dev);
327 if (!idev->dead) {
328 pr_warning("Freeing alive inet6 device %p\n", idev);
329 return;
331 snmp6_free_dev(idev);
332 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
335 EXPORT_SYMBOL(in6_dev_finish_destroy);
337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
339 struct inet6_dev *ndev;
341 ASSERT_RTNL();
343 if (dev->mtu < IPV6_MIN_MTU)
344 return NULL;
346 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
348 if (ndev == NULL)
349 return NULL;
351 rwlock_init(&ndev->lock);
352 ndev->dev = dev;
353 INIT_LIST_HEAD(&ndev->addr_list);
355 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
356 ndev->cnf.mtu6 = dev->mtu;
357 ndev->cnf.sysctl = NULL;
358 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
359 if (ndev->nd_parms == NULL) {
360 kfree(ndev);
361 return NULL;
363 if (ndev->cnf.forwarding)
364 dev_disable_lro(dev);
365 /* We refer to the device */
366 dev_hold(dev);
368 if (snmp6_alloc_dev(ndev) < 0) {
369 ADBG((KERN_WARNING
370 "%s(): cannot allocate memory for statistics; dev=%s.\n",
371 __func__, dev->name));
372 neigh_parms_release(&nd_tbl, ndev->nd_parms);
373 ndev->dead = 1;
374 in6_dev_finish_destroy(ndev);
375 return NULL;
378 if (snmp6_register_dev(ndev) < 0) {
379 ADBG((KERN_WARNING
380 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
381 __func__, dev->name));
382 neigh_parms_release(&nd_tbl, ndev->nd_parms);
383 ndev->dead = 1;
384 in6_dev_finish_destroy(ndev);
385 return NULL;
388 /* One reference from device. We must do this before
389 * we invoke __ipv6_regen_rndid().
391 in6_dev_hold(ndev);
393 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
394 ndev->cnf.accept_dad = -1;
396 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
397 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
398 printk(KERN_INFO
399 "%s: Disabled Multicast RS\n",
400 dev->name);
401 ndev->cnf.rtr_solicits = 0;
403 #endif
405 #ifdef CONFIG_IPV6_PRIVACY
406 INIT_LIST_HEAD(&ndev->tempaddr_list);
407 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
408 if ((dev->flags&IFF_LOOPBACK) ||
409 dev->type == ARPHRD_TUNNEL ||
410 dev->type == ARPHRD_TUNNEL6 ||
411 dev->type == ARPHRD_SIT ||
412 dev->type == ARPHRD_NONE) {
413 printk(KERN_INFO
414 "%s: Disabled Privacy Extensions\n",
415 dev->name);
416 ndev->cnf.use_tempaddr = -1;
417 } else {
418 in6_dev_hold(ndev);
419 ipv6_regen_rndid((unsigned long) ndev);
421 #endif
423 if (netif_running(dev) && addrconf_qdisc_ok(dev))
424 ndev->if_flags |= IF_READY;
426 ipv6_mc_init_dev(ndev);
427 ndev->tstamp = jiffies;
428 addrconf_sysctl_register(ndev);
429 /* protected by rtnl_lock */
430 rcu_assign_pointer(dev->ip6_ptr, ndev);
432 /* Join all-node multicast group */
433 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
435 return ndev;
438 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
440 struct inet6_dev *idev;
442 ASSERT_RTNL();
444 idev = __in6_dev_get(dev);
445 if (!idev) {
446 idev = ipv6_add_dev(dev);
447 if (!idev)
448 return NULL;
451 if (dev->flags&IFF_UP)
452 ipv6_mc_up(idev);
453 return idev;
456 #ifdef CONFIG_SYSCTL
457 static void dev_forward_change(struct inet6_dev *idev)
459 struct net_device *dev;
460 struct inet6_ifaddr *ifa;
462 if (!idev)
463 return;
464 dev = idev->dev;
465 if (idev->cnf.forwarding)
466 dev_disable_lro(dev);
467 if (dev && (dev->flags & IFF_MULTICAST)) {
468 if (idev->cnf.forwarding)
469 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
470 else
471 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
474 list_for_each_entry(ifa, &idev->addr_list, if_list) {
475 if (ifa->flags&IFA_F_TENTATIVE)
476 continue;
477 if (idev->cnf.forwarding)
478 addrconf_join_anycast(ifa);
479 else
480 addrconf_leave_anycast(ifa);
485 static void addrconf_forward_change(struct net *net, __s32 newf)
487 struct net_device *dev;
488 struct inet6_dev *idev;
490 rcu_read_lock();
491 for_each_netdev_rcu(net, dev) {
492 idev = __in6_dev_get(dev);
493 if (idev) {
494 int changed = (!idev->cnf.forwarding) ^ (!newf);
495 idev->cnf.forwarding = newf;
496 if (changed)
497 dev_forward_change(idev);
500 rcu_read_unlock();
503 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
505 struct net *net;
507 net = (struct net *)table->extra2;
508 if (p == &net->ipv6.devconf_dflt->forwarding)
509 return 0;
511 if (!rtnl_trylock()) {
512 /* Restore the original values before restarting */
513 *p = old;
514 return restart_syscall();
517 if (p == &net->ipv6.devconf_all->forwarding) {
518 __s32 newf = net->ipv6.devconf_all->forwarding;
519 net->ipv6.devconf_dflt->forwarding = newf;
520 addrconf_forward_change(net, newf);
521 } else if ((!*p) ^ (!old))
522 dev_forward_change((struct inet6_dev *)table->extra1);
523 rtnl_unlock();
525 if (*p)
526 rt6_purge_dflt_routers(net);
527 return 1;
529 #endif
531 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
533 struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
534 kfree(ifp);
537 /* Nobody refers to this ifaddr, destroy it */
538 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
540 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
542 #ifdef NET_REFCNT_DEBUG
543 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
544 #endif
546 in6_dev_put(ifp->idev);
548 if (del_timer(&ifp->timer))
549 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
551 if (!ifp->dead) {
552 pr_warning("Freeing alive inet6 address %p\n", ifp);
553 return;
555 dst_release(&ifp->rt->u.dst);
557 call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
560 static void
561 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
563 struct list_head *p;
564 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
567 * Each device address list is sorted in order of scope -
568 * global before linklocal.
570 list_for_each(p, &idev->addr_list) {
571 struct inet6_ifaddr *ifa
572 = list_entry(p, struct inet6_ifaddr, if_list);
573 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
574 break;
577 list_add(&ifp->if_list, p);
580 static u32 ipv6_addr_hash(const struct in6_addr *addr)
583 * We perform the hash function over the last 64 bits of the address
584 * This will include the IEEE address token on links that support it.
586 return jhash_2words(addr->s6_addr32[2], addr->s6_addr32[3], 0)
587 & (IN6_ADDR_HSIZE - 1);
590 /* On success it returns ifp with increased reference count */
592 static struct inet6_ifaddr *
593 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
594 int scope, u32 flags)
596 struct inet6_ifaddr *ifa = NULL;
597 struct rt6_info *rt;
598 unsigned int hash;
599 int err = 0;
600 int addr_type = ipv6_addr_type(addr);
602 if (addr_type == IPV6_ADDR_ANY ||
603 addr_type & IPV6_ADDR_MULTICAST ||
604 (!(idev->dev->flags & IFF_LOOPBACK) &&
605 addr_type & IPV6_ADDR_LOOPBACK))
606 return ERR_PTR(-EADDRNOTAVAIL);
608 rcu_read_lock_bh();
609 if (idev->dead) {
610 err = -ENODEV; /*XXX*/
611 goto out2;
614 if (idev->cnf.disable_ipv6) {
615 err = -EACCES;
616 goto out2;
619 spin_lock(&addrconf_hash_lock);
621 /* Ignore adding duplicate addresses on an interface */
622 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
623 ADBG(("ipv6_add_addr: already assigned\n"));
624 err = -EEXIST;
625 goto out;
628 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
630 if (ifa == NULL) {
631 ADBG(("ipv6_add_addr: malloc failed\n"));
632 err = -ENOBUFS;
633 goto out;
636 rt = addrconf_dst_alloc(idev, addr, 0);
637 if (IS_ERR(rt)) {
638 err = PTR_ERR(rt);
639 goto out;
642 ipv6_addr_copy(&ifa->addr, addr);
644 spin_lock_init(&ifa->lock);
645 init_timer(&ifa->timer);
646 INIT_HLIST_NODE(&ifa->addr_lst);
647 ifa->timer.data = (unsigned long) ifa;
648 ifa->scope = scope;
649 ifa->prefix_len = pfxlen;
650 ifa->flags = flags | IFA_F_TENTATIVE;
651 ifa->cstamp = ifa->tstamp = jiffies;
653 ifa->rt = rt;
656 * part one of RFC 4429, section 3.3
657 * We should not configure an address as
658 * optimistic if we do not yet know the link
659 * layer address of our nexhop router
662 if (rt->rt6i_nexthop == NULL)
663 ifa->flags &= ~IFA_F_OPTIMISTIC;
665 ifa->idev = idev;
666 in6_dev_hold(idev);
667 /* For caller */
668 in6_ifa_hold(ifa);
670 /* Add to big hash table */
671 hash = ipv6_addr_hash(addr);
673 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
674 in6_ifa_hold(ifa);
675 spin_unlock(&addrconf_hash_lock);
677 write_lock(&idev->lock);
678 /* Add to inet6_dev unicast addr list. */
679 ipv6_link_dev_addr(idev, ifa);
681 #ifdef CONFIG_IPV6_PRIVACY
682 if (ifa->flags&IFA_F_TEMPORARY) {
683 list_add(&ifa->tmp_list, &idev->tempaddr_list);
684 in6_ifa_hold(ifa);
686 #endif
688 in6_ifa_hold(ifa);
689 write_unlock(&idev->lock);
690 out2:
691 rcu_read_unlock_bh();
693 if (likely(err == 0))
694 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
695 else {
696 kfree(ifa);
697 ifa = ERR_PTR(err);
700 return ifa;
701 out:
702 spin_unlock(&addrconf_hash_lock);
703 goto out2;
706 /* This function wants to get referenced ifp and releases it before return */
708 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
710 struct inet6_ifaddr *ifa, *ifn;
711 struct inet6_dev *idev = ifp->idev;
712 int hash;
713 int deleted = 0, onlink = 0;
714 unsigned long expires = jiffies;
716 hash = ipv6_addr_hash(&ifp->addr);
718 ifp->dead = 1;
720 spin_lock_bh(&addrconf_hash_lock);
721 hlist_del_init_rcu(&ifp->addr_lst);
722 __in6_ifa_put(ifp);
723 spin_unlock_bh(&addrconf_hash_lock);
725 write_lock_bh(&idev->lock);
726 #ifdef CONFIG_IPV6_PRIVACY
727 if (ifp->flags&IFA_F_TEMPORARY) {
728 list_del(&ifp->tmp_list);
729 if (ifp->ifpub) {
730 in6_ifa_put(ifp->ifpub);
731 ifp->ifpub = NULL;
733 __in6_ifa_put(ifp);
735 #endif
737 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
738 if (ifa == ifp) {
739 list_del_init(&ifp->if_list);
740 __in6_ifa_put(ifp);
742 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
743 break;
744 deleted = 1;
745 continue;
746 } else if (ifp->flags & IFA_F_PERMANENT) {
747 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
748 ifp->prefix_len)) {
749 if (ifa->flags & IFA_F_PERMANENT) {
750 onlink = 1;
751 if (deleted)
752 break;
753 } else {
754 unsigned long lifetime;
756 if (!onlink)
757 onlink = -1;
759 spin_lock(&ifa->lock);
761 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
763 * Note: Because this address is
764 * not permanent, lifetime <
765 * LONG_MAX / HZ here.
767 if (time_before(expires,
768 ifa->tstamp + lifetime * HZ))
769 expires = ifa->tstamp + lifetime * HZ;
770 spin_unlock(&ifa->lock);
775 write_unlock_bh(&idev->lock);
777 addrconf_del_timer(ifp);
779 ipv6_ifa_notify(RTM_DELADDR, ifp);
781 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
784 * Purge or update corresponding prefix
786 * 1) we don't purge prefix here if address was not permanent.
787 * prefix is managed by its own lifetime.
788 * 2) if there're no addresses, delete prefix.
789 * 3) if there're still other permanent address(es),
790 * corresponding prefix is still permanent.
791 * 4) otherwise, update prefix lifetime to the
792 * longest valid lifetime among the corresponding
793 * addresses on the device.
794 * Note: subsequent RA will update lifetime.
796 * --yoshfuji
798 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
799 struct in6_addr prefix;
800 struct rt6_info *rt;
801 struct net *net = dev_net(ifp->idev->dev);
802 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
803 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
805 if (rt && addrconf_is_prefix_route(rt)) {
806 if (onlink == 0) {
807 ip6_del_rt(rt);
808 rt = NULL;
809 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
810 rt->rt6i_expires = expires;
811 rt->rt6i_flags |= RTF_EXPIRES;
814 dst_release(&rt->u.dst);
817 in6_ifa_put(ifp);
820 #ifdef CONFIG_IPV6_PRIVACY
821 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
823 struct inet6_dev *idev = ifp->idev;
824 struct in6_addr addr, *tmpaddr;
825 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
826 unsigned long regen_advance;
827 int tmp_plen;
828 int ret = 0;
829 int max_addresses;
830 u32 addr_flags;
832 write_lock(&idev->lock);
833 if (ift) {
834 spin_lock_bh(&ift->lock);
835 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
836 spin_unlock_bh(&ift->lock);
837 tmpaddr = &addr;
838 } else {
839 tmpaddr = NULL;
841 retry:
842 in6_dev_hold(idev);
843 if (idev->cnf.use_tempaddr <= 0) {
844 write_unlock(&idev->lock);
845 printk(KERN_INFO
846 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
847 in6_dev_put(idev);
848 ret = -1;
849 goto out;
851 spin_lock_bh(&ifp->lock);
852 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
853 idev->cnf.use_tempaddr = -1; /*XXX*/
854 spin_unlock_bh(&ifp->lock);
855 write_unlock(&idev->lock);
856 printk(KERN_WARNING
857 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
858 in6_dev_put(idev);
859 ret = -1;
860 goto out;
862 in6_ifa_hold(ifp);
863 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
864 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
865 spin_unlock_bh(&ifp->lock);
866 write_unlock(&idev->lock);
867 printk(KERN_WARNING
868 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
869 in6_ifa_put(ifp);
870 in6_dev_put(idev);
871 ret = -1;
872 goto out;
874 memcpy(&addr.s6_addr[8], idev->rndid, 8);
875 tmp_valid_lft = min_t(__u32,
876 ifp->valid_lft,
877 idev->cnf.temp_valid_lft);
878 tmp_prefered_lft = min_t(__u32,
879 ifp->prefered_lft,
880 idev->cnf.temp_prefered_lft - desync_factor / HZ);
881 tmp_plen = ifp->prefix_len;
882 max_addresses = idev->cnf.max_addresses;
883 tmp_cstamp = ifp->cstamp;
884 tmp_tstamp = ifp->tstamp;
885 spin_unlock_bh(&ifp->lock);
887 regen_advance = idev->cnf.regen_max_retry *
888 idev->cnf.dad_transmits *
889 idev->nd_parms->retrans_time / HZ;
890 write_unlock(&idev->lock);
892 /* A temporary address is created only if this calculated Preferred
893 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
894 * an implementation must not create a temporary address with a zero
895 * Preferred Lifetime.
897 if (tmp_prefered_lft <= regen_advance) {
898 in6_ifa_put(ifp);
899 in6_dev_put(idev);
900 ret = -1;
901 goto out;
904 addr_flags = IFA_F_TEMPORARY;
905 /* set in addrconf_prefix_rcv() */
906 if (ifp->flags & IFA_F_OPTIMISTIC)
907 addr_flags |= IFA_F_OPTIMISTIC;
909 ift = !max_addresses ||
910 ipv6_count_addresses(idev) < max_addresses ?
911 ipv6_add_addr(idev, &addr, tmp_plen,
912 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
913 addr_flags) : NULL;
914 if (!ift || IS_ERR(ift)) {
915 in6_ifa_put(ifp);
916 in6_dev_put(idev);
917 printk(KERN_INFO
918 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
919 tmpaddr = &addr;
920 write_lock(&idev->lock);
921 goto retry;
924 spin_lock_bh(&ift->lock);
925 ift->ifpub = ifp;
926 ift->valid_lft = tmp_valid_lft;
927 ift->prefered_lft = tmp_prefered_lft;
928 ift->cstamp = tmp_cstamp;
929 ift->tstamp = tmp_tstamp;
930 spin_unlock_bh(&ift->lock);
932 addrconf_dad_start(ift, 0);
933 in6_ifa_put(ift);
934 in6_dev_put(idev);
935 out:
936 return ret;
938 #endif
941 * Choose an appropriate source address (RFC3484)
943 enum {
944 IPV6_SADDR_RULE_INIT = 0,
945 IPV6_SADDR_RULE_LOCAL,
946 IPV6_SADDR_RULE_SCOPE,
947 IPV6_SADDR_RULE_PREFERRED,
948 #ifdef CONFIG_IPV6_MIP6
949 IPV6_SADDR_RULE_HOA,
950 #endif
951 IPV6_SADDR_RULE_OIF,
952 IPV6_SADDR_RULE_LABEL,
953 #ifdef CONFIG_IPV6_PRIVACY
954 IPV6_SADDR_RULE_PRIVACY,
955 #endif
956 IPV6_SADDR_RULE_ORCHID,
957 IPV6_SADDR_RULE_PREFIX,
958 IPV6_SADDR_RULE_MAX
961 struct ipv6_saddr_score {
962 int rule;
963 int addr_type;
964 struct inet6_ifaddr *ifa;
965 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
966 int scopedist;
967 int matchlen;
970 struct ipv6_saddr_dst {
971 const struct in6_addr *addr;
972 int ifindex;
973 int scope;
974 int label;
975 unsigned int prefs;
978 static inline int ipv6_saddr_preferred(int type)
980 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
981 return 1;
982 return 0;
985 static int ipv6_get_saddr_eval(struct net *net,
986 struct ipv6_saddr_score *score,
987 struct ipv6_saddr_dst *dst,
988 int i)
990 int ret;
992 if (i <= score->rule) {
993 switch (i) {
994 case IPV6_SADDR_RULE_SCOPE:
995 ret = score->scopedist;
996 break;
997 case IPV6_SADDR_RULE_PREFIX:
998 ret = score->matchlen;
999 break;
1000 default:
1001 ret = !!test_bit(i, score->scorebits);
1003 goto out;
1006 switch (i) {
1007 case IPV6_SADDR_RULE_INIT:
1008 /* Rule 0: remember if hiscore is not ready yet */
1009 ret = !!score->ifa;
1010 break;
1011 case IPV6_SADDR_RULE_LOCAL:
1012 /* Rule 1: Prefer same address */
1013 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1014 break;
1015 case IPV6_SADDR_RULE_SCOPE:
1016 /* Rule 2: Prefer appropriate scope
1018 * ret
1020 * -1 | d 15
1021 * ---+--+-+---> scope
1023 * | d is scope of the destination.
1024 * B-d | \
1025 * | \ <- smaller scope is better if
1026 * B-15 | \ if scope is enough for destinaion.
1027 * | ret = B - scope (-1 <= scope >= d <= 15).
1028 * d-C-1 | /
1029 * |/ <- greater is better
1030 * -C / if scope is not enough for destination.
1031 * /| ret = scope - C (-1 <= d < scope <= 15).
1033 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1034 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1035 * Assume B = 0 and we get C > 29.
1037 ret = __ipv6_addr_src_scope(score->addr_type);
1038 if (ret >= dst->scope)
1039 ret = -ret;
1040 else
1041 ret -= 128; /* 30 is enough */
1042 score->scopedist = ret;
1043 break;
1044 case IPV6_SADDR_RULE_PREFERRED:
1045 /* Rule 3: Avoid deprecated and optimistic addresses */
1046 ret = ipv6_saddr_preferred(score->addr_type) ||
1047 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1048 break;
1049 #ifdef CONFIG_IPV6_MIP6
1050 case IPV6_SADDR_RULE_HOA:
1052 /* Rule 4: Prefer home address */
1053 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1054 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1055 break;
1057 #endif
1058 case IPV6_SADDR_RULE_OIF:
1059 /* Rule 5: Prefer outgoing interface */
1060 ret = (!dst->ifindex ||
1061 dst->ifindex == score->ifa->idev->dev->ifindex);
1062 break;
1063 case IPV6_SADDR_RULE_LABEL:
1064 /* Rule 6: Prefer matching label */
1065 ret = ipv6_addr_label(net,
1066 &score->ifa->addr, score->addr_type,
1067 score->ifa->idev->dev->ifindex) == dst->label;
1068 break;
1069 #ifdef CONFIG_IPV6_PRIVACY
1070 case IPV6_SADDR_RULE_PRIVACY:
1072 /* Rule 7: Prefer public address
1073 * Note: prefer temprary address if use_tempaddr >= 2
1075 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1076 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1077 score->ifa->idev->cnf.use_tempaddr >= 2;
1078 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1079 break;
1081 #endif
1082 case IPV6_SADDR_RULE_ORCHID:
1083 /* Rule 8-: Prefer ORCHID vs ORCHID or
1084 * non-ORCHID vs non-ORCHID
1086 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1087 ipv6_addr_orchid(dst->addr));
1088 break;
1089 case IPV6_SADDR_RULE_PREFIX:
1090 /* Rule 8: Use longest matching prefix */
1091 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1092 dst->addr);
1093 break;
1094 default:
1095 ret = 0;
1098 if (ret)
1099 __set_bit(i, score->scorebits);
1100 score->rule = i;
1101 out:
1102 return ret;
1105 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1106 const struct in6_addr *daddr, unsigned int prefs,
1107 struct in6_addr *saddr)
1109 struct ipv6_saddr_score scores[2],
1110 *score = &scores[0], *hiscore = &scores[1];
1111 struct ipv6_saddr_dst dst;
1112 struct net_device *dev;
1113 int dst_type;
1115 dst_type = __ipv6_addr_type(daddr);
1116 dst.addr = daddr;
1117 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1118 dst.scope = __ipv6_addr_src_scope(dst_type);
1119 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1120 dst.prefs = prefs;
1122 hiscore->rule = -1;
1123 hiscore->ifa = NULL;
1125 rcu_read_lock();
1127 for_each_netdev_rcu(net, dev) {
1128 struct inet6_dev *idev;
1130 /* Candidate Source Address (section 4)
1131 * - multicast and link-local destination address,
1132 * the set of candidate source address MUST only
1133 * include addresses assigned to interfaces
1134 * belonging to the same link as the outgoing
1135 * interface.
1136 * (- For site-local destination addresses, the
1137 * set of candidate source addresses MUST only
1138 * include addresses assigned to interfaces
1139 * belonging to the same site as the outgoing
1140 * interface.)
1142 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1143 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1144 dst.ifindex && dev->ifindex != dst.ifindex)
1145 continue;
1147 idev = __in6_dev_get(dev);
1148 if (!idev)
1149 continue;
1151 read_lock_bh(&idev->lock);
1152 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1153 int i;
1156 * - Tentative Address (RFC2462 section 5.4)
1157 * - A tentative address is not considered
1158 * "assigned to an interface" in the traditional
1159 * sense, unless it is also flagged as optimistic.
1160 * - Candidate Source Address (section 4)
1161 * - In any case, anycast addresses, multicast
1162 * addresses, and the unspecified address MUST
1163 * NOT be included in a candidate set.
1165 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1166 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1167 continue;
1169 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1171 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1172 score->addr_type & IPV6_ADDR_MULTICAST)) {
1173 LIMIT_NETDEBUG(KERN_DEBUG
1174 "ADDRCONF: unspecified / multicast address "
1175 "assigned as unicast address on %s",
1176 dev->name);
1177 continue;
1180 score->rule = -1;
1181 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1183 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1184 int minihiscore, miniscore;
1186 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1187 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1189 if (minihiscore > miniscore) {
1190 if (i == IPV6_SADDR_RULE_SCOPE &&
1191 score->scopedist > 0) {
1193 * special case:
1194 * each remaining entry
1195 * has too small (not enough)
1196 * scope, because ifa entries
1197 * are sorted by their scope
1198 * values.
1200 goto try_nextdev;
1202 break;
1203 } else if (minihiscore < miniscore) {
1204 if (hiscore->ifa)
1205 in6_ifa_put(hiscore->ifa);
1207 in6_ifa_hold(score->ifa);
1209 swap(hiscore, score);
1211 /* restore our iterator */
1212 score->ifa = hiscore->ifa;
1214 break;
1218 try_nextdev:
1219 read_unlock_bh(&idev->lock);
1221 rcu_read_unlock();
1223 if (!hiscore->ifa)
1224 return -EADDRNOTAVAIL;
1226 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1227 in6_ifa_put(hiscore->ifa);
1228 return 0;
1230 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1232 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1233 unsigned char banned_flags)
1235 struct inet6_dev *idev;
1236 int err = -EADDRNOTAVAIL;
1238 rcu_read_lock();
1239 idev = __in6_dev_get(dev);
1240 if (idev) {
1241 struct inet6_ifaddr *ifp;
1243 read_lock_bh(&idev->lock);
1244 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1245 if (ifp->scope == IFA_LINK &&
1246 !(ifp->flags & banned_flags)) {
1247 ipv6_addr_copy(addr, &ifp->addr);
1248 err = 0;
1249 break;
1252 read_unlock_bh(&idev->lock);
1254 rcu_read_unlock();
1255 return err;
1258 static int ipv6_count_addresses(struct inet6_dev *idev)
1260 int cnt = 0;
1261 struct inet6_ifaddr *ifp;
1263 read_lock_bh(&idev->lock);
1264 list_for_each_entry(ifp, &idev->addr_list, if_list)
1265 cnt++;
1266 read_unlock_bh(&idev->lock);
1267 return cnt;
1270 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1271 struct net_device *dev, int strict)
1273 struct inet6_ifaddr *ifp = NULL;
1274 struct hlist_node *node;
1275 unsigned int hash = ipv6_addr_hash(addr);
1277 rcu_read_lock_bh();
1278 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1279 if (!net_eq(dev_net(ifp->idev->dev), net))
1280 continue;
1281 if (ipv6_addr_equal(&ifp->addr, addr) &&
1282 !(ifp->flags&IFA_F_TENTATIVE)) {
1283 if (dev == NULL || ifp->idev->dev == dev ||
1284 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1285 break;
1288 rcu_read_unlock_bh();
1290 return ifp != NULL;
1292 EXPORT_SYMBOL(ipv6_chk_addr);
1294 static
1295 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1296 struct net_device *dev)
1298 struct inet6_ifaddr *ifp;
1299 struct hlist_node *node;
1300 unsigned int hash = ipv6_addr_hash(addr);
1302 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1303 if (!net_eq(dev_net(ifp->idev->dev), net))
1304 continue;
1305 if (ipv6_addr_equal(&ifp->addr, addr)) {
1306 if (dev == NULL || ifp->idev->dev == dev)
1307 break;
1310 return ifp != NULL;
1313 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1315 struct inet6_dev *idev;
1316 struct inet6_ifaddr *ifa;
1317 int onlink;
1319 onlink = 0;
1320 rcu_read_lock();
1321 idev = __in6_dev_get(dev);
1322 if (idev) {
1323 read_lock_bh(&idev->lock);
1324 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1325 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1326 ifa->prefix_len);
1327 if (onlink)
1328 break;
1330 read_unlock_bh(&idev->lock);
1332 rcu_read_unlock();
1333 return onlink;
1336 EXPORT_SYMBOL(ipv6_chk_prefix);
1338 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1339 struct net_device *dev, int strict)
1341 struct inet6_ifaddr *ifp = NULL;
1342 struct hlist_node *node;
1343 unsigned int hash = ipv6_addr_hash(addr);
1345 rcu_read_lock_bh();
1346 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1347 if (!net_eq(dev_net(ifp->idev->dev), net))
1348 continue;
1349 if (ipv6_addr_equal(&ifp->addr, addr)) {
1350 if (dev == NULL || ifp->idev->dev == dev ||
1351 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1352 in6_ifa_hold(ifp);
1353 break;
1357 rcu_read_unlock_bh();
1359 return ifp;
1362 /* Gets referenced address, destroys ifaddr */
1364 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1366 if (ifp->flags&IFA_F_PERMANENT) {
1367 spin_lock_bh(&ifp->lock);
1368 addrconf_del_timer(ifp);
1369 ifp->flags |= IFA_F_TENTATIVE;
1370 if (dad_failed)
1371 ifp->flags |= IFA_F_DADFAILED;
1372 spin_unlock_bh(&ifp->lock);
1373 if (dad_failed)
1374 ipv6_ifa_notify(0, ifp);
1375 in6_ifa_put(ifp);
1376 #ifdef CONFIG_IPV6_PRIVACY
1377 } else if (ifp->flags&IFA_F_TEMPORARY) {
1378 struct inet6_ifaddr *ifpub;
1379 spin_lock_bh(&ifp->lock);
1380 ifpub = ifp->ifpub;
1381 if (ifpub) {
1382 in6_ifa_hold(ifpub);
1383 spin_unlock_bh(&ifp->lock);
1384 ipv6_create_tempaddr(ifpub, ifp);
1385 in6_ifa_put(ifpub);
1386 } else {
1387 spin_unlock_bh(&ifp->lock);
1389 ipv6_del_addr(ifp);
1390 #endif
1391 } else
1392 ipv6_del_addr(ifp);
1395 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1397 struct inet6_dev *idev = ifp->idev;
1399 if (net_ratelimit())
1400 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1401 ifp->idev->dev->name, &ifp->addr);
1403 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1404 struct in6_addr addr;
1406 addr.s6_addr32[0] = htonl(0xfe800000);
1407 addr.s6_addr32[1] = 0;
1409 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1410 ipv6_addr_equal(&ifp->addr, &addr)) {
1411 /* DAD failed for link-local based on MAC address */
1412 idev->cnf.disable_ipv6 = 1;
1414 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1415 ifp->idev->dev->name);
1419 addrconf_dad_stop(ifp, 1);
1422 /* Join to solicited addr multicast group. */
1424 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1426 struct in6_addr maddr;
1428 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1429 return;
1431 addrconf_addr_solict_mult(addr, &maddr);
1432 ipv6_dev_mc_inc(dev, &maddr);
1435 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1437 struct in6_addr maddr;
1439 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1440 return;
1442 addrconf_addr_solict_mult(addr, &maddr);
1443 __ipv6_dev_mc_dec(idev, &maddr);
1446 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1448 struct in6_addr addr;
1449 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1450 if (ipv6_addr_any(&addr))
1451 return;
1452 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1455 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1457 struct in6_addr addr;
1458 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1459 if (ipv6_addr_any(&addr))
1460 return;
1461 __ipv6_dev_ac_dec(ifp->idev, &addr);
1464 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1466 if (dev->addr_len != ETH_ALEN)
1467 return -1;
1468 memcpy(eui, dev->dev_addr, 3);
1469 memcpy(eui + 5, dev->dev_addr + 3, 3);
1472 * The zSeries OSA network cards can be shared among various
1473 * OS instances, but the OSA cards have only one MAC address.
1474 * This leads to duplicate address conflicts in conjunction
1475 * with IPv6 if more than one instance uses the same card.
1477 * The driver for these cards can deliver a unique 16-bit
1478 * identifier for each instance sharing the same card. It is
1479 * placed instead of 0xFFFE in the interface identifier. The
1480 * "u" bit of the interface identifier is not inverted in this
1481 * case. Hence the resulting interface identifier has local
1482 * scope according to RFC2373.
1484 if (dev->dev_id) {
1485 eui[3] = (dev->dev_id >> 8) & 0xFF;
1486 eui[4] = dev->dev_id & 0xFF;
1487 } else {
1488 eui[3] = 0xFF;
1489 eui[4] = 0xFE;
1490 eui[0] ^= 2;
1492 return 0;
1495 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1497 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1498 if (dev->addr_len != ARCNET_ALEN)
1499 return -1;
1500 memset(eui, 0, 7);
1501 eui[7] = *(u8*)dev->dev_addr;
1502 return 0;
1505 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1507 if (dev->addr_len != INFINIBAND_ALEN)
1508 return -1;
1509 memcpy(eui, dev->dev_addr + 12, 8);
1510 eui[0] |= 2;
1511 return 0;
1514 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1516 if (addr == 0)
1517 return -1;
1518 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1519 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1520 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1521 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1522 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1523 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1524 eui[1] = 0;
1525 eui[2] = 0x5E;
1526 eui[3] = 0xFE;
1527 memcpy(eui + 4, &addr, 4);
1528 return 0;
1530 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1532 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1534 if (dev->priv_flags & IFF_ISATAP)
1535 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1536 return -1;
1539 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1541 switch (dev->type) {
1542 case ARPHRD_ETHER:
1543 case ARPHRD_FDDI:
1544 case ARPHRD_IEEE802_TR:
1545 return addrconf_ifid_eui48(eui, dev);
1546 case ARPHRD_ARCNET:
1547 return addrconf_ifid_arcnet(eui, dev);
1548 case ARPHRD_INFINIBAND:
1549 return addrconf_ifid_infiniband(eui, dev);
1550 case ARPHRD_SIT:
1551 return addrconf_ifid_sit(eui, dev);
1553 return -1;
1556 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1558 int err = -1;
1559 struct inet6_ifaddr *ifp;
1561 read_lock_bh(&idev->lock);
1562 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1563 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1564 memcpy(eui, ifp->addr.s6_addr+8, 8);
1565 err = 0;
1566 break;
1569 read_unlock_bh(&idev->lock);
1570 return err;
1573 #ifdef CONFIG_IPV6_PRIVACY
1574 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1575 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1577 regen:
1578 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1579 idev->rndid[0] &= ~0x02;
1582 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1583 * check if generated address is not inappropriate
1585 * - Reserved subnet anycast (RFC 2526)
1586 * 11111101 11....11 1xxxxxxx
1587 * - ISATAP (RFC4214) 6.1
1588 * 00-00-5E-FE-xx-xx-xx-xx
1589 * - value 0
1590 * - XXX: already assigned to an address on the device
1592 if (idev->rndid[0] == 0xfd &&
1593 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1594 (idev->rndid[7]&0x80))
1595 goto regen;
1596 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1597 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1598 goto regen;
1599 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1600 goto regen;
1603 return 0;
1606 static void ipv6_regen_rndid(unsigned long data)
1608 struct inet6_dev *idev = (struct inet6_dev *) data;
1609 unsigned long expires;
1611 rcu_read_lock_bh();
1612 write_lock_bh(&idev->lock);
1614 if (idev->dead)
1615 goto out;
1617 if (__ipv6_regen_rndid(idev) < 0)
1618 goto out;
1620 expires = jiffies +
1621 idev->cnf.temp_prefered_lft * HZ -
1622 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1623 if (time_before(expires, jiffies)) {
1624 printk(KERN_WARNING
1625 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1626 idev->dev->name);
1627 goto out;
1630 if (!mod_timer(&idev->regen_timer, expires))
1631 in6_dev_hold(idev);
1633 out:
1634 write_unlock_bh(&idev->lock);
1635 rcu_read_unlock_bh();
1636 in6_dev_put(idev);
1639 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1640 int ret = 0;
1642 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1643 ret = __ipv6_regen_rndid(idev);
1644 return ret;
1646 #endif
1649 * Add prefix route.
1652 static void
1653 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1654 unsigned long expires, u32 flags)
1656 struct fib6_config cfg = {
1657 .fc_table = RT6_TABLE_PREFIX,
1658 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1659 .fc_ifindex = dev->ifindex,
1660 .fc_expires = expires,
1661 .fc_dst_len = plen,
1662 .fc_flags = RTF_UP | flags,
1663 .fc_nlinfo.nl_net = dev_net(dev),
1664 .fc_protocol = RTPROT_KERNEL,
1667 ipv6_addr_copy(&cfg.fc_dst, pfx);
1669 /* Prevent useless cloning on PtP SIT.
1670 This thing is done here expecting that the whole
1671 class of non-broadcast devices need not cloning.
1673 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1674 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1675 cfg.fc_flags |= RTF_NONEXTHOP;
1676 #endif
1678 ip6_route_add(&cfg);
1681 /* Create "default" multicast route to the interface */
1683 static void addrconf_add_mroute(struct net_device *dev)
1685 struct fib6_config cfg = {
1686 .fc_table = RT6_TABLE_LOCAL,
1687 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1688 .fc_ifindex = dev->ifindex,
1689 .fc_dst_len = 8,
1690 .fc_flags = RTF_UP,
1691 .fc_nlinfo.nl_net = dev_net(dev),
1694 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1696 ip6_route_add(&cfg);
1699 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1700 static void sit_route_add(struct net_device *dev)
1702 struct fib6_config cfg = {
1703 .fc_table = RT6_TABLE_MAIN,
1704 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1705 .fc_ifindex = dev->ifindex,
1706 .fc_dst_len = 96,
1707 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1708 .fc_nlinfo.nl_net = dev_net(dev),
1711 /* prefix length - 96 bits "::d.d.d.d" */
1712 ip6_route_add(&cfg);
1714 #endif
1716 static void addrconf_add_lroute(struct net_device *dev)
1718 struct in6_addr addr;
1720 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1721 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1724 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1726 struct inet6_dev *idev;
1728 ASSERT_RTNL();
1730 idev = ipv6_find_idev(dev);
1731 if (!idev)
1732 return NULL;
1734 /* Add default multicast route */
1735 addrconf_add_mroute(dev);
1737 /* Add link local route */
1738 addrconf_add_lroute(dev);
1739 return idev;
1742 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1744 struct prefix_info *pinfo;
1745 __u32 valid_lft;
1746 __u32 prefered_lft;
1747 int addr_type;
1748 struct inet6_dev *in6_dev;
1749 struct net *net = dev_net(dev);
1751 pinfo = (struct prefix_info *) opt;
1753 if (len < sizeof(struct prefix_info)) {
1754 ADBG(("addrconf: prefix option too short\n"));
1755 return;
1759 * Validation checks ([ADDRCONF], page 19)
1762 addr_type = ipv6_addr_type(&pinfo->prefix);
1764 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1765 return;
1767 valid_lft = ntohl(pinfo->valid);
1768 prefered_lft = ntohl(pinfo->prefered);
1770 if (prefered_lft > valid_lft) {
1771 if (net_ratelimit())
1772 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1773 return;
1776 in6_dev = in6_dev_get(dev);
1778 if (in6_dev == NULL) {
1779 if (net_ratelimit())
1780 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1781 return;
1785 * Two things going on here:
1786 * 1) Add routes for on-link prefixes
1787 * 2) Configure prefixes with the auto flag set
1790 if (pinfo->onlink) {
1791 struct rt6_info *rt;
1792 unsigned long rt_expires;
1794 /* Avoid arithmetic overflow. Really, we could
1795 * save rt_expires in seconds, likely valid_lft,
1796 * but it would require division in fib gc, that it
1797 * not good.
1799 if (HZ > USER_HZ)
1800 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1801 else
1802 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1804 if (addrconf_finite_timeout(rt_expires))
1805 rt_expires *= HZ;
1807 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1808 dev->ifindex, 1);
1810 if (rt && addrconf_is_prefix_route(rt)) {
1811 /* Autoconf prefix route */
1812 if (valid_lft == 0) {
1813 ip6_del_rt(rt);
1814 rt = NULL;
1815 } else if (addrconf_finite_timeout(rt_expires)) {
1816 /* not infinity */
1817 rt->rt6i_expires = jiffies + rt_expires;
1818 rt->rt6i_flags |= RTF_EXPIRES;
1819 } else {
1820 rt->rt6i_flags &= ~RTF_EXPIRES;
1821 rt->rt6i_expires = 0;
1823 } else if (valid_lft) {
1824 clock_t expires = 0;
1825 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1826 if (addrconf_finite_timeout(rt_expires)) {
1827 /* not infinity */
1828 flags |= RTF_EXPIRES;
1829 expires = jiffies_to_clock_t(rt_expires);
1831 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1832 dev, expires, flags);
1834 if (rt)
1835 dst_release(&rt->u.dst);
1838 /* Try to figure out our local address for this prefix */
1840 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1841 struct inet6_ifaddr * ifp;
1842 struct in6_addr addr;
1843 int create = 0, update_lft = 0;
1845 if (pinfo->prefix_len == 64) {
1846 memcpy(&addr, &pinfo->prefix, 8);
1847 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1848 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1849 in6_dev_put(in6_dev);
1850 return;
1852 goto ok;
1854 if (net_ratelimit())
1855 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1856 pinfo->prefix_len);
1857 in6_dev_put(in6_dev);
1858 return;
1862 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1864 if (ifp == NULL && valid_lft) {
1865 int max_addresses = in6_dev->cnf.max_addresses;
1866 u32 addr_flags = 0;
1868 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1869 if (in6_dev->cnf.optimistic_dad &&
1870 !net->ipv6.devconf_all->forwarding)
1871 addr_flags = IFA_F_OPTIMISTIC;
1872 #endif
1874 /* Do not allow to create too much of autoconfigured
1875 * addresses; this would be too easy way to crash kernel.
1877 if (!max_addresses ||
1878 ipv6_count_addresses(in6_dev) < max_addresses)
1879 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1880 addr_type&IPV6_ADDR_SCOPE_MASK,
1881 addr_flags);
1883 if (!ifp || IS_ERR(ifp)) {
1884 in6_dev_put(in6_dev);
1885 return;
1888 update_lft = create = 1;
1889 ifp->cstamp = jiffies;
1890 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1893 if (ifp) {
1894 int flags;
1895 unsigned long now;
1896 #ifdef CONFIG_IPV6_PRIVACY
1897 struct inet6_ifaddr *ift;
1898 #endif
1899 u32 stored_lft;
1901 /* update lifetime (RFC2462 5.5.3 e) */
1902 spin_lock(&ifp->lock);
1903 now = jiffies;
1904 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1905 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1906 else
1907 stored_lft = 0;
1908 if (!update_lft && stored_lft) {
1909 if (valid_lft > MIN_VALID_LIFETIME ||
1910 valid_lft > stored_lft)
1911 update_lft = 1;
1912 else if (stored_lft <= MIN_VALID_LIFETIME) {
1913 /* valid_lft <= stored_lft is always true */
1915 * RFC 4862 Section 5.5.3e:
1916 * "Note that the preferred lifetime of
1917 * the corresponding address is always
1918 * reset to the Preferred Lifetime in
1919 * the received Prefix Information
1920 * option, regardless of whether the
1921 * valid lifetime is also reset or
1922 * ignored."
1924 * So if the preferred lifetime in
1925 * this advertisement is different
1926 * than what we have stored, but the
1927 * valid lifetime is invalid, just
1928 * reset prefered_lft.
1930 * We must set the valid lifetime
1931 * to the stored lifetime since we'll
1932 * be updating the timestamp below,
1933 * else we'll set it back to the
1934 * minumum.
1936 if (prefered_lft != ifp->prefered_lft) {
1937 valid_lft = stored_lft;
1938 update_lft = 1;
1940 } else {
1941 valid_lft = MIN_VALID_LIFETIME;
1942 if (valid_lft < prefered_lft)
1943 prefered_lft = valid_lft;
1944 update_lft = 1;
1948 if (update_lft) {
1949 ifp->valid_lft = valid_lft;
1950 ifp->prefered_lft = prefered_lft;
1951 ifp->tstamp = now;
1952 flags = ifp->flags;
1953 ifp->flags &= ~IFA_F_DEPRECATED;
1954 spin_unlock(&ifp->lock);
1956 if (!(flags&IFA_F_TENTATIVE))
1957 ipv6_ifa_notify(0, ifp);
1958 } else
1959 spin_unlock(&ifp->lock);
1961 #ifdef CONFIG_IPV6_PRIVACY
1962 read_lock_bh(&in6_dev->lock);
1963 /* update all temporary addresses in the list */
1964 list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1966 * When adjusting the lifetimes of an existing
1967 * temporary address, only lower the lifetimes.
1968 * Implementations must not increase the
1969 * lifetimes of an existing temporary address
1970 * when processing a Prefix Information Option.
1972 if (ifp != ift->ifpub)
1973 continue;
1975 spin_lock(&ift->lock);
1976 flags = ift->flags;
1977 if (ift->valid_lft > valid_lft &&
1978 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1979 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1980 if (ift->prefered_lft > prefered_lft &&
1981 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1982 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1983 spin_unlock(&ift->lock);
1984 if (!(flags&IFA_F_TENTATIVE))
1985 ipv6_ifa_notify(0, ift);
1988 if (create && in6_dev->cnf.use_tempaddr > 0) {
1990 * When a new public address is created as described in [ADDRCONF],
1991 * also create a new temporary address.
1993 read_unlock_bh(&in6_dev->lock);
1994 ipv6_create_tempaddr(ifp, NULL);
1995 } else {
1996 read_unlock_bh(&in6_dev->lock);
1998 #endif
1999 in6_ifa_put(ifp);
2000 addrconf_verify(0);
2003 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2004 in6_dev_put(in6_dev);
2008 * Set destination address.
2009 * Special case for SIT interfaces where we create a new "virtual"
2010 * device.
2012 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2014 struct in6_ifreq ireq;
2015 struct net_device *dev;
2016 int err = -EINVAL;
2018 rtnl_lock();
2020 err = -EFAULT;
2021 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2022 goto err_exit;
2024 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2026 err = -ENODEV;
2027 if (dev == NULL)
2028 goto err_exit;
2030 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2031 if (dev->type == ARPHRD_SIT) {
2032 const struct net_device_ops *ops = dev->netdev_ops;
2033 struct ifreq ifr;
2034 struct ip_tunnel_parm p;
2036 err = -EADDRNOTAVAIL;
2037 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2038 goto err_exit;
2040 memset(&p, 0, sizeof(p));
2041 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2042 p.iph.saddr = 0;
2043 p.iph.version = 4;
2044 p.iph.ihl = 5;
2045 p.iph.protocol = IPPROTO_IPV6;
2046 p.iph.ttl = 64;
2047 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2049 if (ops->ndo_do_ioctl) {
2050 mm_segment_t oldfs = get_fs();
2052 set_fs(KERNEL_DS);
2053 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2054 set_fs(oldfs);
2055 } else
2056 err = -EOPNOTSUPP;
2058 if (err == 0) {
2059 err = -ENOBUFS;
2060 dev = __dev_get_by_name(net, p.name);
2061 if (!dev)
2062 goto err_exit;
2063 err = dev_open(dev);
2066 #endif
2068 err_exit:
2069 rtnl_unlock();
2070 return err;
2074 * Manual configuration of address on an interface
2076 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2077 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2078 __u32 valid_lft)
2080 struct inet6_ifaddr *ifp;
2081 struct inet6_dev *idev;
2082 struct net_device *dev;
2083 int scope;
2084 u32 flags;
2085 clock_t expires;
2086 unsigned long timeout;
2088 ASSERT_RTNL();
2090 if (plen > 128)
2091 return -EINVAL;
2093 /* check the lifetime */
2094 if (!valid_lft || prefered_lft > valid_lft)
2095 return -EINVAL;
2097 dev = __dev_get_by_index(net, ifindex);
2098 if (!dev)
2099 return -ENODEV;
2101 if ((idev = addrconf_add_dev(dev)) == NULL)
2102 return -ENOBUFS;
2104 scope = ipv6_addr_scope(pfx);
2106 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2107 if (addrconf_finite_timeout(timeout)) {
2108 expires = jiffies_to_clock_t(timeout * HZ);
2109 valid_lft = timeout;
2110 flags = RTF_EXPIRES;
2111 } else {
2112 expires = 0;
2113 flags = 0;
2114 ifa_flags |= IFA_F_PERMANENT;
2117 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2118 if (addrconf_finite_timeout(timeout)) {
2119 if (timeout == 0)
2120 ifa_flags |= IFA_F_DEPRECATED;
2121 prefered_lft = timeout;
2124 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2126 if (!IS_ERR(ifp)) {
2127 spin_lock_bh(&ifp->lock);
2128 ifp->valid_lft = valid_lft;
2129 ifp->prefered_lft = prefered_lft;
2130 ifp->tstamp = jiffies;
2131 spin_unlock_bh(&ifp->lock);
2133 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2134 expires, flags);
2136 * Note that section 3.1 of RFC 4429 indicates
2137 * that the Optimistic flag should not be set for
2138 * manually configured addresses
2140 addrconf_dad_start(ifp, 0);
2141 in6_ifa_put(ifp);
2142 addrconf_verify(0);
2143 return 0;
2146 return PTR_ERR(ifp);
2149 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2150 unsigned int plen)
2152 struct inet6_ifaddr *ifp;
2153 struct inet6_dev *idev;
2154 struct net_device *dev;
2156 if (plen > 128)
2157 return -EINVAL;
2159 dev = __dev_get_by_index(net, ifindex);
2160 if (!dev)
2161 return -ENODEV;
2163 if ((idev = __in6_dev_get(dev)) == NULL)
2164 return -ENXIO;
2166 read_lock_bh(&idev->lock);
2167 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2168 if (ifp->prefix_len == plen &&
2169 ipv6_addr_equal(pfx, &ifp->addr)) {
2170 in6_ifa_hold(ifp);
2171 read_unlock_bh(&idev->lock);
2173 ipv6_del_addr(ifp);
2175 /* If the last address is deleted administratively,
2176 disable IPv6 on this interface.
2178 if (list_empty(&idev->addr_list))
2179 addrconf_ifdown(idev->dev, 1);
2180 return 0;
2183 read_unlock_bh(&idev->lock);
2184 return -EADDRNOTAVAIL;
2188 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2190 struct in6_ifreq ireq;
2191 int err;
2193 if (!capable(CAP_NET_ADMIN))
2194 return -EPERM;
2196 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2197 return -EFAULT;
2199 rtnl_lock();
2200 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2201 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2202 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2203 rtnl_unlock();
2204 return err;
2207 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2209 struct in6_ifreq ireq;
2210 int err;
2212 if (!capable(CAP_NET_ADMIN))
2213 return -EPERM;
2215 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2216 return -EFAULT;
2218 rtnl_lock();
2219 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2220 ireq.ifr6_prefixlen);
2221 rtnl_unlock();
2222 return err;
2225 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2226 int plen, int scope)
2228 struct inet6_ifaddr *ifp;
2230 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2231 if (!IS_ERR(ifp)) {
2232 spin_lock_bh(&ifp->lock);
2233 ifp->flags &= ~IFA_F_TENTATIVE;
2234 spin_unlock_bh(&ifp->lock);
2235 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2236 in6_ifa_put(ifp);
2240 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2241 static void sit_add_v4_addrs(struct inet6_dev *idev)
2243 struct in6_addr addr;
2244 struct net_device *dev;
2245 struct net *net = dev_net(idev->dev);
2246 int scope;
2248 ASSERT_RTNL();
2250 memset(&addr, 0, sizeof(struct in6_addr));
2251 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2253 if (idev->dev->flags&IFF_POINTOPOINT) {
2254 addr.s6_addr32[0] = htonl(0xfe800000);
2255 scope = IFA_LINK;
2256 } else {
2257 scope = IPV6_ADDR_COMPATv4;
2260 if (addr.s6_addr32[3]) {
2261 add_addr(idev, &addr, 128, scope);
2262 return;
2265 for_each_netdev(net, dev) {
2266 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2267 if (in_dev && (dev->flags & IFF_UP)) {
2268 struct in_ifaddr * ifa;
2270 int flag = scope;
2272 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2273 int plen;
2275 addr.s6_addr32[3] = ifa->ifa_local;
2277 if (ifa->ifa_scope == RT_SCOPE_LINK)
2278 continue;
2279 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2280 if (idev->dev->flags&IFF_POINTOPOINT)
2281 continue;
2282 flag |= IFA_HOST;
2284 if (idev->dev->flags&IFF_POINTOPOINT)
2285 plen = 64;
2286 else
2287 plen = 96;
2289 add_addr(idev, &addr, plen, flag);
2294 #endif
2296 static void init_loopback(struct net_device *dev)
2298 struct inet6_dev *idev;
2300 /* ::1 */
2302 ASSERT_RTNL();
2304 if ((idev = ipv6_find_idev(dev)) == NULL) {
2305 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2306 return;
2309 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2312 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2314 struct inet6_ifaddr * ifp;
2315 u32 addr_flags = IFA_F_PERMANENT;
2317 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2318 if (idev->cnf.optimistic_dad &&
2319 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2320 addr_flags |= IFA_F_OPTIMISTIC;
2321 #endif
2324 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2325 if (!IS_ERR(ifp)) {
2326 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2327 addrconf_dad_start(ifp, 0);
2328 in6_ifa_put(ifp);
2332 static void addrconf_dev_config(struct net_device *dev)
2334 struct in6_addr addr;
2335 struct inet6_dev * idev;
2337 ASSERT_RTNL();
2339 if ((dev->type != ARPHRD_ETHER) &&
2340 (dev->type != ARPHRD_FDDI) &&
2341 (dev->type != ARPHRD_IEEE802_TR) &&
2342 (dev->type != ARPHRD_ARCNET) &&
2343 (dev->type != ARPHRD_INFINIBAND)) {
2344 /* Alas, we support only Ethernet autoconfiguration. */
2345 return;
2348 idev = addrconf_add_dev(dev);
2349 if (idev == NULL)
2350 return;
2352 memset(&addr, 0, sizeof(struct in6_addr));
2353 addr.s6_addr32[0] = htonl(0xFE800000);
2355 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2356 addrconf_add_linklocal(idev, &addr);
2359 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2360 static void addrconf_sit_config(struct net_device *dev)
2362 struct inet6_dev *idev;
2364 ASSERT_RTNL();
2367 * Configure the tunnel with one of our IPv4
2368 * addresses... we should configure all of
2369 * our v4 addrs in the tunnel
2372 if ((idev = ipv6_find_idev(dev)) == NULL) {
2373 printk(KERN_DEBUG "init sit: add_dev failed\n");
2374 return;
2377 if (dev->priv_flags & IFF_ISATAP) {
2378 struct in6_addr addr;
2380 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2381 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2382 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2383 addrconf_add_linklocal(idev, &addr);
2384 return;
2387 sit_add_v4_addrs(idev);
2389 if (dev->flags&IFF_POINTOPOINT) {
2390 addrconf_add_mroute(dev);
2391 addrconf_add_lroute(dev);
2392 } else
2393 sit_route_add(dev);
2395 #endif
2397 static inline int
2398 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2400 struct in6_addr lladdr;
2402 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2403 addrconf_add_linklocal(idev, &lladdr);
2404 return 0;
2406 return -1;
2409 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2411 struct net_device *link_dev;
2412 struct net *net = dev_net(idev->dev);
2414 /* first try to inherit the link-local address from the link device */
2415 if (idev->dev->iflink &&
2416 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2417 if (!ipv6_inherit_linklocal(idev, link_dev))
2418 return;
2420 /* then try to inherit it from any device */
2421 for_each_netdev(net, link_dev) {
2422 if (!ipv6_inherit_linklocal(idev, link_dev))
2423 return;
2425 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2429 * Autoconfigure tunnel with a link-local address so routing protocols,
2430 * DHCPv6, MLD etc. can be run over the virtual link
2433 static void addrconf_ip6_tnl_config(struct net_device *dev)
2435 struct inet6_dev *idev;
2437 ASSERT_RTNL();
2439 idev = addrconf_add_dev(dev);
2440 if (!idev) {
2441 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2442 return;
2444 ip6_tnl_add_linklocal(idev);
2447 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2448 void * data)
2450 struct net_device *dev = (struct net_device *) data;
2451 struct inet6_dev *idev = __in6_dev_get(dev);
2452 int run_pending = 0;
2453 int err;
2455 switch (event) {
2456 case NETDEV_REGISTER:
2457 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2458 idev = ipv6_add_dev(dev);
2459 if (!idev)
2460 return notifier_from_errno(-ENOMEM);
2462 break;
2464 case NETDEV_UP:
2465 case NETDEV_CHANGE:
2466 if (dev->flags & IFF_SLAVE)
2467 break;
2469 if (event == NETDEV_UP) {
2470 if (!addrconf_qdisc_ok(dev)) {
2471 /* device is not ready yet. */
2472 printk(KERN_INFO
2473 "ADDRCONF(NETDEV_UP): %s: "
2474 "link is not ready\n",
2475 dev->name);
2476 break;
2479 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2480 idev = ipv6_add_dev(dev);
2482 if (idev) {
2483 idev->if_flags |= IF_READY;
2484 run_pending = 1;
2486 } else {
2487 if (!addrconf_qdisc_ok(dev)) {
2488 /* device is still not ready. */
2489 break;
2492 if (idev) {
2493 if (idev->if_flags & IF_READY)
2494 /* device is already configured. */
2495 break;
2496 idev->if_flags |= IF_READY;
2499 printk(KERN_INFO
2500 "ADDRCONF(NETDEV_CHANGE): %s: "
2501 "link becomes ready\n",
2502 dev->name);
2504 run_pending = 1;
2507 switch (dev->type) {
2508 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2509 case ARPHRD_SIT:
2510 addrconf_sit_config(dev);
2511 break;
2512 #endif
2513 case ARPHRD_TUNNEL6:
2514 addrconf_ip6_tnl_config(dev);
2515 break;
2516 case ARPHRD_LOOPBACK:
2517 init_loopback(dev);
2518 break;
2520 default:
2521 addrconf_dev_config(dev);
2522 break;
2525 if (idev) {
2526 if (run_pending)
2527 addrconf_dad_run(idev);
2530 * If the MTU changed during the interface down,
2531 * when the interface up, the changed MTU must be
2532 * reflected in the idev as well as routers.
2534 if (idev->cnf.mtu6 != dev->mtu &&
2535 dev->mtu >= IPV6_MIN_MTU) {
2536 rt6_mtu_change(dev, dev->mtu);
2537 idev->cnf.mtu6 = dev->mtu;
2539 idev->tstamp = jiffies;
2540 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2543 * If the changed mtu during down is lower than
2544 * IPV6_MIN_MTU stop IPv6 on this interface.
2546 if (dev->mtu < IPV6_MIN_MTU)
2547 addrconf_ifdown(dev, 1);
2549 break;
2551 case NETDEV_CHANGEMTU:
2552 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2553 rt6_mtu_change(dev, dev->mtu);
2554 idev->cnf.mtu6 = dev->mtu;
2555 break;
2558 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2559 idev = ipv6_add_dev(dev);
2560 if (idev)
2561 break;
2565 * MTU falled under IPV6_MIN_MTU.
2566 * Stop IPv6 on this interface.
2569 case NETDEV_DOWN:
2570 case NETDEV_UNREGISTER:
2572 * Remove all addresses from this interface.
2574 addrconf_ifdown(dev, event != NETDEV_DOWN);
2575 break;
2577 case NETDEV_CHANGENAME:
2578 if (idev) {
2579 snmp6_unregister_dev(idev);
2580 addrconf_sysctl_unregister(idev);
2581 addrconf_sysctl_register(idev);
2582 err = snmp6_register_dev(idev);
2583 if (err)
2584 return notifier_from_errno(err);
2586 break;
2588 case NETDEV_PRE_TYPE_CHANGE:
2589 case NETDEV_POST_TYPE_CHANGE:
2590 addrconf_type_change(dev, event);
2591 break;
2594 return NOTIFY_OK;
2598 * addrconf module should be notified of a device going up
2600 static struct notifier_block ipv6_dev_notf = {
2601 .notifier_call = addrconf_notify,
2604 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2606 struct inet6_dev *idev;
2607 ASSERT_RTNL();
2609 idev = __in6_dev_get(dev);
2611 if (event == NETDEV_POST_TYPE_CHANGE)
2612 ipv6_mc_remap(idev);
2613 else if (event == NETDEV_PRE_TYPE_CHANGE)
2614 ipv6_mc_unmap(idev);
2617 static int addrconf_ifdown(struct net_device *dev, int how)
2619 struct net *net = dev_net(dev);
2620 struct inet6_dev *idev;
2621 struct inet6_ifaddr *ifa;
2622 LIST_HEAD(keep_list);
2624 ASSERT_RTNL();
2626 rt6_ifdown(net, dev);
2627 neigh_ifdown(&nd_tbl, dev);
2629 idev = __in6_dev_get(dev);
2630 if (idev == NULL)
2631 return -ENODEV;
2634 * Step 1: remove reference to ipv6 device from parent device.
2635 * Do not dev_put!
2637 if (how) {
2638 idev->dead = 1;
2640 /* protected by rtnl_lock */
2641 rcu_assign_pointer(dev->ip6_ptr, NULL);
2643 /* Step 1.5: remove snmp6 entry */
2644 snmp6_unregister_dev(idev);
2648 write_lock_bh(&idev->lock);
2650 /* Step 2: clear flags for stateless addrconf */
2651 if (!how)
2652 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2654 #ifdef CONFIG_IPV6_PRIVACY
2655 if (how && del_timer(&idev->regen_timer))
2656 in6_dev_put(idev);
2658 /* Step 3: clear tempaddr list */
2659 while (!list_empty(&idev->tempaddr_list)) {
2660 ifa = list_first_entry(&idev->tempaddr_list,
2661 struct inet6_ifaddr, tmp_list);
2662 list_del(&ifa->tmp_list);
2663 ifa->dead = 1;
2664 write_unlock_bh(&idev->lock);
2665 spin_lock_bh(&ifa->lock);
2667 if (ifa->ifpub) {
2668 in6_ifa_put(ifa->ifpub);
2669 ifa->ifpub = NULL;
2671 spin_unlock_bh(&ifa->lock);
2672 in6_ifa_put(ifa);
2673 write_lock_bh(&idev->lock);
2675 #endif
2677 while (!list_empty(&idev->addr_list)) {
2678 ifa = list_first_entry(&idev->addr_list,
2679 struct inet6_ifaddr, if_list);
2680 addrconf_del_timer(ifa);
2682 /* If just doing link down, and address is permanent
2683 and not link-local, then retain it. */
2684 if (!how &&
2685 (ifa->flags&IFA_F_PERMANENT) &&
2686 !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2687 list_move_tail(&ifa->if_list, &keep_list);
2689 /* If not doing DAD on this address, just keep it. */
2690 if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2691 idev->cnf.accept_dad <= 0 ||
2692 (ifa->flags & IFA_F_NODAD))
2693 continue;
2695 /* If it was tentative already, no need to notify */
2696 if (ifa->flags & IFA_F_TENTATIVE)
2697 continue;
2699 /* Flag it for later restoration when link comes up */
2700 ifa->flags |= IFA_F_TENTATIVE;
2701 in6_ifa_hold(ifa);
2702 } else {
2703 list_del(&ifa->if_list);
2704 ifa->dead = 1;
2706 write_unlock_bh(&idev->lock);
2708 /* clear hash table */
2709 spin_lock_bh(&addrconf_hash_lock);
2710 hlist_del_init_rcu(&ifa->addr_lst);
2711 __in6_ifa_put(ifa);
2712 spin_unlock_bh(&addrconf_hash_lock);
2714 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2715 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2716 in6_ifa_put(ifa);
2718 write_lock_bh(&idev->lock);
2721 list_splice(&keep_list, &idev->addr_list);
2723 write_unlock_bh(&idev->lock);
2725 /* Step 5: Discard multicast list */
2726 if (how)
2727 ipv6_mc_destroy_dev(idev);
2728 else
2729 ipv6_mc_down(idev);
2731 idev->tstamp = jiffies;
2733 /* Last: Shot the device (if unregistered) */
2734 if (how) {
2735 addrconf_sysctl_unregister(idev);
2736 neigh_parms_release(&nd_tbl, idev->nd_parms);
2737 neigh_ifdown(&nd_tbl, dev);
2738 in6_dev_put(idev);
2740 return 0;
2743 static void addrconf_rs_timer(unsigned long data)
2745 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2746 struct inet6_dev *idev = ifp->idev;
2748 read_lock(&idev->lock);
2749 if (idev->dead || !(idev->if_flags & IF_READY))
2750 goto out;
2752 if (idev->cnf.forwarding)
2753 goto out;
2755 /* Announcement received after solicitation was sent */
2756 if (idev->if_flags & IF_RA_RCVD)
2757 goto out;
2759 spin_lock(&ifp->lock);
2760 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2761 /* The wait after the last probe can be shorter */
2762 addrconf_mod_timer(ifp, AC_RS,
2763 (ifp->probes == idev->cnf.rtr_solicits) ?
2764 idev->cnf.rtr_solicit_delay :
2765 idev->cnf.rtr_solicit_interval);
2766 spin_unlock(&ifp->lock);
2768 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2769 } else {
2770 spin_unlock(&ifp->lock);
2772 * Note: we do not support deprecated "all on-link"
2773 * assumption any longer.
2775 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2776 idev->dev->name);
2779 out:
2780 read_unlock(&idev->lock);
2781 in6_ifa_put(ifp);
2785 * Duplicate Address Detection
2787 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2789 unsigned long rand_num;
2790 struct inet6_dev *idev = ifp->idev;
2792 if (ifp->flags & IFA_F_OPTIMISTIC)
2793 rand_num = 0;
2794 else
2795 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2797 ifp->probes = idev->cnf.dad_transmits;
2798 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2801 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2803 struct inet6_dev *idev = ifp->idev;
2804 struct net_device *dev = idev->dev;
2806 addrconf_join_solict(dev, &ifp->addr);
2808 net_srandom(ifp->addr.s6_addr32[3]);
2810 read_lock_bh(&idev->lock);
2811 if (ifp->dead)
2812 goto out;
2814 spin_lock(&ifp->lock);
2815 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2816 idev->cnf.accept_dad < 1 ||
2817 !(ifp->flags&IFA_F_TENTATIVE) ||
2818 ifp->flags & IFA_F_NODAD) {
2819 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2820 spin_unlock(&ifp->lock);
2821 read_unlock_bh(&idev->lock);
2823 addrconf_dad_completed(ifp);
2824 return;
2827 if (!(idev->if_flags & IF_READY)) {
2828 spin_unlock(&ifp->lock);
2829 read_unlock_bh(&idev->lock);
2831 * If the device is not ready:
2832 * - keep it tentative if it is a permanent address.
2833 * - otherwise, kill it.
2835 in6_ifa_hold(ifp);
2836 addrconf_dad_stop(ifp, 0);
2837 return;
2841 * Optimistic nodes can start receiving
2842 * Frames right away
2844 if (ifp->flags & IFA_F_OPTIMISTIC)
2845 ip6_ins_rt(ifp->rt);
2847 addrconf_dad_kick(ifp);
2848 spin_unlock(&ifp->lock);
2849 out:
2850 read_unlock_bh(&idev->lock);
2853 static void addrconf_dad_timer(unsigned long data)
2855 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2856 struct inet6_dev *idev = ifp->idev;
2857 struct in6_addr mcaddr;
2859 read_lock(&idev->lock);
2860 if (idev->dead || !(idev->if_flags & IF_READY)) {
2861 read_unlock(&idev->lock);
2862 goto out;
2865 spin_lock(&ifp->lock);
2866 if (ifp->probes == 0) {
2868 * DAD was successful
2871 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2872 spin_unlock(&ifp->lock);
2873 read_unlock(&idev->lock);
2875 addrconf_dad_completed(ifp);
2877 goto out;
2880 ifp->probes--;
2881 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2882 spin_unlock(&ifp->lock);
2883 read_unlock(&idev->lock);
2885 /* send a neighbour solicitation for our addr */
2886 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2887 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2888 out:
2889 in6_ifa_put(ifp);
2892 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2894 struct net_device *dev = ifp->idev->dev;
2897 * Configure the address for reception. Now it is valid.
2900 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2902 /* If added prefix is link local and forwarding is off,
2903 start sending router solicitations.
2906 if (ifp->idev->cnf.forwarding == 0 &&
2907 ifp->idev->cnf.rtr_solicits > 0 &&
2908 (dev->flags&IFF_LOOPBACK) == 0 &&
2909 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2911 * If a host as already performed a random delay
2912 * [...] as part of DAD [...] there is no need
2913 * to delay again before sending the first RS
2915 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2917 spin_lock_bh(&ifp->lock);
2918 ifp->probes = 1;
2919 ifp->idev->if_flags |= IF_RS_SENT;
2920 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2921 spin_unlock_bh(&ifp->lock);
2925 static void addrconf_dad_run(struct inet6_dev *idev)
2927 struct inet6_ifaddr *ifp;
2929 read_lock_bh(&idev->lock);
2930 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2931 spin_lock(&ifp->lock);
2932 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2933 spin_unlock(&ifp->lock);
2934 continue;
2936 spin_unlock(&ifp->lock);
2937 addrconf_dad_kick(ifp);
2939 read_unlock_bh(&idev->lock);
2942 #ifdef CONFIG_PROC_FS
2943 struct if6_iter_state {
2944 struct seq_net_private p;
2945 int bucket;
2948 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2950 struct inet6_ifaddr *ifa = NULL;
2951 struct if6_iter_state *state = seq->private;
2952 struct net *net = seq_file_net(seq);
2954 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2955 struct hlist_node *n;
2956 hlist_for_each_entry_rcu(ifa, n, &inet6_addr_lst[state->bucket],
2957 addr_lst)
2958 if (net_eq(dev_net(ifa->idev->dev), net))
2959 return ifa;
2961 return NULL;
2964 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2965 struct inet6_ifaddr *ifa)
2967 struct if6_iter_state *state = seq->private;
2968 struct net *net = seq_file_net(seq);
2969 struct hlist_node *n = &ifa->addr_lst;
2971 hlist_for_each_entry_continue_rcu(ifa, n, addr_lst)
2972 if (net_eq(dev_net(ifa->idev->dev), net))
2973 return ifa;
2975 while (++state->bucket < IN6_ADDR_HSIZE) {
2976 hlist_for_each_entry(ifa, n,
2977 &inet6_addr_lst[state->bucket], addr_lst) {
2978 if (net_eq(dev_net(ifa->idev->dev), net))
2979 return ifa;
2983 return NULL;
2986 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2988 struct inet6_ifaddr *ifa = if6_get_first(seq);
2990 if (ifa)
2991 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2992 --pos;
2993 return pos ? NULL : ifa;
2996 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2997 __acquires(rcu)
2999 rcu_read_lock_bh();
3000 return if6_get_idx(seq, *pos);
3003 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3005 struct inet6_ifaddr *ifa;
3007 ifa = if6_get_next(seq, v);
3008 ++*pos;
3009 return ifa;
3012 static void if6_seq_stop(struct seq_file *seq, void *v)
3013 __releases(rcu)
3015 rcu_read_unlock_bh();
3018 static int if6_seq_show(struct seq_file *seq, void *v)
3020 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3021 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3022 &ifp->addr,
3023 ifp->idev->dev->ifindex,
3024 ifp->prefix_len,
3025 ifp->scope,
3026 ifp->flags,
3027 ifp->idev->dev->name);
3028 return 0;
3031 static const struct seq_operations if6_seq_ops = {
3032 .start = if6_seq_start,
3033 .next = if6_seq_next,
3034 .show = if6_seq_show,
3035 .stop = if6_seq_stop,
3038 static int if6_seq_open(struct inode *inode, struct file *file)
3040 return seq_open_net(inode, file, &if6_seq_ops,
3041 sizeof(struct if6_iter_state));
3044 static const struct file_operations if6_fops = {
3045 .owner = THIS_MODULE,
3046 .open = if6_seq_open,
3047 .read = seq_read,
3048 .llseek = seq_lseek,
3049 .release = seq_release_net,
3052 static int __net_init if6_proc_net_init(struct net *net)
3054 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3055 return -ENOMEM;
3056 return 0;
3059 static void __net_exit if6_proc_net_exit(struct net *net)
3061 proc_net_remove(net, "if_inet6");
3064 static struct pernet_operations if6_proc_net_ops = {
3065 .init = if6_proc_net_init,
3066 .exit = if6_proc_net_exit,
3069 int __init if6_proc_init(void)
3071 return register_pernet_subsys(&if6_proc_net_ops);
3074 void if6_proc_exit(void)
3076 unregister_pernet_subsys(&if6_proc_net_ops);
3078 #endif /* CONFIG_PROC_FS */
3080 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3081 /* Check if address is a home address configured on any interface. */
3082 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3084 int ret = 0;
3085 struct inet6_ifaddr *ifp = NULL;
3086 struct hlist_node *n;
3087 unsigned int hash = ipv6_addr_hash(addr);
3089 rcu_read_lock_bh();
3090 hlist_for_each_entry_rcu(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3091 if (!net_eq(dev_net(ifp->idev->dev), net))
3092 continue;
3093 if (ipv6_addr_equal(&ifp->addr, addr) &&
3094 (ifp->flags & IFA_F_HOMEADDRESS)) {
3095 ret = 1;
3096 break;
3099 rcu_read_unlock_bh();
3100 return ret;
3102 #endif
3105 * Periodic address status verification
3108 static void addrconf_verify(unsigned long foo)
3110 struct inet6_ifaddr *ifp;
3111 struct hlist_node *node;
3112 unsigned long now, next;
3113 int i;
3115 rcu_read_lock_bh();
3116 spin_lock(&addrconf_verify_lock);
3117 now = jiffies;
3118 next = now + ADDR_CHECK_FREQUENCY;
3120 del_timer(&addr_chk_timer);
3122 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3123 restart:
3124 hlist_for_each_entry_rcu(ifp, node,
3125 &inet6_addr_lst[i], addr_lst) {
3126 unsigned long age;
3128 if (ifp->flags & IFA_F_PERMANENT)
3129 continue;
3131 spin_lock(&ifp->lock);
3132 age = (now - ifp->tstamp) / HZ;
3134 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3135 age >= ifp->valid_lft) {
3136 spin_unlock(&ifp->lock);
3137 in6_ifa_hold(ifp);
3138 ipv6_del_addr(ifp);
3139 goto restart;
3140 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3141 spin_unlock(&ifp->lock);
3142 continue;
3143 } else if (age >= ifp->prefered_lft) {
3144 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3145 int deprecate = 0;
3147 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3148 deprecate = 1;
3149 ifp->flags |= IFA_F_DEPRECATED;
3152 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3153 next = ifp->tstamp + ifp->valid_lft * HZ;
3155 spin_unlock(&ifp->lock);
3157 if (deprecate) {
3158 in6_ifa_hold(ifp);
3160 ipv6_ifa_notify(0, ifp);
3161 in6_ifa_put(ifp);
3162 goto restart;
3164 #ifdef CONFIG_IPV6_PRIVACY
3165 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3166 !(ifp->flags&IFA_F_TENTATIVE)) {
3167 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3168 ifp->idev->cnf.dad_transmits *
3169 ifp->idev->nd_parms->retrans_time / HZ;
3171 if (age >= ifp->prefered_lft - regen_advance) {
3172 struct inet6_ifaddr *ifpub = ifp->ifpub;
3173 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3174 next = ifp->tstamp + ifp->prefered_lft * HZ;
3175 if (!ifp->regen_count && ifpub) {
3176 ifp->regen_count++;
3177 in6_ifa_hold(ifp);
3178 in6_ifa_hold(ifpub);
3179 spin_unlock(&ifp->lock);
3181 spin_lock(&ifpub->lock);
3182 ifpub->regen_count = 0;
3183 spin_unlock(&ifpub->lock);
3184 ipv6_create_tempaddr(ifpub, ifp);
3185 in6_ifa_put(ifpub);
3186 in6_ifa_put(ifp);
3187 goto restart;
3189 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3190 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3191 spin_unlock(&ifp->lock);
3192 #endif
3193 } else {
3194 /* ifp->prefered_lft <= ifp->valid_lft */
3195 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3196 next = ifp->tstamp + ifp->prefered_lft * HZ;
3197 spin_unlock(&ifp->lock);
3202 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3203 add_timer(&addr_chk_timer);
3204 spin_unlock(&addrconf_verify_lock);
3205 rcu_read_unlock_bh();
3208 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3210 struct in6_addr *pfx = NULL;
3212 if (addr)
3213 pfx = nla_data(addr);
3215 if (local) {
3216 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3217 pfx = NULL;
3218 else
3219 pfx = nla_data(local);
3222 return pfx;
3225 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3226 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3227 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3228 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3231 static int
3232 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3234 struct net *net = sock_net(skb->sk);
3235 struct ifaddrmsg *ifm;
3236 struct nlattr *tb[IFA_MAX+1];
3237 struct in6_addr *pfx;
3238 int err;
3240 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3241 if (err < 0)
3242 return err;
3244 ifm = nlmsg_data(nlh);
3245 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3246 if (pfx == NULL)
3247 return -EINVAL;
3249 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3252 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3253 u32 prefered_lft, u32 valid_lft)
3255 u32 flags;
3256 clock_t expires;
3257 unsigned long timeout;
3259 if (!valid_lft || (prefered_lft > valid_lft))
3260 return -EINVAL;
3262 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3263 if (addrconf_finite_timeout(timeout)) {
3264 expires = jiffies_to_clock_t(timeout * HZ);
3265 valid_lft = timeout;
3266 flags = RTF_EXPIRES;
3267 } else {
3268 expires = 0;
3269 flags = 0;
3270 ifa_flags |= IFA_F_PERMANENT;
3273 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3274 if (addrconf_finite_timeout(timeout)) {
3275 if (timeout == 0)
3276 ifa_flags |= IFA_F_DEPRECATED;
3277 prefered_lft = timeout;
3280 spin_lock_bh(&ifp->lock);
3281 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3282 ifp->tstamp = jiffies;
3283 ifp->valid_lft = valid_lft;
3284 ifp->prefered_lft = prefered_lft;
3286 spin_unlock_bh(&ifp->lock);
3287 if (!(ifp->flags&IFA_F_TENTATIVE))
3288 ipv6_ifa_notify(0, ifp);
3290 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3291 expires, flags);
3292 addrconf_verify(0);
3294 return 0;
3297 static int
3298 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3300 struct net *net = sock_net(skb->sk);
3301 struct ifaddrmsg *ifm;
3302 struct nlattr *tb[IFA_MAX+1];
3303 struct in6_addr *pfx;
3304 struct inet6_ifaddr *ifa;
3305 struct net_device *dev;
3306 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3307 u8 ifa_flags;
3308 int err;
3310 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3311 if (err < 0)
3312 return err;
3314 ifm = nlmsg_data(nlh);
3315 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3316 if (pfx == NULL)
3317 return -EINVAL;
3319 if (tb[IFA_CACHEINFO]) {
3320 struct ifa_cacheinfo *ci;
3322 ci = nla_data(tb[IFA_CACHEINFO]);
3323 valid_lft = ci->ifa_valid;
3324 preferred_lft = ci->ifa_prefered;
3325 } else {
3326 preferred_lft = INFINITY_LIFE_TIME;
3327 valid_lft = INFINITY_LIFE_TIME;
3330 dev = __dev_get_by_index(net, ifm->ifa_index);
3331 if (dev == NULL)
3332 return -ENODEV;
3334 /* We ignore other flags so far. */
3335 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3337 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3338 if (ifa == NULL) {
3340 * It would be best to check for !NLM_F_CREATE here but
3341 * userspace alreay relies on not having to provide this.
3343 return inet6_addr_add(net, ifm->ifa_index, pfx,
3344 ifm->ifa_prefixlen, ifa_flags,
3345 preferred_lft, valid_lft);
3348 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3349 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3350 err = -EEXIST;
3351 else
3352 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3354 in6_ifa_put(ifa);
3356 return err;
3359 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3360 u8 scope, int ifindex)
3362 struct ifaddrmsg *ifm;
3364 ifm = nlmsg_data(nlh);
3365 ifm->ifa_family = AF_INET6;
3366 ifm->ifa_prefixlen = prefixlen;
3367 ifm->ifa_flags = flags;
3368 ifm->ifa_scope = scope;
3369 ifm->ifa_index = ifindex;
3372 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3373 unsigned long tstamp, u32 preferred, u32 valid)
3375 struct ifa_cacheinfo ci;
3377 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3378 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3379 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3380 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3381 ci.ifa_prefered = preferred;
3382 ci.ifa_valid = valid;
3384 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3387 static inline int rt_scope(int ifa_scope)
3389 if (ifa_scope & IFA_HOST)
3390 return RT_SCOPE_HOST;
3391 else if (ifa_scope & IFA_LINK)
3392 return RT_SCOPE_LINK;
3393 else if (ifa_scope & IFA_SITE)
3394 return RT_SCOPE_SITE;
3395 else
3396 return RT_SCOPE_UNIVERSE;
3399 static inline int inet6_ifaddr_msgsize(void)
3401 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3402 + nla_total_size(16) /* IFA_ADDRESS */
3403 + nla_total_size(sizeof(struct ifa_cacheinfo));
3406 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3407 u32 pid, u32 seq, int event, unsigned int flags)
3409 struct nlmsghdr *nlh;
3410 u32 preferred, valid;
3412 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3413 if (nlh == NULL)
3414 return -EMSGSIZE;
3416 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3417 ifa->idev->dev->ifindex);
3419 if (!(ifa->flags&IFA_F_PERMANENT)) {
3420 preferred = ifa->prefered_lft;
3421 valid = ifa->valid_lft;
3422 if (preferred != INFINITY_LIFE_TIME) {
3423 long tval = (jiffies - ifa->tstamp)/HZ;
3424 if (preferred > tval)
3425 preferred -= tval;
3426 else
3427 preferred = 0;
3428 if (valid != INFINITY_LIFE_TIME)
3429 valid -= tval;
3431 } else {
3432 preferred = INFINITY_LIFE_TIME;
3433 valid = INFINITY_LIFE_TIME;
3436 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3437 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3438 nlmsg_cancel(skb, nlh);
3439 return -EMSGSIZE;
3442 return nlmsg_end(skb, nlh);
3445 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3446 u32 pid, u32 seq, int event, u16 flags)
3448 struct nlmsghdr *nlh;
3449 u8 scope = RT_SCOPE_UNIVERSE;
3450 int ifindex = ifmca->idev->dev->ifindex;
3452 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3453 scope = RT_SCOPE_SITE;
3455 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3456 if (nlh == NULL)
3457 return -EMSGSIZE;
3459 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3460 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3461 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3462 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3463 nlmsg_cancel(skb, nlh);
3464 return -EMSGSIZE;
3467 return nlmsg_end(skb, nlh);
3470 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3471 u32 pid, u32 seq, int event, unsigned int flags)
3473 struct nlmsghdr *nlh;
3474 u8 scope = RT_SCOPE_UNIVERSE;
3475 int ifindex = ifaca->aca_idev->dev->ifindex;
3477 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3478 scope = RT_SCOPE_SITE;
3480 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3481 if (nlh == NULL)
3482 return -EMSGSIZE;
3484 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3485 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3486 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3487 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3488 nlmsg_cancel(skb, nlh);
3489 return -EMSGSIZE;
3492 return nlmsg_end(skb, nlh);
3495 enum addr_type_t {
3496 UNICAST_ADDR,
3497 MULTICAST_ADDR,
3498 ANYCAST_ADDR,
3501 /* called with rcu_read_lock() */
3502 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3503 struct netlink_callback *cb, enum addr_type_t type,
3504 int s_ip_idx, int *p_ip_idx)
3506 struct ifmcaddr6 *ifmca;
3507 struct ifacaddr6 *ifaca;
3508 int err = 1;
3509 int ip_idx = *p_ip_idx;
3511 read_lock_bh(&idev->lock);
3512 switch (type) {
3513 case UNICAST_ADDR: {
3514 struct inet6_ifaddr *ifa;
3516 /* unicast address incl. temp addr */
3517 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3518 if (++ip_idx < s_ip_idx)
3519 continue;
3520 err = inet6_fill_ifaddr(skb, ifa,
3521 NETLINK_CB(cb->skb).pid,
3522 cb->nlh->nlmsg_seq,
3523 RTM_NEWADDR,
3524 NLM_F_MULTI);
3525 if (err <= 0)
3526 break;
3528 break;
3530 case MULTICAST_ADDR:
3531 /* multicast address */
3532 for (ifmca = idev->mc_list; ifmca;
3533 ifmca = ifmca->next, ip_idx++) {
3534 if (ip_idx < s_ip_idx)
3535 continue;
3536 err = inet6_fill_ifmcaddr(skb, ifmca,
3537 NETLINK_CB(cb->skb).pid,
3538 cb->nlh->nlmsg_seq,
3539 RTM_GETMULTICAST,
3540 NLM_F_MULTI);
3541 if (err <= 0)
3542 break;
3544 break;
3545 case ANYCAST_ADDR:
3546 /* anycast address */
3547 for (ifaca = idev->ac_list; ifaca;
3548 ifaca = ifaca->aca_next, ip_idx++) {
3549 if (ip_idx < s_ip_idx)
3550 continue;
3551 err = inet6_fill_ifacaddr(skb, ifaca,
3552 NETLINK_CB(cb->skb).pid,
3553 cb->nlh->nlmsg_seq,
3554 RTM_GETANYCAST,
3555 NLM_F_MULTI);
3556 if (err <= 0)
3557 break;
3559 break;
3560 default:
3561 break;
3563 read_unlock_bh(&idev->lock);
3564 *p_ip_idx = ip_idx;
3565 return err;
3568 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3569 enum addr_type_t type)
3571 struct net *net = sock_net(skb->sk);
3572 int h, s_h;
3573 int idx, ip_idx;
3574 int s_idx, s_ip_idx;
3575 struct net_device *dev;
3576 struct inet6_dev *idev;
3577 struct hlist_head *head;
3578 struct hlist_node *node;
3580 s_h = cb->args[0];
3581 s_idx = idx = cb->args[1];
3582 s_ip_idx = ip_idx = cb->args[2];
3584 rcu_read_lock();
3585 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3586 idx = 0;
3587 head = &net->dev_index_head[h];
3588 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3589 if (idx < s_idx)
3590 goto cont;
3591 if (idx > s_idx)
3592 s_ip_idx = 0;
3593 ip_idx = 0;
3594 idev = __in6_dev_get(dev);
3595 if (!idev)
3596 goto cont;
3598 if (in6_dump_addrs(idev, skb, cb, type,
3599 s_ip_idx, &ip_idx) <= 0)
3600 goto done;
3601 cont:
3602 idx++;
3605 done:
3606 rcu_read_unlock();
3607 cb->args[0] = h;
3608 cb->args[1] = idx;
3609 cb->args[2] = ip_idx;
3611 return skb->len;
3614 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3616 enum addr_type_t type = UNICAST_ADDR;
3618 return inet6_dump_addr(skb, cb, type);
3621 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3623 enum addr_type_t type = MULTICAST_ADDR;
3625 return inet6_dump_addr(skb, cb, type);
3629 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3631 enum addr_type_t type = ANYCAST_ADDR;
3633 return inet6_dump_addr(skb, cb, type);
3636 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3637 void *arg)
3639 struct net *net = sock_net(in_skb->sk);
3640 struct ifaddrmsg *ifm;
3641 struct nlattr *tb[IFA_MAX+1];
3642 struct in6_addr *addr = NULL;
3643 struct net_device *dev = NULL;
3644 struct inet6_ifaddr *ifa;
3645 struct sk_buff *skb;
3646 int err;
3648 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3649 if (err < 0)
3650 goto errout;
3652 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3653 if (addr == NULL) {
3654 err = -EINVAL;
3655 goto errout;
3658 ifm = nlmsg_data(nlh);
3659 if (ifm->ifa_index)
3660 dev = __dev_get_by_index(net, ifm->ifa_index);
3662 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3663 if (!ifa) {
3664 err = -EADDRNOTAVAIL;
3665 goto errout;
3668 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3669 if (!skb) {
3670 err = -ENOBUFS;
3671 goto errout_ifa;
3674 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3675 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3676 if (err < 0) {
3677 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3678 WARN_ON(err == -EMSGSIZE);
3679 kfree_skb(skb);
3680 goto errout_ifa;
3682 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3683 errout_ifa:
3684 in6_ifa_put(ifa);
3685 errout:
3686 return err;
3689 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3691 struct sk_buff *skb;
3692 struct net *net = dev_net(ifa->idev->dev);
3693 int err = -ENOBUFS;
3695 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3696 if (skb == NULL)
3697 goto errout;
3699 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3700 if (err < 0) {
3701 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3702 WARN_ON(err == -EMSGSIZE);
3703 kfree_skb(skb);
3704 goto errout;
3706 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3707 return;
3708 errout:
3709 if (err < 0)
3710 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3713 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3714 __s32 *array, int bytes)
3716 BUG_ON(bytes < (DEVCONF_MAX * 4));
3718 memset(array, 0, bytes);
3719 array[DEVCONF_FORWARDING] = cnf->forwarding;
3720 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3721 array[DEVCONF_MTU6] = cnf->mtu6;
3722 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3723 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3724 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3725 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3726 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3727 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3728 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3729 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3730 #ifdef CONFIG_IPV6_PRIVACY
3731 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3732 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3733 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3734 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3735 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3736 #endif
3737 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3738 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3739 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3740 #ifdef CONFIG_IPV6_ROUTER_PREF
3741 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3742 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3743 #ifdef CONFIG_IPV6_ROUTE_INFO
3744 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3745 #endif
3746 #endif
3747 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3748 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3749 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3750 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3751 #endif
3752 #ifdef CONFIG_IPV6_MROUTE
3753 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3754 #endif
3755 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3756 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3757 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3760 static inline size_t inet6_if_nlmsg_size(void)
3762 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3763 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3764 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3765 + nla_total_size(4) /* IFLA_MTU */
3766 + nla_total_size(4) /* IFLA_LINK */
3767 + nla_total_size( /* IFLA_PROTINFO */
3768 nla_total_size(4) /* IFLA_INET6_FLAGS */
3769 + nla_total_size(sizeof(struct ifla_cacheinfo))
3770 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3771 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3772 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3776 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3777 int items, int bytes)
3779 int i;
3780 int pad = bytes - sizeof(u64) * items;
3781 BUG_ON(pad < 0);
3783 /* Use put_unaligned() because stats may not be aligned for u64. */
3784 put_unaligned(items, &stats[0]);
3785 for (i = 1; i < items; i++)
3786 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3788 memset(&stats[items], 0, pad);
3791 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3792 int bytes)
3794 switch (attrtype) {
3795 case IFLA_INET6_STATS:
3796 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3797 break;
3798 case IFLA_INET6_ICMP6STATS:
3799 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3800 break;
3804 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3805 u32 pid, u32 seq, int event, unsigned int flags)
3807 struct net_device *dev = idev->dev;
3808 struct nlattr *nla;
3809 struct ifinfomsg *hdr;
3810 struct nlmsghdr *nlh;
3811 void *protoinfo;
3812 struct ifla_cacheinfo ci;
3814 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3815 if (nlh == NULL)
3816 return -EMSGSIZE;
3818 hdr = nlmsg_data(nlh);
3819 hdr->ifi_family = AF_INET6;
3820 hdr->__ifi_pad = 0;
3821 hdr->ifi_type = dev->type;
3822 hdr->ifi_index = dev->ifindex;
3823 hdr->ifi_flags = dev_get_flags(dev);
3824 hdr->ifi_change = 0;
3826 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3828 if (dev->addr_len)
3829 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3831 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3832 if (dev->ifindex != dev->iflink)
3833 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3835 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3836 if (protoinfo == NULL)
3837 goto nla_put_failure;
3839 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3841 ci.max_reasm_len = IPV6_MAXPLEN;
3842 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3843 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3844 ci.reachable_time = idev->nd_parms->reachable_time;
3845 ci.retrans_time = idev->nd_parms->retrans_time;
3846 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3848 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3849 if (nla == NULL)
3850 goto nla_put_failure;
3851 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3853 /* XXX - MC not implemented */
3855 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3856 if (nla == NULL)
3857 goto nla_put_failure;
3858 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3860 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3861 if (nla == NULL)
3862 goto nla_put_failure;
3863 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3865 nla_nest_end(skb, protoinfo);
3866 return nlmsg_end(skb, nlh);
3868 nla_put_failure:
3869 nlmsg_cancel(skb, nlh);
3870 return -EMSGSIZE;
3873 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3875 struct net *net = sock_net(skb->sk);
3876 int h, s_h;
3877 int idx = 0, s_idx;
3878 struct net_device *dev;
3879 struct inet6_dev *idev;
3880 struct hlist_head *head;
3881 struct hlist_node *node;
3883 s_h = cb->args[0];
3884 s_idx = cb->args[1];
3886 rcu_read_lock();
3887 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3888 idx = 0;
3889 head = &net->dev_index_head[h];
3890 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3891 if (idx < s_idx)
3892 goto cont;
3893 idev = __in6_dev_get(dev);
3894 if (!idev)
3895 goto cont;
3896 if (inet6_fill_ifinfo(skb, idev,
3897 NETLINK_CB(cb->skb).pid,
3898 cb->nlh->nlmsg_seq,
3899 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3900 goto out;
3901 cont:
3902 idx++;
3905 out:
3906 rcu_read_unlock();
3907 cb->args[1] = idx;
3908 cb->args[0] = h;
3910 return skb->len;
3913 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3915 struct sk_buff *skb;
3916 struct net *net = dev_net(idev->dev);
3917 int err = -ENOBUFS;
3919 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3920 if (skb == NULL)
3921 goto errout;
3923 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3924 if (err < 0) {
3925 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3926 WARN_ON(err == -EMSGSIZE);
3927 kfree_skb(skb);
3928 goto errout;
3930 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3931 return;
3932 errout:
3933 if (err < 0)
3934 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3937 static inline size_t inet6_prefix_nlmsg_size(void)
3939 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3940 + nla_total_size(sizeof(struct in6_addr))
3941 + nla_total_size(sizeof(struct prefix_cacheinfo));
3944 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3945 struct prefix_info *pinfo, u32 pid, u32 seq,
3946 int event, unsigned int flags)
3948 struct prefixmsg *pmsg;
3949 struct nlmsghdr *nlh;
3950 struct prefix_cacheinfo ci;
3952 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3953 if (nlh == NULL)
3954 return -EMSGSIZE;
3956 pmsg = nlmsg_data(nlh);
3957 pmsg->prefix_family = AF_INET6;
3958 pmsg->prefix_pad1 = 0;
3959 pmsg->prefix_pad2 = 0;
3960 pmsg->prefix_ifindex = idev->dev->ifindex;
3961 pmsg->prefix_len = pinfo->prefix_len;
3962 pmsg->prefix_type = pinfo->type;
3963 pmsg->prefix_pad3 = 0;
3964 pmsg->prefix_flags = 0;
3965 if (pinfo->onlink)
3966 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3967 if (pinfo->autoconf)
3968 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3970 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3972 ci.preferred_time = ntohl(pinfo->prefered);
3973 ci.valid_time = ntohl(pinfo->valid);
3974 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3976 return nlmsg_end(skb, nlh);
3978 nla_put_failure:
3979 nlmsg_cancel(skb, nlh);
3980 return -EMSGSIZE;
3983 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3984 struct prefix_info *pinfo)
3986 struct sk_buff *skb;
3987 struct net *net = dev_net(idev->dev);
3988 int err = -ENOBUFS;
3990 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3991 if (skb == NULL)
3992 goto errout;
3994 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3995 if (err < 0) {
3996 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3997 WARN_ON(err == -EMSGSIZE);
3998 kfree_skb(skb);
3999 goto errout;
4001 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4002 return;
4003 errout:
4004 if (err < 0)
4005 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4008 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4010 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4012 switch (event) {
4013 case RTM_NEWADDR:
4015 * If the address was optimistic
4016 * we inserted the route at the start of
4017 * our DAD process, so we don't need
4018 * to do it again
4020 if (!(ifp->rt->rt6i_node))
4021 ip6_ins_rt(ifp->rt);
4022 if (ifp->idev->cnf.forwarding)
4023 addrconf_join_anycast(ifp);
4024 break;
4025 case RTM_DELADDR:
4026 if (ifp->idev->cnf.forwarding)
4027 addrconf_leave_anycast(ifp);
4028 addrconf_leave_solict(ifp->idev, &ifp->addr);
4029 dst_hold(&ifp->rt->u.dst);
4030 if (ip6_del_rt(ifp->rt))
4031 dst_free(&ifp->rt->u.dst);
4032 break;
4036 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4038 rcu_read_lock_bh();
4039 if (likely(ifp->idev->dead == 0))
4040 __ipv6_ifa_notify(event, ifp);
4041 rcu_read_unlock_bh();
4044 #ifdef CONFIG_SYSCTL
4046 static
4047 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4048 void __user *buffer, size_t *lenp, loff_t *ppos)
4050 int *valp = ctl->data;
4051 int val = *valp;
4052 loff_t pos = *ppos;
4053 int ret;
4055 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4057 if (write)
4058 ret = addrconf_fixup_forwarding(ctl, valp, val);
4059 if (ret)
4060 *ppos = pos;
4061 return ret;
4064 static void dev_disable_change(struct inet6_dev *idev)
4066 if (!idev || !idev->dev)
4067 return;
4069 if (idev->cnf.disable_ipv6)
4070 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4071 else
4072 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4075 static void addrconf_disable_change(struct net *net, __s32 newf)
4077 struct net_device *dev;
4078 struct inet6_dev *idev;
4080 rcu_read_lock();
4081 for_each_netdev_rcu(net, dev) {
4082 idev = __in6_dev_get(dev);
4083 if (idev) {
4084 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4085 idev->cnf.disable_ipv6 = newf;
4086 if (changed)
4087 dev_disable_change(idev);
4090 rcu_read_unlock();
4093 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4095 struct net *net;
4097 net = (struct net *)table->extra2;
4099 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4100 return 0;
4102 if (!rtnl_trylock()) {
4103 /* Restore the original values before restarting */
4104 *p = old;
4105 return restart_syscall();
4108 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4109 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4110 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4111 addrconf_disable_change(net, newf);
4112 } else if ((!*p) ^ (!old))
4113 dev_disable_change((struct inet6_dev *)table->extra1);
4115 rtnl_unlock();
4116 return 0;
4119 static
4120 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4121 void __user *buffer, size_t *lenp, loff_t *ppos)
4123 int *valp = ctl->data;
4124 int val = *valp;
4125 loff_t pos = *ppos;
4126 int ret;
4128 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4130 if (write)
4131 ret = addrconf_disable_ipv6(ctl, valp, val);
4132 if (ret)
4133 *ppos = pos;
4134 return ret;
4137 static struct addrconf_sysctl_table
4139 struct ctl_table_header *sysctl_header;
4140 ctl_table addrconf_vars[DEVCONF_MAX+1];
4141 char *dev_name;
4142 } addrconf_sysctl __read_mostly = {
4143 .sysctl_header = NULL,
4144 .addrconf_vars = {
4146 .procname = "forwarding",
4147 .data = &ipv6_devconf.forwarding,
4148 .maxlen = sizeof(int),
4149 .mode = 0644,
4150 .proc_handler = addrconf_sysctl_forward,
4153 .procname = "hop_limit",
4154 .data = &ipv6_devconf.hop_limit,
4155 .maxlen = sizeof(int),
4156 .mode = 0644,
4157 .proc_handler = proc_dointvec,
4160 .procname = "mtu",
4161 .data = &ipv6_devconf.mtu6,
4162 .maxlen = sizeof(int),
4163 .mode = 0644,
4164 .proc_handler = proc_dointvec,
4167 .procname = "accept_ra",
4168 .data = &ipv6_devconf.accept_ra,
4169 .maxlen = sizeof(int),
4170 .mode = 0644,
4171 .proc_handler = proc_dointvec,
4174 .procname = "accept_redirects",
4175 .data = &ipv6_devconf.accept_redirects,
4176 .maxlen = sizeof(int),
4177 .mode = 0644,
4178 .proc_handler = proc_dointvec,
4181 .procname = "autoconf",
4182 .data = &ipv6_devconf.autoconf,
4183 .maxlen = sizeof(int),
4184 .mode = 0644,
4185 .proc_handler = proc_dointvec,
4188 .procname = "dad_transmits",
4189 .data = &ipv6_devconf.dad_transmits,
4190 .maxlen = sizeof(int),
4191 .mode = 0644,
4192 .proc_handler = proc_dointvec,
4195 .procname = "router_solicitations",
4196 .data = &ipv6_devconf.rtr_solicits,
4197 .maxlen = sizeof(int),
4198 .mode = 0644,
4199 .proc_handler = proc_dointvec,
4202 .procname = "router_solicitation_interval",
4203 .data = &ipv6_devconf.rtr_solicit_interval,
4204 .maxlen = sizeof(int),
4205 .mode = 0644,
4206 .proc_handler = proc_dointvec_jiffies,
4209 .procname = "router_solicitation_delay",
4210 .data = &ipv6_devconf.rtr_solicit_delay,
4211 .maxlen = sizeof(int),
4212 .mode = 0644,
4213 .proc_handler = proc_dointvec_jiffies,
4216 .procname = "force_mld_version",
4217 .data = &ipv6_devconf.force_mld_version,
4218 .maxlen = sizeof(int),
4219 .mode = 0644,
4220 .proc_handler = proc_dointvec,
4222 #ifdef CONFIG_IPV6_PRIVACY
4224 .procname = "use_tempaddr",
4225 .data = &ipv6_devconf.use_tempaddr,
4226 .maxlen = sizeof(int),
4227 .mode = 0644,
4228 .proc_handler = proc_dointvec,
4231 .procname = "temp_valid_lft",
4232 .data = &ipv6_devconf.temp_valid_lft,
4233 .maxlen = sizeof(int),
4234 .mode = 0644,
4235 .proc_handler = proc_dointvec,
4238 .procname = "temp_prefered_lft",
4239 .data = &ipv6_devconf.temp_prefered_lft,
4240 .maxlen = sizeof(int),
4241 .mode = 0644,
4242 .proc_handler = proc_dointvec,
4245 .procname = "regen_max_retry",
4246 .data = &ipv6_devconf.regen_max_retry,
4247 .maxlen = sizeof(int),
4248 .mode = 0644,
4249 .proc_handler = proc_dointvec,
4252 .procname = "max_desync_factor",
4253 .data = &ipv6_devconf.max_desync_factor,
4254 .maxlen = sizeof(int),
4255 .mode = 0644,
4256 .proc_handler = proc_dointvec,
4258 #endif
4260 .procname = "max_addresses",
4261 .data = &ipv6_devconf.max_addresses,
4262 .maxlen = sizeof(int),
4263 .mode = 0644,
4264 .proc_handler = proc_dointvec,
4267 .procname = "accept_ra_defrtr",
4268 .data = &ipv6_devconf.accept_ra_defrtr,
4269 .maxlen = sizeof(int),
4270 .mode = 0644,
4271 .proc_handler = proc_dointvec,
4274 .procname = "accept_ra_pinfo",
4275 .data = &ipv6_devconf.accept_ra_pinfo,
4276 .maxlen = sizeof(int),
4277 .mode = 0644,
4278 .proc_handler = proc_dointvec,
4280 #ifdef CONFIG_IPV6_ROUTER_PREF
4282 .procname = "accept_ra_rtr_pref",
4283 .data = &ipv6_devconf.accept_ra_rtr_pref,
4284 .maxlen = sizeof(int),
4285 .mode = 0644,
4286 .proc_handler = proc_dointvec,
4289 .procname = "router_probe_interval",
4290 .data = &ipv6_devconf.rtr_probe_interval,
4291 .maxlen = sizeof(int),
4292 .mode = 0644,
4293 .proc_handler = proc_dointvec_jiffies,
4295 #ifdef CONFIG_IPV6_ROUTE_INFO
4297 .procname = "accept_ra_rt_info_max_plen",
4298 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4299 .maxlen = sizeof(int),
4300 .mode = 0644,
4301 .proc_handler = proc_dointvec,
4303 #endif
4304 #endif
4306 .procname = "proxy_ndp",
4307 .data = &ipv6_devconf.proxy_ndp,
4308 .maxlen = sizeof(int),
4309 .mode = 0644,
4310 .proc_handler = proc_dointvec,
4313 .procname = "accept_source_route",
4314 .data = &ipv6_devconf.accept_source_route,
4315 .maxlen = sizeof(int),
4316 .mode = 0644,
4317 .proc_handler = proc_dointvec,
4319 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4321 .procname = "optimistic_dad",
4322 .data = &ipv6_devconf.optimistic_dad,
4323 .maxlen = sizeof(int),
4324 .mode = 0644,
4325 .proc_handler = proc_dointvec,
4328 #endif
4329 #ifdef CONFIG_IPV6_MROUTE
4331 .procname = "mc_forwarding",
4332 .data = &ipv6_devconf.mc_forwarding,
4333 .maxlen = sizeof(int),
4334 .mode = 0444,
4335 .proc_handler = proc_dointvec,
4337 #endif
4339 .procname = "disable_ipv6",
4340 .data = &ipv6_devconf.disable_ipv6,
4341 .maxlen = sizeof(int),
4342 .mode = 0644,
4343 .proc_handler = addrconf_sysctl_disable,
4346 .procname = "accept_dad",
4347 .data = &ipv6_devconf.accept_dad,
4348 .maxlen = sizeof(int),
4349 .mode = 0644,
4350 .proc_handler = proc_dointvec,
4353 .procname = "force_tllao",
4354 .data = &ipv6_devconf.force_tllao,
4355 .maxlen = sizeof(int),
4356 .mode = 0644,
4357 .proc_handler = proc_dointvec
4360 /* sentinel */
4365 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4366 struct inet6_dev *idev, struct ipv6_devconf *p)
4368 int i;
4369 struct addrconf_sysctl_table *t;
4371 #define ADDRCONF_CTL_PATH_DEV 3
4373 struct ctl_path addrconf_ctl_path[] = {
4374 { .procname = "net", },
4375 { .procname = "ipv6", },
4376 { .procname = "conf", },
4377 { /* to be set */ },
4378 { },
4382 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4383 if (t == NULL)
4384 goto out;
4386 for (i = 0; t->addrconf_vars[i].data; i++) {
4387 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4388 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4389 t->addrconf_vars[i].extra2 = net;
4393 * Make a copy of dev_name, because '.procname' is regarded as const
4394 * by sysctl and we wouldn't want anyone to change it under our feet
4395 * (see SIOCSIFNAME).
4397 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4398 if (!t->dev_name)
4399 goto free;
4401 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4403 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4404 t->addrconf_vars);
4405 if (t->sysctl_header == NULL)
4406 goto free_procname;
4408 p->sysctl = t;
4409 return 0;
4411 free_procname:
4412 kfree(t->dev_name);
4413 free:
4414 kfree(t);
4415 out:
4416 return -ENOBUFS;
4419 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4421 struct addrconf_sysctl_table *t;
4423 if (p->sysctl == NULL)
4424 return;
4426 t = p->sysctl;
4427 p->sysctl = NULL;
4428 unregister_sysctl_table(t->sysctl_header);
4429 kfree(t->dev_name);
4430 kfree(t);
4433 static void addrconf_sysctl_register(struct inet6_dev *idev)
4435 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4436 &ndisc_ifinfo_sysctl_change);
4437 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4438 idev, &idev->cnf);
4441 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4443 __addrconf_sysctl_unregister(&idev->cnf);
4444 neigh_sysctl_unregister(idev->nd_parms);
4448 #endif
4450 static int __net_init addrconf_init_net(struct net *net)
4452 int err;
4453 struct ipv6_devconf *all, *dflt;
4455 err = -ENOMEM;
4456 all = &ipv6_devconf;
4457 dflt = &ipv6_devconf_dflt;
4459 if (!net_eq(net, &init_net)) {
4460 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4461 if (all == NULL)
4462 goto err_alloc_all;
4464 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4465 if (dflt == NULL)
4466 goto err_alloc_dflt;
4467 } else {
4468 /* these will be inherited by all namespaces */
4469 dflt->autoconf = ipv6_defaults.autoconf;
4470 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4473 net->ipv6.devconf_all = all;
4474 net->ipv6.devconf_dflt = dflt;
4476 #ifdef CONFIG_SYSCTL
4477 err = __addrconf_sysctl_register(net, "all", NULL, all);
4478 if (err < 0)
4479 goto err_reg_all;
4481 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4482 if (err < 0)
4483 goto err_reg_dflt;
4484 #endif
4485 return 0;
4487 #ifdef CONFIG_SYSCTL
4488 err_reg_dflt:
4489 __addrconf_sysctl_unregister(all);
4490 err_reg_all:
4491 kfree(dflt);
4492 #endif
4493 err_alloc_dflt:
4494 kfree(all);
4495 err_alloc_all:
4496 return err;
4499 static void __net_exit addrconf_exit_net(struct net *net)
4501 #ifdef CONFIG_SYSCTL
4502 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4503 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4504 #endif
4505 if (!net_eq(net, &init_net)) {
4506 kfree(net->ipv6.devconf_dflt);
4507 kfree(net->ipv6.devconf_all);
4511 static struct pernet_operations addrconf_ops = {
4512 .init = addrconf_init_net,
4513 .exit = addrconf_exit_net,
4517 * Device notifier
4520 int register_inet6addr_notifier(struct notifier_block *nb)
4522 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4524 EXPORT_SYMBOL(register_inet6addr_notifier);
4526 int unregister_inet6addr_notifier(struct notifier_block *nb)
4528 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4530 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4533 * Init / cleanup code
4536 int __init addrconf_init(void)
4538 int i, err;
4540 err = ipv6_addr_label_init();
4541 if (err < 0) {
4542 printk(KERN_CRIT "IPv6 Addrconf:"
4543 " cannot initialize default policy table: %d.\n", err);
4544 return err;
4547 register_pernet_subsys(&addrconf_ops);
4549 /* The addrconf netdev notifier requires that loopback_dev
4550 * has it's ipv6 private information allocated and setup
4551 * before it can bring up and give link-local addresses
4552 * to other devices which are up.
4554 * Unfortunately, loopback_dev is not necessarily the first
4555 * entry in the global dev_base list of net devices. In fact,
4556 * it is likely to be the very last entry on that list.
4557 * So this causes the notifier registry below to try and
4558 * give link-local addresses to all devices besides loopback_dev
4559 * first, then loopback_dev, which cases all the non-loopback_dev
4560 * devices to fail to get a link-local address.
4562 * So, as a temporary fix, allocate the ipv6 structure for
4563 * loopback_dev first by hand.
4564 * Longer term, all of the dependencies ipv6 has upon the loopback
4565 * device and it being up should be removed.
4567 rtnl_lock();
4568 if (!ipv6_add_dev(init_net.loopback_dev))
4569 err = -ENOMEM;
4570 rtnl_unlock();
4571 if (err)
4572 goto errlo;
4574 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4575 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4577 register_netdevice_notifier(&ipv6_dev_notf);
4579 addrconf_verify(0);
4581 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4582 if (err < 0)
4583 goto errout;
4585 /* Only the first call to __rtnl_register can fail */
4586 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4587 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4588 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4589 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4590 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4592 ipv6_addr_label_rtnl_register();
4594 return 0;
4595 errout:
4596 unregister_netdevice_notifier(&ipv6_dev_notf);
4597 errlo:
4598 unregister_pernet_subsys(&addrconf_ops);
4600 return err;
4603 void addrconf_cleanup(void)
4605 struct net_device *dev;
4606 int i;
4608 unregister_netdevice_notifier(&ipv6_dev_notf);
4609 unregister_pernet_subsys(&addrconf_ops);
4611 rtnl_lock();
4613 /* clean dev list */
4614 for_each_netdev(&init_net, dev) {
4615 if (__in6_dev_get(dev) == NULL)
4616 continue;
4617 addrconf_ifdown(dev, 1);
4619 addrconf_ifdown(init_net.loopback_dev, 2);
4622 * Check hash table.
4624 spin_lock_bh(&addrconf_hash_lock);
4625 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4626 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4627 spin_unlock_bh(&addrconf_hash_lock);
4629 del_timer(&addr_chk_timer);
4630 rtnl_unlock();