Linux 2.6.34.3
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / addrconf.c
blob844ffc5a7420f423441ff35bd12450f00b094536
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 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
100 #define INFINITY_LIFE_TIME 0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
114 #endif
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
124 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
125 static int ipv6_count_addresses(struct inet6_dev *idev);
128 * Configured unicast address hash table
130 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
131 static DEFINE_RWLOCK(addrconf_hash_lock);
133 static void addrconf_verify(unsigned long);
135 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
136 static DEFINE_SPINLOCK(addrconf_verify_lock);
138 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
139 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
141 static void addrconf_bonding_change(struct net_device *dev,
142 unsigned long event);
143 static int addrconf_ifdown(struct net_device *dev, int how);
145 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
146 static void addrconf_dad_timer(unsigned long data);
147 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
148 static void addrconf_dad_run(struct inet6_dev *idev);
149 static void addrconf_rs_timer(unsigned long data);
150 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
151 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
153 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
154 struct prefix_info *pinfo);
155 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
156 struct net_device *dev);
158 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
160 static struct ipv6_devconf ipv6_devconf __read_mostly = {
161 .forwarding = 0,
162 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
163 .mtu6 = IPV6_MIN_MTU,
164 .accept_ra = 1,
165 .accept_redirects = 1,
166 .autoconf = 1,
167 .force_mld_version = 0,
168 .dad_transmits = 1,
169 .rtr_solicits = MAX_RTR_SOLICITATIONS,
170 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
171 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
172 #ifdef CONFIG_IPV6_PRIVACY
173 .use_tempaddr = 0,
174 .temp_valid_lft = TEMP_VALID_LIFETIME,
175 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
176 .regen_max_retry = REGEN_MAX_RETRY,
177 .max_desync_factor = MAX_DESYNC_FACTOR,
178 #endif
179 .max_addresses = IPV6_MAX_ADDRESSES,
180 .accept_ra_defrtr = 1,
181 .accept_ra_pinfo = 1,
182 #ifdef CONFIG_IPV6_ROUTER_PREF
183 .accept_ra_rtr_pref = 1,
184 .rtr_probe_interval = 60 * HZ,
185 #ifdef CONFIG_IPV6_ROUTE_INFO
186 .accept_ra_rt_info_max_plen = 0,
187 #endif
188 #endif
189 .proxy_ndp = 0,
190 .accept_source_route = 0, /* we do not accept RH0 by default. */
191 .disable_ipv6 = 0,
192 .accept_dad = 1,
195 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
196 .forwarding = 0,
197 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
198 .mtu6 = IPV6_MIN_MTU,
199 .accept_ra = 1,
200 .accept_redirects = 1,
201 .autoconf = 1,
202 .dad_transmits = 1,
203 .rtr_solicits = MAX_RTR_SOLICITATIONS,
204 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
205 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
206 #ifdef CONFIG_IPV6_PRIVACY
207 .use_tempaddr = 0,
208 .temp_valid_lft = TEMP_VALID_LIFETIME,
209 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
210 .regen_max_retry = REGEN_MAX_RETRY,
211 .max_desync_factor = MAX_DESYNC_FACTOR,
212 #endif
213 .max_addresses = IPV6_MAX_ADDRESSES,
214 .accept_ra_defrtr = 1,
215 .accept_ra_pinfo = 1,
216 #ifdef CONFIG_IPV6_ROUTER_PREF
217 .accept_ra_rtr_pref = 1,
218 .rtr_probe_interval = 60 * HZ,
219 #ifdef CONFIG_IPV6_ROUTE_INFO
220 .accept_ra_rt_info_max_plen = 0,
221 #endif
222 #endif
223 .proxy_ndp = 0,
224 .accept_source_route = 0, /* we do not accept RH0 by default. */
225 .disable_ipv6 = 0,
226 .accept_dad = 1,
229 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
230 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
231 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
232 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
233 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
235 /* Check if a valid qdisc is available */
236 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
238 return !qdisc_tx_is_noop(dev);
241 /* Check if a route is valid prefix route */
242 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
244 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
247 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
249 if (del_timer(&ifp->timer))
250 __in6_ifa_put(ifp);
253 enum addrconf_timer_t
255 AC_NONE,
256 AC_DAD,
257 AC_RS,
260 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
261 enum addrconf_timer_t what,
262 unsigned long when)
264 if (!del_timer(&ifp->timer))
265 in6_ifa_hold(ifp);
267 switch (what) {
268 case AC_DAD:
269 ifp->timer.function = addrconf_dad_timer;
270 break;
271 case AC_RS:
272 ifp->timer.function = addrconf_rs_timer;
273 break;
274 default:;
276 ifp->timer.expires = jiffies + when;
277 add_timer(&ifp->timer);
280 static int snmp6_alloc_dev(struct inet6_dev *idev)
282 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
283 sizeof(struct ipstats_mib)) < 0)
284 goto err_ip;
285 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
286 sizeof(struct icmpv6_mib)) < 0)
287 goto err_icmp;
288 if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
289 sizeof(struct icmpv6msg_mib)) < 0)
290 goto err_icmpmsg;
292 return 0;
294 err_icmpmsg:
295 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
296 err_icmp:
297 snmp_mib_free((void __percpu **)idev->stats.ipv6);
298 err_ip:
299 return -ENOMEM;
302 static void snmp6_free_dev(struct inet6_dev *idev)
304 snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
305 snmp_mib_free((void __percpu **)idev->stats.icmpv6);
306 snmp_mib_free((void __percpu **)idev->stats.ipv6);
309 /* Nobody refers to this device, we may destroy it. */
311 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
313 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
314 kfree(idev);
317 void in6_dev_finish_destroy(struct inet6_dev *idev)
319 struct net_device *dev = idev->dev;
321 WARN_ON(idev->addr_list != NULL);
322 WARN_ON(idev->mc_list != NULL);
324 #ifdef NET_REFCNT_DEBUG
325 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
326 #endif
327 dev_put(dev);
328 if (!idev->dead) {
329 printk("Freeing alive inet6 device %p\n", idev);
330 return;
332 snmp6_free_dev(idev);
333 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
336 EXPORT_SYMBOL(in6_dev_finish_destroy);
338 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
340 struct inet6_dev *ndev;
342 ASSERT_RTNL();
344 if (dev->mtu < IPV6_MIN_MTU)
345 return NULL;
347 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
349 if (ndev == NULL)
350 return NULL;
352 rwlock_init(&ndev->lock);
353 ndev->dev = dev;
354 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
355 ndev->cnf.mtu6 = dev->mtu;
356 ndev->cnf.sysctl = NULL;
357 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
358 if (ndev->nd_parms == NULL) {
359 kfree(ndev);
360 return NULL;
362 if (ndev->cnf.forwarding)
363 dev_disable_lro(dev);
364 /* We refer to the device */
365 dev_hold(dev);
367 if (snmp6_alloc_dev(ndev) < 0) {
368 ADBG((KERN_WARNING
369 "%s(): cannot allocate memory for statistics; dev=%s.\n",
370 __func__, dev->name));
371 neigh_parms_release(&nd_tbl, ndev->nd_parms);
372 ndev->dead = 1;
373 in6_dev_finish_destroy(ndev);
374 return NULL;
377 if (snmp6_register_dev(ndev) < 0) {
378 ADBG((KERN_WARNING
379 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
380 __func__, dev->name));
381 neigh_parms_release(&nd_tbl, ndev->nd_parms);
382 ndev->dead = 1;
383 in6_dev_finish_destroy(ndev);
384 return NULL;
387 /* One reference from device. We must do this before
388 * we invoke __ipv6_regen_rndid().
390 in6_dev_hold(ndev);
392 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
393 ndev->cnf.accept_dad = -1;
395 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
396 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
397 printk(KERN_INFO
398 "%s: Disabled Multicast RS\n",
399 dev->name);
400 ndev->cnf.rtr_solicits = 0;
402 #endif
404 #ifdef CONFIG_IPV6_PRIVACY
405 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
406 if ((dev->flags&IFF_LOOPBACK) ||
407 dev->type == ARPHRD_TUNNEL ||
408 dev->type == ARPHRD_TUNNEL6 ||
409 dev->type == ARPHRD_SIT ||
410 dev->type == ARPHRD_NONE) {
411 printk(KERN_INFO
412 "%s: Disabled Privacy Extensions\n",
413 dev->name);
414 ndev->cnf.use_tempaddr = -1;
415 } else {
416 in6_dev_hold(ndev);
417 ipv6_regen_rndid((unsigned long) ndev);
419 #endif
421 if (netif_running(dev) && addrconf_qdisc_ok(dev))
422 ndev->if_flags |= IF_READY;
424 ipv6_mc_init_dev(ndev);
425 ndev->tstamp = jiffies;
426 addrconf_sysctl_register(ndev);
427 /* protected by rtnl_lock */
428 rcu_assign_pointer(dev->ip6_ptr, ndev);
430 /* Join all-node multicast group */
431 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
433 return ndev;
436 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
438 struct inet6_dev *idev;
440 ASSERT_RTNL();
442 if ((idev = __in6_dev_get(dev)) == NULL) {
443 if ((idev = ipv6_add_dev(dev)) == NULL)
444 return NULL;
447 if (dev->flags&IFF_UP)
448 ipv6_mc_up(idev);
449 return idev;
452 #ifdef CONFIG_SYSCTL
453 static void dev_forward_change(struct inet6_dev *idev)
455 struct net_device *dev;
456 struct inet6_ifaddr *ifa;
458 if (!idev)
459 return;
460 dev = idev->dev;
461 if (idev->cnf.forwarding)
462 dev_disable_lro(dev);
463 if (dev && (dev->flags & IFF_MULTICAST)) {
464 if (idev->cnf.forwarding)
465 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
466 else
467 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
469 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
470 if (ifa->flags&IFA_F_TENTATIVE)
471 continue;
472 if (idev->cnf.forwarding)
473 addrconf_join_anycast(ifa);
474 else
475 addrconf_leave_anycast(ifa);
480 static void addrconf_forward_change(struct net *net, __s32 newf)
482 struct net_device *dev;
483 struct inet6_dev *idev;
485 rcu_read_lock();
486 for_each_netdev_rcu(net, dev) {
487 idev = __in6_dev_get(dev);
488 if (idev) {
489 int changed = (!idev->cnf.forwarding) ^ (!newf);
490 idev->cnf.forwarding = newf;
491 if (changed)
492 dev_forward_change(idev);
495 rcu_read_unlock();
498 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
500 struct net *net;
502 net = (struct net *)table->extra2;
503 if (p == &net->ipv6.devconf_dflt->forwarding)
504 return 0;
506 if (!rtnl_trylock()) {
507 /* Restore the original values before restarting */
508 *p = old;
509 return restart_syscall();
512 if (p == &net->ipv6.devconf_all->forwarding) {
513 __s32 newf = net->ipv6.devconf_all->forwarding;
514 net->ipv6.devconf_dflt->forwarding = newf;
515 addrconf_forward_change(net, newf);
516 } else if ((!*p) ^ (!old))
517 dev_forward_change((struct inet6_dev *)table->extra1);
518 rtnl_unlock();
520 if (*p)
521 rt6_purge_dflt_routers(net);
522 return 1;
524 #endif
526 /* Nobody refers to this ifaddr, destroy it */
528 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
530 WARN_ON(ifp->if_next != NULL);
531 WARN_ON(ifp->lst_next != NULL);
533 #ifdef NET_REFCNT_DEBUG
534 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
535 #endif
537 in6_dev_put(ifp->idev);
539 if (del_timer(&ifp->timer))
540 printk("Timer is still running, when freeing ifa=%p\n", ifp);
542 if (!ifp->dead) {
543 printk("Freeing alive inet6 address %p\n", ifp);
544 return;
546 dst_release(&ifp->rt->u.dst);
548 kfree(ifp);
551 static void
552 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
554 struct inet6_ifaddr *ifa, **ifap;
555 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
558 * Each device address list is sorted in order of scope -
559 * global before linklocal.
561 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
562 ifap = &ifa->if_next) {
563 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
564 break;
567 ifp->if_next = *ifap;
568 *ifap = ifp;
572 * Hash function taken from net_alias.c
574 static u8 ipv6_addr_hash(const struct in6_addr *addr)
576 __u32 word;
579 * We perform the hash function over the last 64 bits of the address
580 * This will include the IEEE address token on links that support it.
583 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
584 word ^= (word >> 16);
585 word ^= (word >> 8);
587 return ((word ^ (word >> 4)) & 0x0f);
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 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 write_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 ifa->timer.data = (unsigned long) ifa;
647 ifa->scope = scope;
648 ifa->prefix_len = pfxlen;
649 ifa->flags = flags | IFA_F_TENTATIVE;
650 ifa->cstamp = ifa->tstamp = jiffies;
652 ifa->rt = rt;
655 * part one of RFC 4429, section 3.3
656 * We should not configure an address as
657 * optimistic if we do not yet know the link
658 * layer address of our nexhop router
661 if (rt->rt6i_nexthop == NULL)
662 ifa->flags &= ~IFA_F_OPTIMISTIC;
664 ifa->idev = idev;
665 in6_dev_hold(idev);
666 /* For caller */
667 in6_ifa_hold(ifa);
669 /* Add to big hash table */
670 hash = ipv6_addr_hash(addr);
672 ifa->lst_next = inet6_addr_lst[hash];
673 inet6_addr_lst[hash] = ifa;
674 in6_ifa_hold(ifa);
675 write_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 ifa->tmp_next = idev->tempaddr_list;
684 idev->tempaddr_list = ifa;
685 in6_ifa_hold(ifa);
687 #endif
689 in6_ifa_hold(ifa);
690 write_unlock(&idev->lock);
691 out2:
692 rcu_read_unlock_bh();
694 if (likely(err == 0))
695 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
696 else {
697 kfree(ifa);
698 ifa = ERR_PTR(err);
701 return ifa;
702 out:
703 write_unlock(&addrconf_hash_lock);
704 goto out2;
707 /* This function wants to get referenced ifp and releases it before return */
709 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
711 struct inet6_ifaddr *ifa, **ifap;
712 struct inet6_dev *idev = ifp->idev;
713 int hash;
714 int deleted = 0, onlink = 0;
715 unsigned long expires = jiffies;
717 hash = ipv6_addr_hash(&ifp->addr);
719 ifp->dead = 1;
721 write_lock_bh(&addrconf_hash_lock);
722 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
723 ifap = &ifa->lst_next) {
724 if (ifa == ifp) {
725 *ifap = ifa->lst_next;
726 __in6_ifa_put(ifp);
727 ifa->lst_next = NULL;
728 break;
731 write_unlock_bh(&addrconf_hash_lock);
733 write_lock_bh(&idev->lock);
734 #ifdef CONFIG_IPV6_PRIVACY
735 if (ifp->flags&IFA_F_TEMPORARY) {
736 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
737 ifap = &ifa->tmp_next) {
738 if (ifa == ifp) {
739 *ifap = ifa->tmp_next;
740 if (ifp->ifpub) {
741 in6_ifa_put(ifp->ifpub);
742 ifp->ifpub = NULL;
744 __in6_ifa_put(ifp);
745 ifa->tmp_next = NULL;
746 break;
750 #endif
752 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
753 if (ifa == ifp) {
754 *ifap = ifa->if_next;
755 __in6_ifa_put(ifp);
756 ifa->if_next = NULL;
757 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
758 break;
759 deleted = 1;
760 continue;
761 } else if (ifp->flags & IFA_F_PERMANENT) {
762 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
763 ifp->prefix_len)) {
764 if (ifa->flags & IFA_F_PERMANENT) {
765 onlink = 1;
766 if (deleted)
767 break;
768 } else {
769 unsigned long lifetime;
771 if (!onlink)
772 onlink = -1;
774 spin_lock(&ifa->lock);
776 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
778 * Note: Because this address is
779 * not permanent, lifetime <
780 * LONG_MAX / HZ here.
782 if (time_before(expires,
783 ifa->tstamp + lifetime * HZ))
784 expires = ifa->tstamp + lifetime * HZ;
785 spin_unlock(&ifa->lock);
789 ifap = &ifa->if_next;
791 write_unlock_bh(&idev->lock);
793 addrconf_del_timer(ifp);
795 ipv6_ifa_notify(RTM_DELADDR, ifp);
797 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
800 * Purge or update corresponding prefix
802 * 1) we don't purge prefix here if address was not permanent.
803 * prefix is managed by its own lifetime.
804 * 2) if there're no addresses, delete prefix.
805 * 3) if there're still other permanent address(es),
806 * corresponding prefix is still permanent.
807 * 4) otherwise, update prefix lifetime to the
808 * longest valid lifetime among the corresponding
809 * addresses on the device.
810 * Note: subsequent RA will update lifetime.
812 * --yoshfuji
814 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
815 struct in6_addr prefix;
816 struct rt6_info *rt;
817 struct net *net = dev_net(ifp->idev->dev);
818 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
819 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
821 if (rt && addrconf_is_prefix_route(rt)) {
822 if (onlink == 0) {
823 ip6_del_rt(rt);
824 rt = NULL;
825 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
826 rt->rt6i_expires = expires;
827 rt->rt6i_flags |= RTF_EXPIRES;
830 dst_release(&rt->u.dst);
833 in6_ifa_put(ifp);
836 #ifdef CONFIG_IPV6_PRIVACY
837 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
839 struct inet6_dev *idev = ifp->idev;
840 struct in6_addr addr, *tmpaddr;
841 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
842 unsigned long regen_advance;
843 int tmp_plen;
844 int ret = 0;
845 int max_addresses;
846 u32 addr_flags;
848 write_lock(&idev->lock);
849 if (ift) {
850 spin_lock_bh(&ift->lock);
851 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
852 spin_unlock_bh(&ift->lock);
853 tmpaddr = &addr;
854 } else {
855 tmpaddr = NULL;
857 retry:
858 in6_dev_hold(idev);
859 if (idev->cnf.use_tempaddr <= 0) {
860 write_unlock(&idev->lock);
861 printk(KERN_INFO
862 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
863 in6_dev_put(idev);
864 ret = -1;
865 goto out;
867 spin_lock_bh(&ifp->lock);
868 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
869 idev->cnf.use_tempaddr = -1; /*XXX*/
870 spin_unlock_bh(&ifp->lock);
871 write_unlock(&idev->lock);
872 printk(KERN_WARNING
873 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
874 in6_dev_put(idev);
875 ret = -1;
876 goto out;
878 in6_ifa_hold(ifp);
879 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
880 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
881 spin_unlock_bh(&ifp->lock);
882 write_unlock(&idev->lock);
883 printk(KERN_WARNING
884 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
885 in6_ifa_put(ifp);
886 in6_dev_put(idev);
887 ret = -1;
888 goto out;
890 memcpy(&addr.s6_addr[8], idev->rndid, 8);
891 tmp_valid_lft = min_t(__u32,
892 ifp->valid_lft,
893 idev->cnf.temp_valid_lft);
894 tmp_prefered_lft = min_t(__u32,
895 ifp->prefered_lft,
896 idev->cnf.temp_prefered_lft - desync_factor / HZ);
897 tmp_plen = ifp->prefix_len;
898 max_addresses = idev->cnf.max_addresses;
899 tmp_cstamp = ifp->cstamp;
900 tmp_tstamp = ifp->tstamp;
901 spin_unlock_bh(&ifp->lock);
903 regen_advance = idev->cnf.regen_max_retry *
904 idev->cnf.dad_transmits *
905 idev->nd_parms->retrans_time / HZ;
906 write_unlock(&idev->lock);
908 /* A temporary address is created only if this calculated Preferred
909 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
910 * an implementation must not create a temporary address with a zero
911 * Preferred Lifetime.
913 if (tmp_prefered_lft <= regen_advance) {
914 in6_ifa_put(ifp);
915 in6_dev_put(idev);
916 ret = -1;
917 goto out;
920 addr_flags = IFA_F_TEMPORARY;
921 /* set in addrconf_prefix_rcv() */
922 if (ifp->flags & IFA_F_OPTIMISTIC)
923 addr_flags |= IFA_F_OPTIMISTIC;
925 ift = !max_addresses ||
926 ipv6_count_addresses(idev) < max_addresses ?
927 ipv6_add_addr(idev, &addr, tmp_plen,
928 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
929 addr_flags) : NULL;
930 if (!ift || IS_ERR(ift)) {
931 in6_ifa_put(ifp);
932 in6_dev_put(idev);
933 printk(KERN_INFO
934 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
935 tmpaddr = &addr;
936 write_lock(&idev->lock);
937 goto retry;
940 spin_lock_bh(&ift->lock);
941 ift->ifpub = ifp;
942 ift->valid_lft = tmp_valid_lft;
943 ift->prefered_lft = tmp_prefered_lft;
944 ift->cstamp = tmp_cstamp;
945 ift->tstamp = tmp_tstamp;
946 spin_unlock_bh(&ift->lock);
948 addrconf_dad_start(ift, 0);
949 in6_ifa_put(ift);
950 in6_dev_put(idev);
951 out:
952 return ret;
954 #endif
957 * Choose an appropriate source address (RFC3484)
959 enum {
960 IPV6_SADDR_RULE_INIT = 0,
961 IPV6_SADDR_RULE_LOCAL,
962 IPV6_SADDR_RULE_SCOPE,
963 IPV6_SADDR_RULE_PREFERRED,
964 #ifdef CONFIG_IPV6_MIP6
965 IPV6_SADDR_RULE_HOA,
966 #endif
967 IPV6_SADDR_RULE_OIF,
968 IPV6_SADDR_RULE_LABEL,
969 #ifdef CONFIG_IPV6_PRIVACY
970 IPV6_SADDR_RULE_PRIVACY,
971 #endif
972 IPV6_SADDR_RULE_ORCHID,
973 IPV6_SADDR_RULE_PREFIX,
974 IPV6_SADDR_RULE_MAX
977 struct ipv6_saddr_score {
978 int rule;
979 int addr_type;
980 struct inet6_ifaddr *ifa;
981 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
982 int scopedist;
983 int matchlen;
986 struct ipv6_saddr_dst {
987 const struct in6_addr *addr;
988 int ifindex;
989 int scope;
990 int label;
991 unsigned int prefs;
994 static inline int ipv6_saddr_preferred(int type)
996 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
997 return 1;
998 return 0;
1001 static int ipv6_get_saddr_eval(struct net *net,
1002 struct ipv6_saddr_score *score,
1003 struct ipv6_saddr_dst *dst,
1004 int i)
1006 int ret;
1008 if (i <= score->rule) {
1009 switch (i) {
1010 case IPV6_SADDR_RULE_SCOPE:
1011 ret = score->scopedist;
1012 break;
1013 case IPV6_SADDR_RULE_PREFIX:
1014 ret = score->matchlen;
1015 break;
1016 default:
1017 ret = !!test_bit(i, score->scorebits);
1019 goto out;
1022 switch (i) {
1023 case IPV6_SADDR_RULE_INIT:
1024 /* Rule 0: remember if hiscore is not ready yet */
1025 ret = !!score->ifa;
1026 break;
1027 case IPV6_SADDR_RULE_LOCAL:
1028 /* Rule 1: Prefer same address */
1029 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030 break;
1031 case IPV6_SADDR_RULE_SCOPE:
1032 /* Rule 2: Prefer appropriate scope
1034 * ret
1036 * -1 | d 15
1037 * ---+--+-+---> scope
1039 * | d is scope of the destination.
1040 * B-d | \
1041 * | \ <- smaller scope is better if
1042 * B-15 | \ if scope is enough for destinaion.
1043 * | ret = B - scope (-1 <= scope >= d <= 15).
1044 * d-C-1 | /
1045 * |/ <- greater is better
1046 * -C / if scope is not enough for destination.
1047 * /| ret = scope - C (-1 <= d < scope <= 15).
1049 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051 * Assume B = 0 and we get C > 29.
1053 ret = __ipv6_addr_src_scope(score->addr_type);
1054 if (ret >= dst->scope)
1055 ret = -ret;
1056 else
1057 ret -= 128; /* 30 is enough */
1058 score->scopedist = ret;
1059 break;
1060 case IPV6_SADDR_RULE_PREFERRED:
1061 /* Rule 3: Avoid deprecated and optimistic addresses */
1062 ret = ipv6_saddr_preferred(score->addr_type) ||
1063 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064 break;
1065 #ifdef CONFIG_IPV6_MIP6
1066 case IPV6_SADDR_RULE_HOA:
1068 /* Rule 4: Prefer home address */
1069 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071 break;
1073 #endif
1074 case IPV6_SADDR_RULE_OIF:
1075 /* Rule 5: Prefer outgoing interface */
1076 ret = (!dst->ifindex ||
1077 dst->ifindex == score->ifa->idev->dev->ifindex);
1078 break;
1079 case IPV6_SADDR_RULE_LABEL:
1080 /* Rule 6: Prefer matching label */
1081 ret = ipv6_addr_label(net,
1082 &score->ifa->addr, score->addr_type,
1083 score->ifa->idev->dev->ifindex) == dst->label;
1084 break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086 case IPV6_SADDR_RULE_PRIVACY:
1088 /* Rule 7: Prefer public address
1089 * Note: prefer temprary address if use_tempaddr >= 2
1091 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093 score->ifa->idev->cnf.use_tempaddr >= 2;
1094 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095 break;
1097 #endif
1098 case IPV6_SADDR_RULE_ORCHID:
1099 /* Rule 8-: Prefer ORCHID vs ORCHID or
1100 * non-ORCHID vs non-ORCHID
1102 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103 ipv6_addr_orchid(dst->addr));
1104 break;
1105 case IPV6_SADDR_RULE_PREFIX:
1106 /* Rule 8: Use longest matching prefix */
1107 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108 dst->addr);
1109 break;
1110 default:
1111 ret = 0;
1114 if (ret)
1115 __set_bit(i, score->scorebits);
1116 score->rule = i;
1117 out:
1118 return ret;
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122 const struct in6_addr *daddr, unsigned int prefs,
1123 struct in6_addr *saddr)
1125 struct ipv6_saddr_score scores[2],
1126 *score = &scores[0], *hiscore = &scores[1];
1127 struct ipv6_saddr_dst dst;
1128 struct net_device *dev;
1129 int dst_type;
1131 dst_type = __ipv6_addr_type(daddr);
1132 dst.addr = daddr;
1133 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134 dst.scope = __ipv6_addr_src_scope(dst_type);
1135 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136 dst.prefs = prefs;
1138 hiscore->rule = -1;
1139 hiscore->ifa = NULL;
1141 rcu_read_lock();
1143 for_each_netdev_rcu(net, dev) {
1144 struct inet6_dev *idev;
1146 /* Candidate Source Address (section 4)
1147 * - multicast and link-local destination address,
1148 * the set of candidate source address MUST only
1149 * include addresses assigned to interfaces
1150 * belonging to the same link as the outgoing
1151 * interface.
1152 * (- For site-local destination addresses, the
1153 * set of candidate source addresses MUST only
1154 * include addresses assigned to interfaces
1155 * belonging to the same site as the outgoing
1156 * interface.)
1158 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160 dst.ifindex && dev->ifindex != dst.ifindex)
1161 continue;
1163 idev = __in6_dev_get(dev);
1164 if (!idev)
1165 continue;
1167 read_lock_bh(&idev->lock);
1168 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1169 int i;
1172 * - Tentative Address (RFC2462 section 5.4)
1173 * - A tentative address is not considered
1174 * "assigned to an interface" in the traditional
1175 * sense, unless it is also flagged as optimistic.
1176 * - Candidate Source Address (section 4)
1177 * - In any case, anycast addresses, multicast
1178 * addresses, and the unspecified address MUST
1179 * NOT be included in a candidate set.
1181 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183 continue;
1185 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1187 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188 score->addr_type & IPV6_ADDR_MULTICAST)) {
1189 LIMIT_NETDEBUG(KERN_DEBUG
1190 "ADDRCONF: unspecified / multicast address "
1191 "assigned as unicast address on %s",
1192 dev->name);
1193 continue;
1196 score->rule = -1;
1197 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1199 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200 int minihiscore, miniscore;
1202 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1205 if (minihiscore > miniscore) {
1206 if (i == IPV6_SADDR_RULE_SCOPE &&
1207 score->scopedist > 0) {
1209 * special case:
1210 * each remaining entry
1211 * has too small (not enough)
1212 * scope, because ifa entries
1213 * are sorted by their scope
1214 * values.
1216 goto try_nextdev;
1218 break;
1219 } else if (minihiscore < miniscore) {
1220 if (hiscore->ifa)
1221 in6_ifa_put(hiscore->ifa);
1223 in6_ifa_hold(score->ifa);
1225 swap(hiscore, score);
1227 /* restore our iterator */
1228 score->ifa = hiscore->ifa;
1230 break;
1234 try_nextdev:
1235 read_unlock_bh(&idev->lock);
1237 rcu_read_unlock();
1239 if (!hiscore->ifa)
1240 return -EADDRNOTAVAIL;
1242 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243 in6_ifa_put(hiscore->ifa);
1244 return 0;
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250 unsigned char banned_flags)
1252 struct inet6_dev *idev;
1253 int err = -EADDRNOTAVAIL;
1255 rcu_read_lock();
1256 if ((idev = __in6_dev_get(dev)) != NULL) {
1257 struct inet6_ifaddr *ifp;
1259 read_lock_bh(&idev->lock);
1260 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1261 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1262 ipv6_addr_copy(addr, &ifp->addr);
1263 err = 0;
1264 break;
1267 read_unlock_bh(&idev->lock);
1269 rcu_read_unlock();
1270 return err;
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1275 int cnt = 0;
1276 struct inet6_ifaddr *ifp;
1278 read_lock_bh(&idev->lock);
1279 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1280 cnt++;
1281 read_unlock_bh(&idev->lock);
1282 return cnt;
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286 struct net_device *dev, int strict)
1288 struct inet6_ifaddr * ifp;
1289 u8 hash = ipv6_addr_hash(addr);
1291 read_lock_bh(&addrconf_hash_lock);
1292 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1293 if (!net_eq(dev_net(ifp->idev->dev), net))
1294 continue;
1295 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296 !(ifp->flags&IFA_F_TENTATIVE)) {
1297 if (dev == NULL || ifp->idev->dev == dev ||
1298 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1299 break;
1302 read_unlock_bh(&addrconf_hash_lock);
1303 return ifp != NULL;
1305 EXPORT_SYMBOL(ipv6_chk_addr);
1307 static
1308 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309 struct net_device *dev)
1311 struct inet6_ifaddr * ifp;
1312 u8 hash = ipv6_addr_hash(addr);
1314 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1315 if (!net_eq(dev_net(ifp->idev->dev), net))
1316 continue;
1317 if (ipv6_addr_equal(&ifp->addr, addr)) {
1318 if (dev == NULL || ifp->idev->dev == dev)
1319 break;
1322 return ifp != NULL;
1325 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1327 struct inet6_dev *idev;
1328 struct inet6_ifaddr *ifa;
1329 int onlink;
1331 onlink = 0;
1332 rcu_read_lock();
1333 idev = __in6_dev_get(dev);
1334 if (idev) {
1335 read_lock_bh(&idev->lock);
1336 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1337 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1338 ifa->prefix_len);
1339 if (onlink)
1340 break;
1342 read_unlock_bh(&idev->lock);
1344 rcu_read_unlock();
1345 return onlink;
1348 EXPORT_SYMBOL(ipv6_chk_prefix);
1350 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1351 struct net_device *dev, int strict)
1353 struct inet6_ifaddr * ifp;
1354 u8 hash = ipv6_addr_hash(addr);
1356 read_lock_bh(&addrconf_hash_lock);
1357 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1358 if (!net_eq(dev_net(ifp->idev->dev), net))
1359 continue;
1360 if (ipv6_addr_equal(&ifp->addr, addr)) {
1361 if (dev == NULL || ifp->idev->dev == dev ||
1362 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1363 in6_ifa_hold(ifp);
1364 break;
1368 read_unlock_bh(&addrconf_hash_lock);
1370 return ifp;
1373 /* Gets referenced address, destroys ifaddr */
1375 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1377 if (ifp->flags&IFA_F_PERMANENT) {
1378 spin_lock_bh(&ifp->lock);
1379 addrconf_del_timer(ifp);
1380 ifp->flags |= IFA_F_TENTATIVE;
1381 if (dad_failed)
1382 ifp->flags |= IFA_F_DADFAILED;
1383 spin_unlock_bh(&ifp->lock);
1384 if (dad_failed)
1385 ipv6_ifa_notify(0, ifp);
1386 in6_ifa_put(ifp);
1387 #ifdef CONFIG_IPV6_PRIVACY
1388 } else if (ifp->flags&IFA_F_TEMPORARY) {
1389 struct inet6_ifaddr *ifpub;
1390 spin_lock_bh(&ifp->lock);
1391 ifpub = ifp->ifpub;
1392 if (ifpub) {
1393 in6_ifa_hold(ifpub);
1394 spin_unlock_bh(&ifp->lock);
1395 ipv6_create_tempaddr(ifpub, ifp);
1396 in6_ifa_put(ifpub);
1397 } else {
1398 spin_unlock_bh(&ifp->lock);
1400 ipv6_del_addr(ifp);
1401 #endif
1402 } else
1403 ipv6_del_addr(ifp);
1406 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1408 struct inet6_dev *idev = ifp->idev;
1410 if (net_ratelimit())
1411 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1412 ifp->idev->dev->name, &ifp->addr);
1414 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1415 struct in6_addr addr;
1417 addr.s6_addr32[0] = htonl(0xfe800000);
1418 addr.s6_addr32[1] = 0;
1420 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1421 ipv6_addr_equal(&ifp->addr, &addr)) {
1422 /* DAD failed for link-local based on MAC address */
1423 idev->cnf.disable_ipv6 = 1;
1425 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1426 ifp->idev->dev->name);
1430 addrconf_dad_stop(ifp, 1);
1433 /* Join to solicited addr multicast group. */
1435 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1437 struct in6_addr maddr;
1439 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1440 return;
1442 addrconf_addr_solict_mult(addr, &maddr);
1443 ipv6_dev_mc_inc(dev, &maddr);
1446 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1448 struct in6_addr maddr;
1450 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1451 return;
1453 addrconf_addr_solict_mult(addr, &maddr);
1454 __ipv6_dev_mc_dec(idev, &maddr);
1457 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1459 struct in6_addr addr;
1460 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1461 if (ipv6_addr_any(&addr))
1462 return;
1463 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1466 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1468 struct in6_addr addr;
1469 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1470 if (ipv6_addr_any(&addr))
1471 return;
1472 __ipv6_dev_ac_dec(ifp->idev, &addr);
1475 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1477 if (dev->addr_len != ETH_ALEN)
1478 return -1;
1479 memcpy(eui, dev->dev_addr, 3);
1480 memcpy(eui + 5, dev->dev_addr + 3, 3);
1483 * The zSeries OSA network cards can be shared among various
1484 * OS instances, but the OSA cards have only one MAC address.
1485 * This leads to duplicate address conflicts in conjunction
1486 * with IPv6 if more than one instance uses the same card.
1488 * The driver for these cards can deliver a unique 16-bit
1489 * identifier for each instance sharing the same card. It is
1490 * placed instead of 0xFFFE in the interface identifier. The
1491 * "u" bit of the interface identifier is not inverted in this
1492 * case. Hence the resulting interface identifier has local
1493 * scope according to RFC2373.
1495 if (dev->dev_id) {
1496 eui[3] = (dev->dev_id >> 8) & 0xFF;
1497 eui[4] = dev->dev_id & 0xFF;
1498 } else {
1499 eui[3] = 0xFF;
1500 eui[4] = 0xFE;
1501 eui[0] ^= 2;
1503 return 0;
1506 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1508 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1509 if (dev->addr_len != ARCNET_ALEN)
1510 return -1;
1511 memset(eui, 0, 7);
1512 eui[7] = *(u8*)dev->dev_addr;
1513 return 0;
1516 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1518 if (dev->addr_len != INFINIBAND_ALEN)
1519 return -1;
1520 memcpy(eui, dev->dev_addr + 12, 8);
1521 eui[0] |= 2;
1522 return 0;
1525 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1527 if (addr == 0)
1528 return -1;
1529 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1530 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1531 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1532 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1533 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1534 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1535 eui[1] = 0;
1536 eui[2] = 0x5E;
1537 eui[3] = 0xFE;
1538 memcpy(eui + 4, &addr, 4);
1539 return 0;
1541 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1543 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1545 if (dev->priv_flags & IFF_ISATAP)
1546 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1547 return -1;
1550 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1552 switch (dev->type) {
1553 case ARPHRD_ETHER:
1554 case ARPHRD_FDDI:
1555 case ARPHRD_IEEE802_TR:
1556 return addrconf_ifid_eui48(eui, dev);
1557 case ARPHRD_ARCNET:
1558 return addrconf_ifid_arcnet(eui, dev);
1559 case ARPHRD_INFINIBAND:
1560 return addrconf_ifid_infiniband(eui, dev);
1561 case ARPHRD_SIT:
1562 return addrconf_ifid_sit(eui, dev);
1564 return -1;
1567 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1569 int err = -1;
1570 struct inet6_ifaddr *ifp;
1572 read_lock_bh(&idev->lock);
1573 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1574 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1575 memcpy(eui, ifp->addr.s6_addr+8, 8);
1576 err = 0;
1577 break;
1580 read_unlock_bh(&idev->lock);
1581 return err;
1584 #ifdef CONFIG_IPV6_PRIVACY
1585 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1586 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1588 regen:
1589 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1590 idev->rndid[0] &= ~0x02;
1593 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1594 * check if generated address is not inappropriate
1596 * - Reserved subnet anycast (RFC 2526)
1597 * 11111101 11....11 1xxxxxxx
1598 * - ISATAP (RFC4214) 6.1
1599 * 00-00-5E-FE-xx-xx-xx-xx
1600 * - value 0
1601 * - XXX: already assigned to an address on the device
1603 if (idev->rndid[0] == 0xfd &&
1604 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1605 (idev->rndid[7]&0x80))
1606 goto regen;
1607 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1608 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1609 goto regen;
1610 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1611 goto regen;
1614 return 0;
1617 static void ipv6_regen_rndid(unsigned long data)
1619 struct inet6_dev *idev = (struct inet6_dev *) data;
1620 unsigned long expires;
1622 rcu_read_lock_bh();
1623 write_lock_bh(&idev->lock);
1625 if (idev->dead)
1626 goto out;
1628 if (__ipv6_regen_rndid(idev) < 0)
1629 goto out;
1631 expires = jiffies +
1632 idev->cnf.temp_prefered_lft * HZ -
1633 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1634 if (time_before(expires, jiffies)) {
1635 printk(KERN_WARNING
1636 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1637 idev->dev->name);
1638 goto out;
1641 if (!mod_timer(&idev->regen_timer, expires))
1642 in6_dev_hold(idev);
1644 out:
1645 write_unlock_bh(&idev->lock);
1646 rcu_read_unlock_bh();
1647 in6_dev_put(idev);
1650 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1651 int ret = 0;
1653 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1654 ret = __ipv6_regen_rndid(idev);
1655 return ret;
1657 #endif
1660 * Add prefix route.
1663 static void
1664 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1665 unsigned long expires, u32 flags)
1667 struct fib6_config cfg = {
1668 .fc_table = RT6_TABLE_PREFIX,
1669 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1670 .fc_ifindex = dev->ifindex,
1671 .fc_expires = expires,
1672 .fc_dst_len = plen,
1673 .fc_flags = RTF_UP | flags,
1674 .fc_nlinfo.nl_net = dev_net(dev),
1675 .fc_protocol = RTPROT_KERNEL,
1678 ipv6_addr_copy(&cfg.fc_dst, pfx);
1680 /* Prevent useless cloning on PtP SIT.
1681 This thing is done here expecting that the whole
1682 class of non-broadcast devices need not cloning.
1684 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1685 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1686 cfg.fc_flags |= RTF_NONEXTHOP;
1687 #endif
1689 ip6_route_add(&cfg);
1692 /* Create "default" multicast route to the interface */
1694 static void addrconf_add_mroute(struct net_device *dev)
1696 struct fib6_config cfg = {
1697 .fc_table = RT6_TABLE_LOCAL,
1698 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1699 .fc_ifindex = dev->ifindex,
1700 .fc_dst_len = 8,
1701 .fc_flags = RTF_UP,
1702 .fc_nlinfo.nl_net = dev_net(dev),
1705 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1707 ip6_route_add(&cfg);
1710 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1711 static void sit_route_add(struct net_device *dev)
1713 struct fib6_config cfg = {
1714 .fc_table = RT6_TABLE_MAIN,
1715 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1716 .fc_ifindex = dev->ifindex,
1717 .fc_dst_len = 96,
1718 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1719 .fc_nlinfo.nl_net = dev_net(dev),
1722 /* prefix length - 96 bits "::d.d.d.d" */
1723 ip6_route_add(&cfg);
1725 #endif
1727 static void addrconf_add_lroute(struct net_device *dev)
1729 struct in6_addr addr;
1731 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1732 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1735 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1737 struct inet6_dev *idev;
1739 ASSERT_RTNL();
1741 if ((idev = ipv6_find_idev(dev)) == NULL)
1742 return NULL;
1744 /* Add default multicast route */
1745 addrconf_add_mroute(dev);
1747 /* Add link local route */
1748 addrconf_add_lroute(dev);
1749 return idev;
1752 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1754 struct prefix_info *pinfo;
1755 __u32 valid_lft;
1756 __u32 prefered_lft;
1757 int addr_type;
1758 struct inet6_dev *in6_dev;
1759 struct net *net = dev_net(dev);
1761 pinfo = (struct prefix_info *) opt;
1763 if (len < sizeof(struct prefix_info)) {
1764 ADBG(("addrconf: prefix option too short\n"));
1765 return;
1769 * Validation checks ([ADDRCONF], page 19)
1772 addr_type = ipv6_addr_type(&pinfo->prefix);
1774 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1775 return;
1777 valid_lft = ntohl(pinfo->valid);
1778 prefered_lft = ntohl(pinfo->prefered);
1780 if (prefered_lft > valid_lft) {
1781 if (net_ratelimit())
1782 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1783 return;
1786 in6_dev = in6_dev_get(dev);
1788 if (in6_dev == NULL) {
1789 if (net_ratelimit())
1790 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1791 return;
1795 * Two things going on here:
1796 * 1) Add routes for on-link prefixes
1797 * 2) Configure prefixes with the auto flag set
1800 if (pinfo->onlink) {
1801 struct rt6_info *rt;
1802 unsigned long rt_expires;
1804 /* Avoid arithmetic overflow. Really, we could
1805 * save rt_expires in seconds, likely valid_lft,
1806 * but it would require division in fib gc, that it
1807 * not good.
1809 if (HZ > USER_HZ)
1810 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1811 else
1812 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1814 if (addrconf_finite_timeout(rt_expires))
1815 rt_expires *= HZ;
1817 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1818 dev->ifindex, 1);
1820 if (rt && addrconf_is_prefix_route(rt)) {
1821 /* Autoconf prefix route */
1822 if (valid_lft == 0) {
1823 ip6_del_rt(rt);
1824 rt = NULL;
1825 } else if (addrconf_finite_timeout(rt_expires)) {
1826 /* not infinity */
1827 rt->rt6i_expires = jiffies + rt_expires;
1828 rt->rt6i_flags |= RTF_EXPIRES;
1829 } else {
1830 rt->rt6i_flags &= ~RTF_EXPIRES;
1831 rt->rt6i_expires = 0;
1833 } else if (valid_lft) {
1834 clock_t expires = 0;
1835 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1836 if (addrconf_finite_timeout(rt_expires)) {
1837 /* not infinity */
1838 flags |= RTF_EXPIRES;
1839 expires = jiffies_to_clock_t(rt_expires);
1841 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1842 dev, expires, flags);
1844 if (rt)
1845 dst_release(&rt->u.dst);
1848 /* Try to figure out our local address for this prefix */
1850 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1851 struct inet6_ifaddr * ifp;
1852 struct in6_addr addr;
1853 int create = 0, update_lft = 0;
1855 if (pinfo->prefix_len == 64) {
1856 memcpy(&addr, &pinfo->prefix, 8);
1857 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1858 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1859 in6_dev_put(in6_dev);
1860 return;
1862 goto ok;
1864 if (net_ratelimit())
1865 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1866 pinfo->prefix_len);
1867 in6_dev_put(in6_dev);
1868 return;
1872 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1874 if (ifp == NULL && valid_lft) {
1875 int max_addresses = in6_dev->cnf.max_addresses;
1876 u32 addr_flags = 0;
1878 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1879 if (in6_dev->cnf.optimistic_dad &&
1880 !net->ipv6.devconf_all->forwarding)
1881 addr_flags = IFA_F_OPTIMISTIC;
1882 #endif
1884 /* Do not allow to create too much of autoconfigured
1885 * addresses; this would be too easy way to crash kernel.
1887 if (!max_addresses ||
1888 ipv6_count_addresses(in6_dev) < max_addresses)
1889 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1890 addr_type&IPV6_ADDR_SCOPE_MASK,
1891 addr_flags);
1893 if (!ifp || IS_ERR(ifp)) {
1894 in6_dev_put(in6_dev);
1895 return;
1898 update_lft = create = 1;
1899 ifp->cstamp = jiffies;
1900 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1903 if (ifp) {
1904 int flags;
1905 unsigned long now;
1906 #ifdef CONFIG_IPV6_PRIVACY
1907 struct inet6_ifaddr *ift;
1908 #endif
1909 u32 stored_lft;
1911 /* update lifetime (RFC2462 5.5.3 e) */
1912 spin_lock(&ifp->lock);
1913 now = jiffies;
1914 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1915 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1916 else
1917 stored_lft = 0;
1918 if (!update_lft && stored_lft) {
1919 if (valid_lft > MIN_VALID_LIFETIME ||
1920 valid_lft > stored_lft)
1921 update_lft = 1;
1922 else if (stored_lft <= MIN_VALID_LIFETIME) {
1923 /* valid_lft <= stored_lft is always true */
1925 * RFC 4862 Section 5.5.3e:
1926 * "Note that the preferred lifetime of
1927 * the corresponding address is always
1928 * reset to the Preferred Lifetime in
1929 * the received Prefix Information
1930 * option, regardless of whether the
1931 * valid lifetime is also reset or
1932 * ignored."
1934 * So if the preferred lifetime in
1935 * this advertisement is different
1936 * than what we have stored, but the
1937 * valid lifetime is invalid, just
1938 * reset prefered_lft.
1940 * We must set the valid lifetime
1941 * to the stored lifetime since we'll
1942 * be updating the timestamp below,
1943 * else we'll set it back to the
1944 * minumum.
1946 if (prefered_lft != ifp->prefered_lft) {
1947 valid_lft = stored_lft;
1948 update_lft = 1;
1950 } else {
1951 valid_lft = MIN_VALID_LIFETIME;
1952 if (valid_lft < prefered_lft)
1953 prefered_lft = valid_lft;
1954 update_lft = 1;
1958 if (update_lft) {
1959 ifp->valid_lft = valid_lft;
1960 ifp->prefered_lft = prefered_lft;
1961 ifp->tstamp = now;
1962 flags = ifp->flags;
1963 ifp->flags &= ~IFA_F_DEPRECATED;
1964 spin_unlock(&ifp->lock);
1966 if (!(flags&IFA_F_TENTATIVE))
1967 ipv6_ifa_notify(0, ifp);
1968 } else
1969 spin_unlock(&ifp->lock);
1971 #ifdef CONFIG_IPV6_PRIVACY
1972 read_lock_bh(&in6_dev->lock);
1973 /* update all temporary addresses in the list */
1974 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1976 * When adjusting the lifetimes of an existing
1977 * temporary address, only lower the lifetimes.
1978 * Implementations must not increase the
1979 * lifetimes of an existing temporary address
1980 * when processing a Prefix Information Option.
1982 if (ifp != ift->ifpub)
1983 continue;
1985 spin_lock(&ift->lock);
1986 flags = ift->flags;
1987 if (ift->valid_lft > valid_lft &&
1988 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1989 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1990 if (ift->prefered_lft > prefered_lft &&
1991 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1992 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1993 spin_unlock(&ift->lock);
1994 if (!(flags&IFA_F_TENTATIVE))
1995 ipv6_ifa_notify(0, ift);
1998 if (create && in6_dev->cnf.use_tempaddr > 0) {
2000 * When a new public address is created as described in [ADDRCONF],
2001 * also create a new temporary address.
2003 read_unlock_bh(&in6_dev->lock);
2004 ipv6_create_tempaddr(ifp, NULL);
2005 } else {
2006 read_unlock_bh(&in6_dev->lock);
2008 #endif
2009 in6_ifa_put(ifp);
2010 addrconf_verify(0);
2013 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2014 in6_dev_put(in6_dev);
2018 * Set destination address.
2019 * Special case for SIT interfaces where we create a new "virtual"
2020 * device.
2022 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2024 struct in6_ifreq ireq;
2025 struct net_device *dev;
2026 int err = -EINVAL;
2028 rtnl_lock();
2030 err = -EFAULT;
2031 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2032 goto err_exit;
2034 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2036 err = -ENODEV;
2037 if (dev == NULL)
2038 goto err_exit;
2040 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2041 if (dev->type == ARPHRD_SIT) {
2042 const struct net_device_ops *ops = dev->netdev_ops;
2043 struct ifreq ifr;
2044 struct ip_tunnel_parm p;
2046 err = -EADDRNOTAVAIL;
2047 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2048 goto err_exit;
2050 memset(&p, 0, sizeof(p));
2051 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2052 p.iph.saddr = 0;
2053 p.iph.version = 4;
2054 p.iph.ihl = 5;
2055 p.iph.protocol = IPPROTO_IPV6;
2056 p.iph.ttl = 64;
2057 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2059 if (ops->ndo_do_ioctl) {
2060 mm_segment_t oldfs = get_fs();
2062 set_fs(KERNEL_DS);
2063 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2064 set_fs(oldfs);
2065 } else
2066 err = -EOPNOTSUPP;
2068 if (err == 0) {
2069 err = -ENOBUFS;
2070 dev = __dev_get_by_name(net, p.name);
2071 if (!dev)
2072 goto err_exit;
2073 err = dev_open(dev);
2076 #endif
2078 err_exit:
2079 rtnl_unlock();
2080 return err;
2084 * Manual configuration of address on an interface
2086 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2087 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2088 __u32 valid_lft)
2090 struct inet6_ifaddr *ifp;
2091 struct inet6_dev *idev;
2092 struct net_device *dev;
2093 int scope;
2094 u32 flags;
2095 clock_t expires;
2096 unsigned long timeout;
2098 ASSERT_RTNL();
2100 if (plen > 128)
2101 return -EINVAL;
2103 /* check the lifetime */
2104 if (!valid_lft || prefered_lft > valid_lft)
2105 return -EINVAL;
2107 dev = __dev_get_by_index(net, ifindex);
2108 if (!dev)
2109 return -ENODEV;
2111 if ((idev = addrconf_add_dev(dev)) == NULL)
2112 return -ENOBUFS;
2114 scope = ipv6_addr_scope(pfx);
2116 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2117 if (addrconf_finite_timeout(timeout)) {
2118 expires = jiffies_to_clock_t(timeout * HZ);
2119 valid_lft = timeout;
2120 flags = RTF_EXPIRES;
2121 } else {
2122 expires = 0;
2123 flags = 0;
2124 ifa_flags |= IFA_F_PERMANENT;
2127 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2128 if (addrconf_finite_timeout(timeout)) {
2129 if (timeout == 0)
2130 ifa_flags |= IFA_F_DEPRECATED;
2131 prefered_lft = timeout;
2134 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2136 if (!IS_ERR(ifp)) {
2137 spin_lock_bh(&ifp->lock);
2138 ifp->valid_lft = valid_lft;
2139 ifp->prefered_lft = prefered_lft;
2140 ifp->tstamp = jiffies;
2141 spin_unlock_bh(&ifp->lock);
2143 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2144 expires, flags);
2146 * Note that section 3.1 of RFC 4429 indicates
2147 * that the Optimistic flag should not be set for
2148 * manually configured addresses
2150 addrconf_dad_start(ifp, 0);
2151 in6_ifa_put(ifp);
2152 addrconf_verify(0);
2153 return 0;
2156 return PTR_ERR(ifp);
2159 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2160 unsigned int plen)
2162 struct inet6_ifaddr *ifp;
2163 struct inet6_dev *idev;
2164 struct net_device *dev;
2166 if (plen > 128)
2167 return -EINVAL;
2169 dev = __dev_get_by_index(net, ifindex);
2170 if (!dev)
2171 return -ENODEV;
2173 if ((idev = __in6_dev_get(dev)) == NULL)
2174 return -ENXIO;
2176 read_lock_bh(&idev->lock);
2177 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2178 if (ifp->prefix_len == plen &&
2179 ipv6_addr_equal(pfx, &ifp->addr)) {
2180 in6_ifa_hold(ifp);
2181 read_unlock_bh(&idev->lock);
2183 ipv6_del_addr(ifp);
2185 /* If the last address is deleted administratively,
2186 disable IPv6 on this interface.
2188 if (idev->addr_list == NULL)
2189 addrconf_ifdown(idev->dev, 1);
2190 return 0;
2193 read_unlock_bh(&idev->lock);
2194 return -EADDRNOTAVAIL;
2198 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2200 struct in6_ifreq ireq;
2201 int err;
2203 if (!capable(CAP_NET_ADMIN))
2204 return -EPERM;
2206 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2207 return -EFAULT;
2209 rtnl_lock();
2210 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2211 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2212 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2213 rtnl_unlock();
2214 return err;
2217 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2219 struct in6_ifreq ireq;
2220 int err;
2222 if (!capable(CAP_NET_ADMIN))
2223 return -EPERM;
2225 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2226 return -EFAULT;
2228 rtnl_lock();
2229 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2230 ireq.ifr6_prefixlen);
2231 rtnl_unlock();
2232 return err;
2235 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2236 int plen, int scope)
2238 struct inet6_ifaddr *ifp;
2240 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2241 if (!IS_ERR(ifp)) {
2242 spin_lock_bh(&ifp->lock);
2243 ifp->flags &= ~IFA_F_TENTATIVE;
2244 spin_unlock_bh(&ifp->lock);
2245 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2246 in6_ifa_put(ifp);
2250 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2251 static void sit_add_v4_addrs(struct inet6_dev *idev)
2253 struct in6_addr addr;
2254 struct net_device *dev;
2255 struct net *net = dev_net(idev->dev);
2256 int scope;
2258 ASSERT_RTNL();
2260 memset(&addr, 0, sizeof(struct in6_addr));
2261 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2263 if (idev->dev->flags&IFF_POINTOPOINT) {
2264 addr.s6_addr32[0] = htonl(0xfe800000);
2265 scope = IFA_LINK;
2266 } else {
2267 scope = IPV6_ADDR_COMPATv4;
2270 if (addr.s6_addr32[3]) {
2271 add_addr(idev, &addr, 128, scope);
2272 return;
2275 for_each_netdev(net, dev) {
2276 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2277 if (in_dev && (dev->flags & IFF_UP)) {
2278 struct in_ifaddr * ifa;
2280 int flag = scope;
2282 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2283 int plen;
2285 addr.s6_addr32[3] = ifa->ifa_local;
2287 if (ifa->ifa_scope == RT_SCOPE_LINK)
2288 continue;
2289 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2290 if (idev->dev->flags&IFF_POINTOPOINT)
2291 continue;
2292 flag |= IFA_HOST;
2294 if (idev->dev->flags&IFF_POINTOPOINT)
2295 plen = 64;
2296 else
2297 plen = 96;
2299 add_addr(idev, &addr, plen, flag);
2304 #endif
2306 static void init_loopback(struct net_device *dev)
2308 struct inet6_dev *idev;
2310 /* ::1 */
2312 ASSERT_RTNL();
2314 if ((idev = ipv6_find_idev(dev)) == NULL) {
2315 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2316 return;
2319 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2322 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2324 struct inet6_ifaddr * ifp;
2325 u32 addr_flags = IFA_F_PERMANENT;
2327 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2328 if (idev->cnf.optimistic_dad &&
2329 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2330 addr_flags |= IFA_F_OPTIMISTIC;
2331 #endif
2334 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2335 if (!IS_ERR(ifp)) {
2336 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2337 addrconf_dad_start(ifp, 0);
2338 in6_ifa_put(ifp);
2342 static void addrconf_dev_config(struct net_device *dev)
2344 struct in6_addr addr;
2345 struct inet6_dev * idev;
2347 ASSERT_RTNL();
2349 if ((dev->type != ARPHRD_ETHER) &&
2350 (dev->type != ARPHRD_FDDI) &&
2351 (dev->type != ARPHRD_IEEE802_TR) &&
2352 (dev->type != ARPHRD_ARCNET) &&
2353 (dev->type != ARPHRD_INFINIBAND)) {
2354 /* Alas, we support only Ethernet autoconfiguration. */
2355 return;
2358 idev = addrconf_add_dev(dev);
2359 if (idev == NULL)
2360 return;
2362 memset(&addr, 0, sizeof(struct in6_addr));
2363 addr.s6_addr32[0] = htonl(0xFE800000);
2365 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2366 addrconf_add_linklocal(idev, &addr);
2369 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2370 static void addrconf_sit_config(struct net_device *dev)
2372 struct inet6_dev *idev;
2374 ASSERT_RTNL();
2377 * Configure the tunnel with one of our IPv4
2378 * addresses... we should configure all of
2379 * our v4 addrs in the tunnel
2382 if ((idev = ipv6_find_idev(dev)) == NULL) {
2383 printk(KERN_DEBUG "init sit: add_dev failed\n");
2384 return;
2387 if (dev->priv_flags & IFF_ISATAP) {
2388 struct in6_addr addr;
2390 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2391 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2392 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2393 addrconf_add_linklocal(idev, &addr);
2394 return;
2397 sit_add_v4_addrs(idev);
2399 if (dev->flags&IFF_POINTOPOINT) {
2400 addrconf_add_mroute(dev);
2401 addrconf_add_lroute(dev);
2402 } else
2403 sit_route_add(dev);
2405 #endif
2407 static inline int
2408 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2410 struct in6_addr lladdr;
2412 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2413 addrconf_add_linklocal(idev, &lladdr);
2414 return 0;
2416 return -1;
2419 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2421 struct net_device *link_dev;
2422 struct net *net = dev_net(idev->dev);
2424 /* first try to inherit the link-local address from the link device */
2425 if (idev->dev->iflink &&
2426 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2427 if (!ipv6_inherit_linklocal(idev, link_dev))
2428 return;
2430 /* then try to inherit it from any device */
2431 for_each_netdev(net, link_dev) {
2432 if (!ipv6_inherit_linklocal(idev, link_dev))
2433 return;
2435 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2439 * Autoconfigure tunnel with a link-local address so routing protocols,
2440 * DHCPv6, MLD etc. can be run over the virtual link
2443 static void addrconf_ip6_tnl_config(struct net_device *dev)
2445 struct inet6_dev *idev;
2447 ASSERT_RTNL();
2449 if ((idev = addrconf_add_dev(dev)) == NULL) {
2450 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2451 return;
2453 ip6_tnl_add_linklocal(idev);
2456 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2457 void * data)
2459 struct net_device *dev = (struct net_device *) data;
2460 struct inet6_dev *idev = __in6_dev_get(dev);
2461 int run_pending = 0;
2462 int err;
2464 switch(event) {
2465 case NETDEV_REGISTER:
2466 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2467 idev = ipv6_add_dev(dev);
2468 if (!idev)
2469 return notifier_from_errno(-ENOMEM);
2471 break;
2472 case NETDEV_UP:
2473 case NETDEV_CHANGE:
2474 if (dev->flags & IFF_SLAVE)
2475 break;
2477 if (event == NETDEV_UP) {
2478 if (!addrconf_qdisc_ok(dev)) {
2479 /* device is not ready yet. */
2480 printk(KERN_INFO
2481 "ADDRCONF(NETDEV_UP): %s: "
2482 "link is not ready\n",
2483 dev->name);
2484 break;
2487 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2488 idev = ipv6_add_dev(dev);
2490 if (idev) {
2491 idev->if_flags |= IF_READY;
2492 run_pending = 1;
2494 } else {
2495 if (!addrconf_qdisc_ok(dev)) {
2496 /* device is still not ready. */
2497 break;
2500 if (idev) {
2501 if (idev->if_flags & IF_READY) {
2502 /* device is already configured. */
2503 break;
2505 idev->if_flags |= IF_READY;
2508 printk(KERN_INFO
2509 "ADDRCONF(NETDEV_CHANGE): %s: "
2510 "link becomes ready\n",
2511 dev->name);
2513 run_pending = 1;
2516 switch(dev->type) {
2517 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2518 case ARPHRD_SIT:
2519 addrconf_sit_config(dev);
2520 break;
2521 #endif
2522 case ARPHRD_TUNNEL6:
2523 addrconf_ip6_tnl_config(dev);
2524 break;
2525 case ARPHRD_LOOPBACK:
2526 init_loopback(dev);
2527 break;
2529 default:
2530 addrconf_dev_config(dev);
2531 break;
2533 if (idev) {
2534 if (run_pending)
2535 addrconf_dad_run(idev);
2537 /* If the MTU changed during the interface down, when the
2538 interface up, the changed MTU must be reflected in the
2539 idev as well as routers.
2541 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2542 rt6_mtu_change(dev, dev->mtu);
2543 idev->cnf.mtu6 = dev->mtu;
2545 idev->tstamp = jiffies;
2546 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2547 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2548 stop IPv6 on this interface.
2550 if (dev->mtu < IPV6_MIN_MTU)
2551 addrconf_ifdown(dev, event != NETDEV_DOWN);
2553 break;
2555 case NETDEV_CHANGEMTU:
2556 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2557 rt6_mtu_change(dev, dev->mtu);
2558 idev->cnf.mtu6 = dev->mtu;
2559 break;
2562 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2563 idev = ipv6_add_dev(dev);
2564 if (idev)
2565 break;
2568 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2570 case NETDEV_DOWN:
2571 case NETDEV_UNREGISTER:
2573 * Remove all addresses from this interface.
2575 addrconf_ifdown(dev, event != NETDEV_DOWN);
2576 break;
2578 case NETDEV_CHANGENAME:
2579 if (idev) {
2580 snmp6_unregister_dev(idev);
2581 addrconf_sysctl_unregister(idev);
2582 addrconf_sysctl_register(idev);
2583 err = snmp6_register_dev(idev);
2584 if (err)
2585 return notifier_from_errno(err);
2587 break;
2588 case NETDEV_BONDING_OLDTYPE:
2589 case NETDEV_BONDING_NEWTYPE:
2590 addrconf_bonding_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,
2602 .priority = 0
2605 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2607 struct inet6_dev *idev;
2608 ASSERT_RTNL();
2610 idev = __in6_dev_get(dev);
2612 if (event == NETDEV_BONDING_NEWTYPE)
2613 ipv6_mc_remap(idev);
2614 else if (event == NETDEV_BONDING_OLDTYPE)
2615 ipv6_mc_unmap(idev);
2618 static int addrconf_ifdown(struct net_device *dev, int how)
2620 struct inet6_dev *idev;
2621 struct inet6_ifaddr *ifa, *keep_list, **bifa;
2622 struct net *net = dev_net(dev);
2623 int i;
2625 ASSERT_RTNL();
2627 rt6_ifdown(net, dev);
2628 neigh_ifdown(&nd_tbl, dev);
2630 idev = __in6_dev_get(dev);
2631 if (idev == NULL)
2632 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 /* Step 2: clear hash table */
2649 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2650 bifa = &inet6_addr_lst[i];
2652 write_lock_bh(&addrconf_hash_lock);
2653 while ((ifa = *bifa) != NULL) {
2654 if (ifa->idev == idev &&
2655 (how || !(ifa->flags&IFA_F_PERMANENT) ||
2656 ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2657 *bifa = ifa->lst_next;
2658 ifa->lst_next = NULL;
2659 __in6_ifa_put(ifa);
2660 continue;
2662 bifa = &ifa->lst_next;
2664 write_unlock_bh(&addrconf_hash_lock);
2667 write_lock_bh(&idev->lock);
2669 /* Step 3: clear flags for stateless addrconf */
2670 if (!how)
2671 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2673 /* Step 4: clear address list */
2674 #ifdef CONFIG_IPV6_PRIVACY
2675 if (how && del_timer(&idev->regen_timer))
2676 in6_dev_put(idev);
2678 /* clear tempaddr list */
2679 while ((ifa = idev->tempaddr_list) != NULL) {
2680 idev->tempaddr_list = ifa->tmp_next;
2681 ifa->tmp_next = NULL;
2682 ifa->dead = 1;
2683 write_unlock_bh(&idev->lock);
2684 spin_lock_bh(&ifa->lock);
2686 if (ifa->ifpub) {
2687 in6_ifa_put(ifa->ifpub);
2688 ifa->ifpub = NULL;
2690 spin_unlock_bh(&ifa->lock);
2691 in6_ifa_put(ifa);
2692 write_lock_bh(&idev->lock);
2694 #endif
2695 keep_list = NULL;
2696 bifa = &keep_list;
2697 while ((ifa = idev->addr_list) != NULL) {
2698 idev->addr_list = ifa->if_next;
2699 ifa->if_next = NULL;
2701 addrconf_del_timer(ifa);
2703 /* If just doing link down, and address is permanent
2704 and not link-local, then retain it. */
2705 if (how == 0 &&
2706 (ifa->flags&IFA_F_PERMANENT) &&
2707 !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2709 /* Move to holding list */
2710 *bifa = ifa;
2711 bifa = &ifa->if_next;
2713 /* If not doing DAD on this address, just keep it. */
2714 if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2715 idev->cnf.accept_dad <= 0 ||
2716 (ifa->flags & IFA_F_NODAD))
2717 continue;
2719 /* If it was tentative already, no need to notify */
2720 if (ifa->flags & IFA_F_TENTATIVE)
2721 continue;
2723 /* Flag it for later restoration when link comes up */
2724 ifa->flags |= IFA_F_TENTATIVE;
2725 in6_ifa_hold(ifa);
2726 } else {
2727 ifa->dead = 1;
2729 write_unlock_bh(&idev->lock);
2731 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2732 if (ifa->dead)
2733 atomic_notifier_call_chain(&inet6addr_chain,
2734 NETDEV_DOWN, ifa);
2735 in6_ifa_put(ifa);
2737 write_lock_bh(&idev->lock);
2740 idev->addr_list = keep_list;
2742 write_unlock_bh(&idev->lock);
2744 /* Step 5: Discard multicast list */
2746 if (how)
2747 ipv6_mc_destroy_dev(idev);
2748 else
2749 ipv6_mc_down(idev);
2751 idev->tstamp = jiffies;
2753 /* Shot the device (if unregistered) */
2755 if (how) {
2756 addrconf_sysctl_unregister(idev);
2757 neigh_parms_release(&nd_tbl, idev->nd_parms);
2758 neigh_ifdown(&nd_tbl, dev);
2759 in6_dev_put(idev);
2761 return 0;
2764 static void addrconf_rs_timer(unsigned long data)
2766 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2767 struct inet6_dev *idev = ifp->idev;
2769 read_lock(&idev->lock);
2770 if (idev->dead || !(idev->if_flags & IF_READY))
2771 goto out;
2773 if (idev->cnf.forwarding)
2774 goto out;
2776 /* Announcement received after solicitation was sent */
2777 if (idev->if_flags & IF_RA_RCVD)
2778 goto out;
2780 spin_lock(&ifp->lock);
2781 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2782 /* The wait after the last probe can be shorter */
2783 addrconf_mod_timer(ifp, AC_RS,
2784 (ifp->probes == idev->cnf.rtr_solicits) ?
2785 idev->cnf.rtr_solicit_delay :
2786 idev->cnf.rtr_solicit_interval);
2787 spin_unlock(&ifp->lock);
2789 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2790 } else {
2791 spin_unlock(&ifp->lock);
2793 * Note: we do not support deprecated "all on-link"
2794 * assumption any longer.
2796 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2797 idev->dev->name);
2800 out:
2801 read_unlock(&idev->lock);
2802 in6_ifa_put(ifp);
2806 * Duplicate Address Detection
2808 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2810 unsigned long rand_num;
2811 struct inet6_dev *idev = ifp->idev;
2813 if (ifp->flags & IFA_F_OPTIMISTIC)
2814 rand_num = 0;
2815 else
2816 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2818 ifp->probes = idev->cnf.dad_transmits;
2819 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2822 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2824 struct inet6_dev *idev = ifp->idev;
2825 struct net_device *dev = idev->dev;
2827 addrconf_join_solict(dev, &ifp->addr);
2829 net_srandom(ifp->addr.s6_addr32[3]);
2831 read_lock_bh(&idev->lock);
2832 if (ifp->dead)
2833 goto out;
2835 spin_lock(&ifp->lock);
2836 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2837 idev->cnf.accept_dad < 1 ||
2838 !(ifp->flags&IFA_F_TENTATIVE) ||
2839 ifp->flags & IFA_F_NODAD) {
2840 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2841 spin_unlock(&ifp->lock);
2842 read_unlock_bh(&idev->lock);
2844 addrconf_dad_completed(ifp);
2845 return;
2848 if (!(idev->if_flags & IF_READY)) {
2849 spin_unlock(&ifp->lock);
2850 read_unlock_bh(&idev->lock);
2852 * If the device is not ready:
2853 * - keep it tentative if it is a permanent address.
2854 * - otherwise, kill it.
2856 in6_ifa_hold(ifp);
2857 addrconf_dad_stop(ifp, 0);
2858 return;
2862 * Optimistic nodes can start receiving
2863 * Frames right away
2865 if(ifp->flags & IFA_F_OPTIMISTIC)
2866 ip6_ins_rt(ifp->rt);
2868 addrconf_dad_kick(ifp);
2869 spin_unlock(&ifp->lock);
2870 out:
2871 read_unlock_bh(&idev->lock);
2874 static void addrconf_dad_timer(unsigned long data)
2876 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2877 struct inet6_dev *idev = ifp->idev;
2878 struct in6_addr mcaddr;
2880 read_lock(&idev->lock);
2881 if (idev->dead || !(idev->if_flags & IF_READY)) {
2882 read_unlock(&idev->lock);
2883 goto out;
2886 spin_lock(&ifp->lock);
2887 if (ifp->probes == 0) {
2889 * DAD was successful
2892 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2893 spin_unlock(&ifp->lock);
2894 read_unlock(&idev->lock);
2896 addrconf_dad_completed(ifp);
2898 goto out;
2901 ifp->probes--;
2902 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2903 spin_unlock(&ifp->lock);
2904 read_unlock(&idev->lock);
2906 /* send a neighbour solicitation for our addr */
2907 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2908 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2909 out:
2910 in6_ifa_put(ifp);
2913 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2915 struct net_device * dev = ifp->idev->dev;
2918 * Configure the address for reception. Now it is valid.
2921 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2923 /* If added prefix is link local and forwarding is off,
2924 start sending router solicitations.
2927 if (ifp->idev->cnf.forwarding == 0 &&
2928 ifp->idev->cnf.rtr_solicits > 0 &&
2929 (dev->flags&IFF_LOOPBACK) == 0 &&
2930 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2932 * If a host as already performed a random delay
2933 * [...] as part of DAD [...] there is no need
2934 * to delay again before sending the first RS
2936 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2938 spin_lock_bh(&ifp->lock);
2939 ifp->probes = 1;
2940 ifp->idev->if_flags |= IF_RS_SENT;
2941 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2942 spin_unlock_bh(&ifp->lock);
2946 static void addrconf_dad_run(struct inet6_dev *idev) {
2947 struct inet6_ifaddr *ifp;
2949 read_lock_bh(&idev->lock);
2950 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2951 spin_lock(&ifp->lock);
2952 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2953 spin_unlock(&ifp->lock);
2954 continue;
2956 spin_unlock(&ifp->lock);
2957 addrconf_dad_kick(ifp);
2959 read_unlock_bh(&idev->lock);
2962 #ifdef CONFIG_PROC_FS
2963 struct if6_iter_state {
2964 struct seq_net_private p;
2965 int bucket;
2968 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2970 struct inet6_ifaddr *ifa = NULL;
2971 struct if6_iter_state *state = seq->private;
2972 struct net *net = seq_file_net(seq);
2974 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2975 ifa = inet6_addr_lst[state->bucket];
2977 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2978 ifa = ifa->lst_next;
2979 if (ifa)
2980 break;
2982 return ifa;
2985 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2987 struct if6_iter_state *state = seq->private;
2988 struct net *net = seq_file_net(seq);
2990 ifa = ifa->lst_next;
2991 try_again:
2992 if (ifa) {
2993 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2994 ifa = ifa->lst_next;
2995 goto try_again;
2999 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
3000 ifa = inet6_addr_lst[state->bucket];
3001 goto try_again;
3004 return ifa;
3007 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3009 struct inet6_ifaddr *ifa = if6_get_first(seq);
3011 if (ifa)
3012 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3013 --pos;
3014 return pos ? NULL : ifa;
3017 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3018 __acquires(addrconf_hash_lock)
3020 read_lock_bh(&addrconf_hash_lock);
3021 return if6_get_idx(seq, *pos);
3024 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3026 struct inet6_ifaddr *ifa;
3028 ifa = if6_get_next(seq, v);
3029 ++*pos;
3030 return ifa;
3033 static void if6_seq_stop(struct seq_file *seq, void *v)
3034 __releases(addrconf_hash_lock)
3036 read_unlock_bh(&addrconf_hash_lock);
3039 static int if6_seq_show(struct seq_file *seq, void *v)
3041 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3042 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3043 &ifp->addr,
3044 ifp->idev->dev->ifindex,
3045 ifp->prefix_len,
3046 ifp->scope,
3047 ifp->flags,
3048 ifp->idev->dev->name);
3049 return 0;
3052 static const struct seq_operations if6_seq_ops = {
3053 .start = if6_seq_start,
3054 .next = if6_seq_next,
3055 .show = if6_seq_show,
3056 .stop = if6_seq_stop,
3059 static int if6_seq_open(struct inode *inode, struct file *file)
3061 return seq_open_net(inode, file, &if6_seq_ops,
3062 sizeof(struct if6_iter_state));
3065 static const struct file_operations if6_fops = {
3066 .owner = THIS_MODULE,
3067 .open = if6_seq_open,
3068 .read = seq_read,
3069 .llseek = seq_lseek,
3070 .release = seq_release_net,
3073 static int __net_init if6_proc_net_init(struct net *net)
3075 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3076 return -ENOMEM;
3077 return 0;
3080 static void __net_exit if6_proc_net_exit(struct net *net)
3082 proc_net_remove(net, "if_inet6");
3085 static struct pernet_operations if6_proc_net_ops = {
3086 .init = if6_proc_net_init,
3087 .exit = if6_proc_net_exit,
3090 int __init if6_proc_init(void)
3092 return register_pernet_subsys(&if6_proc_net_ops);
3095 void if6_proc_exit(void)
3097 unregister_pernet_subsys(&if6_proc_net_ops);
3099 #endif /* CONFIG_PROC_FS */
3101 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3102 /* Check if address is a home address configured on any interface. */
3103 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3105 int ret = 0;
3106 struct inet6_ifaddr * ifp;
3107 u8 hash = ipv6_addr_hash(addr);
3108 read_lock_bh(&addrconf_hash_lock);
3109 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3110 if (!net_eq(dev_net(ifp->idev->dev), net))
3111 continue;
3112 if (ipv6_addr_equal(&ifp->addr, addr) &&
3113 (ifp->flags & IFA_F_HOMEADDRESS)) {
3114 ret = 1;
3115 break;
3118 read_unlock_bh(&addrconf_hash_lock);
3119 return ret;
3121 #endif
3124 * Periodic address status verification
3127 static void addrconf_verify(unsigned long foo)
3129 struct inet6_ifaddr *ifp;
3130 unsigned long now, next;
3131 int i;
3133 spin_lock_bh(&addrconf_verify_lock);
3134 now = jiffies;
3135 next = now + ADDR_CHECK_FREQUENCY;
3137 del_timer(&addr_chk_timer);
3139 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3141 restart:
3142 read_lock(&addrconf_hash_lock);
3143 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3144 unsigned long age;
3145 #ifdef CONFIG_IPV6_PRIVACY
3146 unsigned long regen_advance;
3147 #endif
3149 if (ifp->flags & IFA_F_PERMANENT)
3150 continue;
3152 spin_lock(&ifp->lock);
3153 age = (now - ifp->tstamp) / HZ;
3155 #ifdef CONFIG_IPV6_PRIVACY
3156 regen_advance = ifp->idev->cnf.regen_max_retry *
3157 ifp->idev->cnf.dad_transmits *
3158 ifp->idev->nd_parms->retrans_time / HZ;
3159 #endif
3161 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3162 age >= ifp->valid_lft) {
3163 spin_unlock(&ifp->lock);
3164 in6_ifa_hold(ifp);
3165 read_unlock(&addrconf_hash_lock);
3166 ipv6_del_addr(ifp);
3167 goto restart;
3168 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3169 spin_unlock(&ifp->lock);
3170 continue;
3171 } else if (age >= ifp->prefered_lft) {
3172 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3173 int deprecate = 0;
3175 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3176 deprecate = 1;
3177 ifp->flags |= IFA_F_DEPRECATED;
3180 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3181 next = ifp->tstamp + ifp->valid_lft * HZ;
3183 spin_unlock(&ifp->lock);
3185 if (deprecate) {
3186 in6_ifa_hold(ifp);
3187 read_unlock(&addrconf_hash_lock);
3189 ipv6_ifa_notify(0, ifp);
3190 in6_ifa_put(ifp);
3191 goto restart;
3193 #ifdef CONFIG_IPV6_PRIVACY
3194 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3195 !(ifp->flags&IFA_F_TENTATIVE)) {
3196 if (age >= ifp->prefered_lft - regen_advance) {
3197 struct inet6_ifaddr *ifpub = ifp->ifpub;
3198 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3199 next = ifp->tstamp + ifp->prefered_lft * HZ;
3200 if (!ifp->regen_count && ifpub) {
3201 ifp->regen_count++;
3202 in6_ifa_hold(ifp);
3203 in6_ifa_hold(ifpub);
3204 spin_unlock(&ifp->lock);
3205 read_unlock(&addrconf_hash_lock);
3206 spin_lock(&ifpub->lock);
3207 ifpub->regen_count = 0;
3208 spin_unlock(&ifpub->lock);
3209 ipv6_create_tempaddr(ifpub, ifp);
3210 in6_ifa_put(ifpub);
3211 in6_ifa_put(ifp);
3212 goto restart;
3214 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3215 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3216 spin_unlock(&ifp->lock);
3217 #endif
3218 } else {
3219 /* ifp->prefered_lft <= ifp->valid_lft */
3220 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3221 next = ifp->tstamp + ifp->prefered_lft * HZ;
3222 spin_unlock(&ifp->lock);
3225 read_unlock(&addrconf_hash_lock);
3228 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3229 add_timer(&addr_chk_timer);
3230 spin_unlock_bh(&addrconf_verify_lock);
3233 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3235 struct in6_addr *pfx = NULL;
3237 if (addr)
3238 pfx = nla_data(addr);
3240 if (local) {
3241 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3242 pfx = NULL;
3243 else
3244 pfx = nla_data(local);
3247 return pfx;
3250 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3251 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3252 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3253 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3256 static int
3257 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3259 struct net *net = sock_net(skb->sk);
3260 struct ifaddrmsg *ifm;
3261 struct nlattr *tb[IFA_MAX+1];
3262 struct in6_addr *pfx;
3263 int err;
3265 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3266 if (err < 0)
3267 return err;
3269 ifm = nlmsg_data(nlh);
3270 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3271 if (pfx == NULL)
3272 return -EINVAL;
3274 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3277 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3278 u32 prefered_lft, u32 valid_lft)
3280 u32 flags;
3281 clock_t expires;
3282 unsigned long timeout;
3284 if (!valid_lft || (prefered_lft > valid_lft))
3285 return -EINVAL;
3287 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3288 if (addrconf_finite_timeout(timeout)) {
3289 expires = jiffies_to_clock_t(timeout * HZ);
3290 valid_lft = timeout;
3291 flags = RTF_EXPIRES;
3292 } else {
3293 expires = 0;
3294 flags = 0;
3295 ifa_flags |= IFA_F_PERMANENT;
3298 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3299 if (addrconf_finite_timeout(timeout)) {
3300 if (timeout == 0)
3301 ifa_flags |= IFA_F_DEPRECATED;
3302 prefered_lft = timeout;
3305 spin_lock_bh(&ifp->lock);
3306 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3307 ifp->tstamp = jiffies;
3308 ifp->valid_lft = valid_lft;
3309 ifp->prefered_lft = prefered_lft;
3311 spin_unlock_bh(&ifp->lock);
3312 if (!(ifp->flags&IFA_F_TENTATIVE))
3313 ipv6_ifa_notify(0, ifp);
3315 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3316 expires, flags);
3317 addrconf_verify(0);
3319 return 0;
3322 static int
3323 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3325 struct net *net = sock_net(skb->sk);
3326 struct ifaddrmsg *ifm;
3327 struct nlattr *tb[IFA_MAX+1];
3328 struct in6_addr *pfx;
3329 struct inet6_ifaddr *ifa;
3330 struct net_device *dev;
3331 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3332 u8 ifa_flags;
3333 int err;
3335 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3336 if (err < 0)
3337 return err;
3339 ifm = nlmsg_data(nlh);
3340 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3341 if (pfx == NULL)
3342 return -EINVAL;
3344 if (tb[IFA_CACHEINFO]) {
3345 struct ifa_cacheinfo *ci;
3347 ci = nla_data(tb[IFA_CACHEINFO]);
3348 valid_lft = ci->ifa_valid;
3349 preferred_lft = ci->ifa_prefered;
3350 } else {
3351 preferred_lft = INFINITY_LIFE_TIME;
3352 valid_lft = INFINITY_LIFE_TIME;
3355 dev = __dev_get_by_index(net, ifm->ifa_index);
3356 if (dev == NULL)
3357 return -ENODEV;
3359 /* We ignore other flags so far. */
3360 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3362 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3363 if (ifa == NULL) {
3365 * It would be best to check for !NLM_F_CREATE here but
3366 * userspace alreay relies on not having to provide this.
3368 return inet6_addr_add(net, ifm->ifa_index, pfx,
3369 ifm->ifa_prefixlen, ifa_flags,
3370 preferred_lft, valid_lft);
3373 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3374 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3375 err = -EEXIST;
3376 else
3377 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3379 in6_ifa_put(ifa);
3381 return err;
3384 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3385 u8 scope, int ifindex)
3387 struct ifaddrmsg *ifm;
3389 ifm = nlmsg_data(nlh);
3390 ifm->ifa_family = AF_INET6;
3391 ifm->ifa_prefixlen = prefixlen;
3392 ifm->ifa_flags = flags;
3393 ifm->ifa_scope = scope;
3394 ifm->ifa_index = ifindex;
3397 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3398 unsigned long tstamp, u32 preferred, u32 valid)
3400 struct ifa_cacheinfo ci;
3402 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3403 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3404 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3405 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3406 ci.ifa_prefered = preferred;
3407 ci.ifa_valid = valid;
3409 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3412 static inline int rt_scope(int ifa_scope)
3414 if (ifa_scope & IFA_HOST)
3415 return RT_SCOPE_HOST;
3416 else if (ifa_scope & IFA_LINK)
3417 return RT_SCOPE_LINK;
3418 else if (ifa_scope & IFA_SITE)
3419 return RT_SCOPE_SITE;
3420 else
3421 return RT_SCOPE_UNIVERSE;
3424 static inline int inet6_ifaddr_msgsize(void)
3426 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3427 + nla_total_size(16) /* IFA_ADDRESS */
3428 + nla_total_size(sizeof(struct ifa_cacheinfo));
3431 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3432 u32 pid, u32 seq, int event, unsigned int flags)
3434 struct nlmsghdr *nlh;
3435 u32 preferred, valid;
3437 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3438 if (nlh == NULL)
3439 return -EMSGSIZE;
3441 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3442 ifa->idev->dev->ifindex);
3444 if (!(ifa->flags&IFA_F_PERMANENT)) {
3445 preferred = ifa->prefered_lft;
3446 valid = ifa->valid_lft;
3447 if (preferred != INFINITY_LIFE_TIME) {
3448 long tval = (jiffies - ifa->tstamp)/HZ;
3449 if (preferred > tval)
3450 preferred -= tval;
3451 else
3452 preferred = 0;
3453 if (valid != INFINITY_LIFE_TIME)
3454 valid -= tval;
3456 } else {
3457 preferred = INFINITY_LIFE_TIME;
3458 valid = INFINITY_LIFE_TIME;
3461 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3462 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3463 nlmsg_cancel(skb, nlh);
3464 return -EMSGSIZE;
3467 return nlmsg_end(skb, nlh);
3470 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3471 u32 pid, u32 seq, int event, u16 flags)
3473 struct nlmsghdr *nlh;
3474 u8 scope = RT_SCOPE_UNIVERSE;
3475 int ifindex = ifmca->idev->dev->ifindex;
3477 if (ipv6_addr_scope(&ifmca->mca_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_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3486 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3487 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3488 nlmsg_cancel(skb, nlh);
3489 return -EMSGSIZE;
3492 return nlmsg_end(skb, nlh);
3495 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3496 u32 pid, u32 seq, int event, unsigned int flags)
3498 struct nlmsghdr *nlh;
3499 u8 scope = RT_SCOPE_UNIVERSE;
3500 int ifindex = ifaca->aca_idev->dev->ifindex;
3502 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3503 scope = RT_SCOPE_SITE;
3505 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3506 if (nlh == NULL)
3507 return -EMSGSIZE;
3509 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3510 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3511 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3512 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3513 nlmsg_cancel(skb, nlh);
3514 return -EMSGSIZE;
3517 return nlmsg_end(skb, nlh);
3520 enum addr_type_t
3522 UNICAST_ADDR,
3523 MULTICAST_ADDR,
3524 ANYCAST_ADDR,
3527 /* called with rcu_read_lock() */
3528 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3529 struct netlink_callback *cb, enum addr_type_t type,
3530 int s_ip_idx, int *p_ip_idx)
3532 struct inet6_ifaddr *ifa;
3533 struct ifmcaddr6 *ifmca;
3534 struct ifacaddr6 *ifaca;
3535 int err = 1;
3536 int ip_idx = *p_ip_idx;
3538 read_lock_bh(&idev->lock);
3539 switch (type) {
3540 case UNICAST_ADDR:
3541 /* unicast address incl. temp addr */
3542 for (ifa = idev->addr_list; ifa;
3543 ifa = ifa->if_next, ip_idx++) {
3544 if (ip_idx < s_ip_idx)
3545 continue;
3546 err = inet6_fill_ifaddr(skb, ifa,
3547 NETLINK_CB(cb->skb).pid,
3548 cb->nlh->nlmsg_seq,
3549 RTM_NEWADDR,
3550 NLM_F_MULTI);
3551 if (err <= 0)
3552 break;
3554 break;
3555 case MULTICAST_ADDR:
3556 /* multicast address */
3557 for (ifmca = idev->mc_list; ifmca;
3558 ifmca = ifmca->next, ip_idx++) {
3559 if (ip_idx < s_ip_idx)
3560 continue;
3561 err = inet6_fill_ifmcaddr(skb, ifmca,
3562 NETLINK_CB(cb->skb).pid,
3563 cb->nlh->nlmsg_seq,
3564 RTM_GETMULTICAST,
3565 NLM_F_MULTI);
3566 if (err <= 0)
3567 break;
3569 break;
3570 case ANYCAST_ADDR:
3571 /* anycast address */
3572 for (ifaca = idev->ac_list; ifaca;
3573 ifaca = ifaca->aca_next, ip_idx++) {
3574 if (ip_idx < s_ip_idx)
3575 continue;
3576 err = inet6_fill_ifacaddr(skb, ifaca,
3577 NETLINK_CB(cb->skb).pid,
3578 cb->nlh->nlmsg_seq,
3579 RTM_GETANYCAST,
3580 NLM_F_MULTI);
3581 if (err <= 0)
3582 break;
3584 break;
3585 default:
3586 break;
3588 read_unlock_bh(&idev->lock);
3589 *p_ip_idx = ip_idx;
3590 return err;
3593 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3594 enum addr_type_t type)
3596 struct net *net = sock_net(skb->sk);
3597 int h, s_h;
3598 int idx, ip_idx;
3599 int s_idx, s_ip_idx;
3600 struct net_device *dev;
3601 struct inet6_dev *idev;
3602 struct hlist_head *head;
3603 struct hlist_node *node;
3605 s_h = cb->args[0];
3606 s_idx = idx = cb->args[1];
3607 s_ip_idx = ip_idx = cb->args[2];
3609 rcu_read_lock();
3610 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3611 idx = 0;
3612 head = &net->dev_index_head[h];
3613 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3614 if (idx < s_idx)
3615 goto cont;
3616 if (h > s_h || idx > s_idx)
3617 s_ip_idx = 0;
3618 ip_idx = 0;
3619 if ((idev = __in6_dev_get(dev)) == NULL)
3620 goto cont;
3622 if (in6_dump_addrs(idev, skb, cb, type,
3623 s_ip_idx, &ip_idx) <= 0)
3624 goto done;
3625 cont:
3626 idx++;
3629 done:
3630 rcu_read_unlock();
3631 cb->args[0] = h;
3632 cb->args[1] = idx;
3633 cb->args[2] = ip_idx;
3635 return skb->len;
3638 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3640 enum addr_type_t type = UNICAST_ADDR;
3642 return inet6_dump_addr(skb, cb, type);
3645 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3647 enum addr_type_t type = MULTICAST_ADDR;
3649 return inet6_dump_addr(skb, cb, type);
3653 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3655 enum addr_type_t type = ANYCAST_ADDR;
3657 return inet6_dump_addr(skb, cb, type);
3660 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3661 void *arg)
3663 struct net *net = sock_net(in_skb->sk);
3664 struct ifaddrmsg *ifm;
3665 struct nlattr *tb[IFA_MAX+1];
3666 struct in6_addr *addr = NULL;
3667 struct net_device *dev = NULL;
3668 struct inet6_ifaddr *ifa;
3669 struct sk_buff *skb;
3670 int err;
3672 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3673 if (err < 0)
3674 goto errout;
3676 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3677 if (addr == NULL) {
3678 err = -EINVAL;
3679 goto errout;
3682 ifm = nlmsg_data(nlh);
3683 if (ifm->ifa_index)
3684 dev = __dev_get_by_index(net, ifm->ifa_index);
3686 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3687 err = -EADDRNOTAVAIL;
3688 goto errout;
3691 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3692 err = -ENOBUFS;
3693 goto errout_ifa;
3696 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3697 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3698 if (err < 0) {
3699 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3700 WARN_ON(err == -EMSGSIZE);
3701 kfree_skb(skb);
3702 goto errout_ifa;
3704 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3705 errout_ifa:
3706 in6_ifa_put(ifa);
3707 errout:
3708 return err;
3711 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3713 struct sk_buff *skb;
3714 struct net *net = dev_net(ifa->idev->dev);
3715 int err = -ENOBUFS;
3717 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3718 if (skb == NULL)
3719 goto errout;
3721 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3722 if (err < 0) {
3723 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3724 WARN_ON(err == -EMSGSIZE);
3725 kfree_skb(skb);
3726 goto errout;
3728 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3729 return;
3730 errout:
3731 if (err < 0)
3732 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3735 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3736 __s32 *array, int bytes)
3738 BUG_ON(bytes < (DEVCONF_MAX * 4));
3740 memset(array, 0, bytes);
3741 array[DEVCONF_FORWARDING] = cnf->forwarding;
3742 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3743 array[DEVCONF_MTU6] = cnf->mtu6;
3744 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3745 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3746 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3747 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3748 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3749 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3750 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3751 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3752 #ifdef CONFIG_IPV6_PRIVACY
3753 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3754 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3755 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3756 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3757 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3758 #endif
3759 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3760 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3761 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3762 #ifdef CONFIG_IPV6_ROUTER_PREF
3763 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3764 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3765 #ifdef CONFIG_IPV6_ROUTE_INFO
3766 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3767 #endif
3768 #endif
3769 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3770 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3771 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3772 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3773 #endif
3774 #ifdef CONFIG_IPV6_MROUTE
3775 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3776 #endif
3777 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3778 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3779 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3782 static inline size_t inet6_if_nlmsg_size(void)
3784 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3785 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3786 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3787 + nla_total_size(4) /* IFLA_MTU */
3788 + nla_total_size(4) /* IFLA_LINK */
3789 + nla_total_size( /* IFLA_PROTINFO */
3790 nla_total_size(4) /* IFLA_INET6_FLAGS */
3791 + nla_total_size(sizeof(struct ifla_cacheinfo))
3792 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3793 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3794 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3798 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3799 int items, int bytes)
3801 int i;
3802 int pad = bytes - sizeof(u64) * items;
3803 BUG_ON(pad < 0);
3805 /* Use put_unaligned() because stats may not be aligned for u64. */
3806 put_unaligned(items, &stats[0]);
3807 for (i = 1; i < items; i++)
3808 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3810 memset(&stats[items], 0, pad);
3813 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3814 int bytes)
3816 switch(attrtype) {
3817 case IFLA_INET6_STATS:
3818 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3819 break;
3820 case IFLA_INET6_ICMP6STATS:
3821 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3822 break;
3826 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3827 u32 pid, u32 seq, int event, unsigned int flags)
3829 struct net_device *dev = idev->dev;
3830 struct nlattr *nla;
3831 struct ifinfomsg *hdr;
3832 struct nlmsghdr *nlh;
3833 void *protoinfo;
3834 struct ifla_cacheinfo ci;
3836 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3837 if (nlh == NULL)
3838 return -EMSGSIZE;
3840 hdr = nlmsg_data(nlh);
3841 hdr->ifi_family = AF_INET6;
3842 hdr->__ifi_pad = 0;
3843 hdr->ifi_type = dev->type;
3844 hdr->ifi_index = dev->ifindex;
3845 hdr->ifi_flags = dev_get_flags(dev);
3846 hdr->ifi_change = 0;
3848 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3850 if (dev->addr_len)
3851 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3853 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3854 if (dev->ifindex != dev->iflink)
3855 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3857 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3858 if (protoinfo == NULL)
3859 goto nla_put_failure;
3861 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3863 ci.max_reasm_len = IPV6_MAXPLEN;
3864 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3865 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3866 ci.reachable_time = idev->nd_parms->reachable_time;
3867 ci.retrans_time = idev->nd_parms->retrans_time;
3868 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3870 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3871 if (nla == NULL)
3872 goto nla_put_failure;
3873 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3875 /* XXX - MC not implemented */
3877 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3878 if (nla == NULL)
3879 goto nla_put_failure;
3880 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3882 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3883 if (nla == NULL)
3884 goto nla_put_failure;
3885 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3887 nla_nest_end(skb, protoinfo);
3888 return nlmsg_end(skb, nlh);
3890 nla_put_failure:
3891 nlmsg_cancel(skb, nlh);
3892 return -EMSGSIZE;
3895 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3897 struct net *net = sock_net(skb->sk);
3898 int h, s_h;
3899 int idx = 0, s_idx;
3900 struct net_device *dev;
3901 struct inet6_dev *idev;
3902 struct hlist_head *head;
3903 struct hlist_node *node;
3905 s_h = cb->args[0];
3906 s_idx = cb->args[1];
3908 rcu_read_lock();
3909 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3910 idx = 0;
3911 head = &net->dev_index_head[h];
3912 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3913 if (idx < s_idx)
3914 goto cont;
3915 idev = __in6_dev_get(dev);
3916 if (!idev)
3917 goto cont;
3918 if (inet6_fill_ifinfo(skb, idev,
3919 NETLINK_CB(cb->skb).pid,
3920 cb->nlh->nlmsg_seq,
3921 RTM_NEWLINK, NLM_F_MULTI) <= 0)
3922 goto out;
3923 cont:
3924 idx++;
3927 out:
3928 rcu_read_unlock();
3929 cb->args[1] = idx;
3930 cb->args[0] = h;
3932 return skb->len;
3935 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3937 struct sk_buff *skb;
3938 struct net *net = dev_net(idev->dev);
3939 int err = -ENOBUFS;
3941 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3942 if (skb == NULL)
3943 goto errout;
3945 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3946 if (err < 0) {
3947 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3948 WARN_ON(err == -EMSGSIZE);
3949 kfree_skb(skb);
3950 goto errout;
3952 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3953 return;
3954 errout:
3955 if (err < 0)
3956 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3959 static inline size_t inet6_prefix_nlmsg_size(void)
3961 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3962 + nla_total_size(sizeof(struct in6_addr))
3963 + nla_total_size(sizeof(struct prefix_cacheinfo));
3966 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3967 struct prefix_info *pinfo, u32 pid, u32 seq,
3968 int event, unsigned int flags)
3970 struct prefixmsg *pmsg;
3971 struct nlmsghdr *nlh;
3972 struct prefix_cacheinfo ci;
3974 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3975 if (nlh == NULL)
3976 return -EMSGSIZE;
3978 pmsg = nlmsg_data(nlh);
3979 pmsg->prefix_family = AF_INET6;
3980 pmsg->prefix_pad1 = 0;
3981 pmsg->prefix_pad2 = 0;
3982 pmsg->prefix_ifindex = idev->dev->ifindex;
3983 pmsg->prefix_len = pinfo->prefix_len;
3984 pmsg->prefix_type = pinfo->type;
3985 pmsg->prefix_pad3 = 0;
3986 pmsg->prefix_flags = 0;
3987 if (pinfo->onlink)
3988 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3989 if (pinfo->autoconf)
3990 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3992 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3994 ci.preferred_time = ntohl(pinfo->prefered);
3995 ci.valid_time = ntohl(pinfo->valid);
3996 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3998 return nlmsg_end(skb, nlh);
4000 nla_put_failure:
4001 nlmsg_cancel(skb, nlh);
4002 return -EMSGSIZE;
4005 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4006 struct prefix_info *pinfo)
4008 struct sk_buff *skb;
4009 struct net *net = dev_net(idev->dev);
4010 int err = -ENOBUFS;
4012 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4013 if (skb == NULL)
4014 goto errout;
4016 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4017 if (err < 0) {
4018 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4019 WARN_ON(err == -EMSGSIZE);
4020 kfree_skb(skb);
4021 goto errout;
4023 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4024 return;
4025 errout:
4026 if (err < 0)
4027 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4030 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4032 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4034 switch (event) {
4035 case RTM_NEWADDR:
4037 * If the address was optimistic
4038 * we inserted the route at the start of
4039 * our DAD process, so we don't need
4040 * to do it again
4042 if (!(ifp->rt->rt6i_node))
4043 ip6_ins_rt(ifp->rt);
4044 if (ifp->idev->cnf.forwarding)
4045 addrconf_join_anycast(ifp);
4046 break;
4047 case RTM_DELADDR:
4048 if (ifp->idev->cnf.forwarding)
4049 addrconf_leave_anycast(ifp);
4050 addrconf_leave_solict(ifp->idev, &ifp->addr);
4051 dst_hold(&ifp->rt->u.dst);
4053 if (ifp->dead && ip6_del_rt(ifp->rt))
4054 dst_free(&ifp->rt->u.dst);
4055 break;
4059 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4061 rcu_read_lock_bh();
4062 if (likely(ifp->idev->dead == 0))
4063 __ipv6_ifa_notify(event, ifp);
4064 rcu_read_unlock_bh();
4067 #ifdef CONFIG_SYSCTL
4069 static
4070 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4071 void __user *buffer, size_t *lenp, loff_t *ppos)
4073 int *valp = ctl->data;
4074 int val = *valp;
4075 loff_t pos = *ppos;
4076 int ret;
4078 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4080 if (write)
4081 ret = addrconf_fixup_forwarding(ctl, valp, val);
4082 if (ret)
4083 *ppos = pos;
4084 return ret;
4087 static void dev_disable_change(struct inet6_dev *idev)
4089 if (!idev || !idev->dev)
4090 return;
4092 if (idev->cnf.disable_ipv6)
4093 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4094 else
4095 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4098 static void addrconf_disable_change(struct net *net, __s32 newf)
4100 struct net_device *dev;
4101 struct inet6_dev *idev;
4103 rcu_read_lock();
4104 for_each_netdev_rcu(net, dev) {
4105 idev = __in6_dev_get(dev);
4106 if (idev) {
4107 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4108 idev->cnf.disable_ipv6 = newf;
4109 if (changed)
4110 dev_disable_change(idev);
4113 rcu_read_unlock();
4116 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4118 struct net *net;
4120 net = (struct net *)table->extra2;
4122 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4123 return 0;
4125 if (!rtnl_trylock()) {
4126 /* Restore the original values before restarting */
4127 *p = old;
4128 return restart_syscall();
4131 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4132 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4133 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4134 addrconf_disable_change(net, newf);
4135 } else if ((!*p) ^ (!old))
4136 dev_disable_change((struct inet6_dev *)table->extra1);
4138 rtnl_unlock();
4139 return 0;
4142 static
4143 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4144 void __user *buffer, size_t *lenp, loff_t *ppos)
4146 int *valp = ctl->data;
4147 int val = *valp;
4148 loff_t pos = *ppos;
4149 int ret;
4151 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4153 if (write)
4154 ret = addrconf_disable_ipv6(ctl, valp, val);
4155 if (ret)
4156 *ppos = pos;
4157 return ret;
4160 static struct addrconf_sysctl_table
4162 struct ctl_table_header *sysctl_header;
4163 ctl_table addrconf_vars[DEVCONF_MAX+1];
4164 char *dev_name;
4165 } addrconf_sysctl __read_mostly = {
4166 .sysctl_header = NULL,
4167 .addrconf_vars = {
4169 .procname = "forwarding",
4170 .data = &ipv6_devconf.forwarding,
4171 .maxlen = sizeof(int),
4172 .mode = 0644,
4173 .proc_handler = addrconf_sysctl_forward,
4176 .procname = "hop_limit",
4177 .data = &ipv6_devconf.hop_limit,
4178 .maxlen = sizeof(int),
4179 .mode = 0644,
4180 .proc_handler = proc_dointvec,
4183 .procname = "mtu",
4184 .data = &ipv6_devconf.mtu6,
4185 .maxlen = sizeof(int),
4186 .mode = 0644,
4187 .proc_handler = proc_dointvec,
4190 .procname = "accept_ra",
4191 .data = &ipv6_devconf.accept_ra,
4192 .maxlen = sizeof(int),
4193 .mode = 0644,
4194 .proc_handler = proc_dointvec,
4197 .procname = "accept_redirects",
4198 .data = &ipv6_devconf.accept_redirects,
4199 .maxlen = sizeof(int),
4200 .mode = 0644,
4201 .proc_handler = proc_dointvec,
4204 .procname = "autoconf",
4205 .data = &ipv6_devconf.autoconf,
4206 .maxlen = sizeof(int),
4207 .mode = 0644,
4208 .proc_handler = proc_dointvec,
4211 .procname = "dad_transmits",
4212 .data = &ipv6_devconf.dad_transmits,
4213 .maxlen = sizeof(int),
4214 .mode = 0644,
4215 .proc_handler = proc_dointvec,
4218 .procname = "router_solicitations",
4219 .data = &ipv6_devconf.rtr_solicits,
4220 .maxlen = sizeof(int),
4221 .mode = 0644,
4222 .proc_handler = proc_dointvec,
4225 .procname = "router_solicitation_interval",
4226 .data = &ipv6_devconf.rtr_solicit_interval,
4227 .maxlen = sizeof(int),
4228 .mode = 0644,
4229 .proc_handler = proc_dointvec_jiffies,
4232 .procname = "router_solicitation_delay",
4233 .data = &ipv6_devconf.rtr_solicit_delay,
4234 .maxlen = sizeof(int),
4235 .mode = 0644,
4236 .proc_handler = proc_dointvec_jiffies,
4239 .procname = "force_mld_version",
4240 .data = &ipv6_devconf.force_mld_version,
4241 .maxlen = sizeof(int),
4242 .mode = 0644,
4243 .proc_handler = proc_dointvec,
4245 #ifdef CONFIG_IPV6_PRIVACY
4247 .procname = "use_tempaddr",
4248 .data = &ipv6_devconf.use_tempaddr,
4249 .maxlen = sizeof(int),
4250 .mode = 0644,
4251 .proc_handler = proc_dointvec,
4254 .procname = "temp_valid_lft",
4255 .data = &ipv6_devconf.temp_valid_lft,
4256 .maxlen = sizeof(int),
4257 .mode = 0644,
4258 .proc_handler = proc_dointvec,
4261 .procname = "temp_prefered_lft",
4262 .data = &ipv6_devconf.temp_prefered_lft,
4263 .maxlen = sizeof(int),
4264 .mode = 0644,
4265 .proc_handler = proc_dointvec,
4268 .procname = "regen_max_retry",
4269 .data = &ipv6_devconf.regen_max_retry,
4270 .maxlen = sizeof(int),
4271 .mode = 0644,
4272 .proc_handler = proc_dointvec,
4275 .procname = "max_desync_factor",
4276 .data = &ipv6_devconf.max_desync_factor,
4277 .maxlen = sizeof(int),
4278 .mode = 0644,
4279 .proc_handler = proc_dointvec,
4281 #endif
4283 .procname = "max_addresses",
4284 .data = &ipv6_devconf.max_addresses,
4285 .maxlen = sizeof(int),
4286 .mode = 0644,
4287 .proc_handler = proc_dointvec,
4290 .procname = "accept_ra_defrtr",
4291 .data = &ipv6_devconf.accept_ra_defrtr,
4292 .maxlen = sizeof(int),
4293 .mode = 0644,
4294 .proc_handler = proc_dointvec,
4297 .procname = "accept_ra_pinfo",
4298 .data = &ipv6_devconf.accept_ra_pinfo,
4299 .maxlen = sizeof(int),
4300 .mode = 0644,
4301 .proc_handler = proc_dointvec,
4303 #ifdef CONFIG_IPV6_ROUTER_PREF
4305 .procname = "accept_ra_rtr_pref",
4306 .data = &ipv6_devconf.accept_ra_rtr_pref,
4307 .maxlen = sizeof(int),
4308 .mode = 0644,
4309 .proc_handler = proc_dointvec,
4312 .procname = "router_probe_interval",
4313 .data = &ipv6_devconf.rtr_probe_interval,
4314 .maxlen = sizeof(int),
4315 .mode = 0644,
4316 .proc_handler = proc_dointvec_jiffies,
4318 #ifdef CONFIG_IPV6_ROUTE_INFO
4320 .procname = "accept_ra_rt_info_max_plen",
4321 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4322 .maxlen = sizeof(int),
4323 .mode = 0644,
4324 .proc_handler = proc_dointvec,
4326 #endif
4327 #endif
4329 .procname = "proxy_ndp",
4330 .data = &ipv6_devconf.proxy_ndp,
4331 .maxlen = sizeof(int),
4332 .mode = 0644,
4333 .proc_handler = proc_dointvec,
4336 .procname = "accept_source_route",
4337 .data = &ipv6_devconf.accept_source_route,
4338 .maxlen = sizeof(int),
4339 .mode = 0644,
4340 .proc_handler = proc_dointvec,
4342 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4344 .procname = "optimistic_dad",
4345 .data = &ipv6_devconf.optimistic_dad,
4346 .maxlen = sizeof(int),
4347 .mode = 0644,
4348 .proc_handler = proc_dointvec,
4351 #endif
4352 #ifdef CONFIG_IPV6_MROUTE
4354 .procname = "mc_forwarding",
4355 .data = &ipv6_devconf.mc_forwarding,
4356 .maxlen = sizeof(int),
4357 .mode = 0444,
4358 .proc_handler = proc_dointvec,
4360 #endif
4362 .procname = "disable_ipv6",
4363 .data = &ipv6_devconf.disable_ipv6,
4364 .maxlen = sizeof(int),
4365 .mode = 0644,
4366 .proc_handler = addrconf_sysctl_disable,
4369 .procname = "accept_dad",
4370 .data = &ipv6_devconf.accept_dad,
4371 .maxlen = sizeof(int),
4372 .mode = 0644,
4373 .proc_handler = proc_dointvec,
4376 .procname = "force_tllao",
4377 .data = &ipv6_devconf.force_tllao,
4378 .maxlen = sizeof(int),
4379 .mode = 0644,
4380 .proc_handler = proc_dointvec
4383 /* sentinel */
4388 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4389 struct inet6_dev *idev, struct ipv6_devconf *p)
4391 int i;
4392 struct addrconf_sysctl_table *t;
4394 #define ADDRCONF_CTL_PATH_DEV 3
4396 struct ctl_path addrconf_ctl_path[] = {
4397 { .procname = "net", },
4398 { .procname = "ipv6", },
4399 { .procname = "conf", },
4400 { /* to be set */ },
4401 { },
4405 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4406 if (t == NULL)
4407 goto out;
4409 for (i=0; t->addrconf_vars[i].data; i++) {
4410 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4411 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4412 t->addrconf_vars[i].extra2 = net;
4416 * Make a copy of dev_name, because '.procname' is regarded as const
4417 * by sysctl and we wouldn't want anyone to change it under our feet
4418 * (see SIOCSIFNAME).
4420 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4421 if (!t->dev_name)
4422 goto free;
4424 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4426 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4427 t->addrconf_vars);
4428 if (t->sysctl_header == NULL)
4429 goto free_procname;
4431 p->sysctl = t;
4432 return 0;
4434 free_procname:
4435 kfree(t->dev_name);
4436 free:
4437 kfree(t);
4438 out:
4439 return -ENOBUFS;
4442 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4444 struct addrconf_sysctl_table *t;
4446 if (p->sysctl == NULL)
4447 return;
4449 t = p->sysctl;
4450 p->sysctl = NULL;
4451 unregister_sysctl_table(t->sysctl_header);
4452 kfree(t->dev_name);
4453 kfree(t);
4456 static void addrconf_sysctl_register(struct inet6_dev *idev)
4458 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4459 &ndisc_ifinfo_sysctl_change);
4460 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4461 idev, &idev->cnf);
4464 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4466 __addrconf_sysctl_unregister(&idev->cnf);
4467 neigh_sysctl_unregister(idev->nd_parms);
4471 #endif
4473 static int __net_init addrconf_init_net(struct net *net)
4475 int err;
4476 struct ipv6_devconf *all, *dflt;
4478 err = -ENOMEM;
4479 all = &ipv6_devconf;
4480 dflt = &ipv6_devconf_dflt;
4482 if (!net_eq(net, &init_net)) {
4483 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4484 if (all == NULL)
4485 goto err_alloc_all;
4487 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4488 if (dflt == NULL)
4489 goto err_alloc_dflt;
4490 } else {
4491 /* these will be inherited by all namespaces */
4492 dflt->autoconf = ipv6_defaults.autoconf;
4493 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4496 net->ipv6.devconf_all = all;
4497 net->ipv6.devconf_dflt = dflt;
4499 #ifdef CONFIG_SYSCTL
4500 err = __addrconf_sysctl_register(net, "all", NULL, all);
4501 if (err < 0)
4502 goto err_reg_all;
4504 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4505 if (err < 0)
4506 goto err_reg_dflt;
4507 #endif
4508 return 0;
4510 #ifdef CONFIG_SYSCTL
4511 err_reg_dflt:
4512 __addrconf_sysctl_unregister(all);
4513 err_reg_all:
4514 kfree(dflt);
4515 #endif
4516 err_alloc_dflt:
4517 kfree(all);
4518 err_alloc_all:
4519 return err;
4522 static void __net_exit addrconf_exit_net(struct net *net)
4524 #ifdef CONFIG_SYSCTL
4525 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4526 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4527 #endif
4528 if (!net_eq(net, &init_net)) {
4529 kfree(net->ipv6.devconf_dflt);
4530 kfree(net->ipv6.devconf_all);
4534 static struct pernet_operations addrconf_ops = {
4535 .init = addrconf_init_net,
4536 .exit = addrconf_exit_net,
4540 * Device notifier
4543 int register_inet6addr_notifier(struct notifier_block *nb)
4545 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4548 EXPORT_SYMBOL(register_inet6addr_notifier);
4550 int unregister_inet6addr_notifier(struct notifier_block *nb)
4552 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4555 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4558 * Init / cleanup code
4561 int __init addrconf_init(void)
4563 int err;
4565 if ((err = ipv6_addr_label_init()) < 0) {
4566 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4567 err);
4568 return err;
4571 register_pernet_subsys(&addrconf_ops);
4573 /* The addrconf netdev notifier requires that loopback_dev
4574 * has it's ipv6 private information allocated and setup
4575 * before it can bring up and give link-local addresses
4576 * to other devices which are up.
4578 * Unfortunately, loopback_dev is not necessarily the first
4579 * entry in the global dev_base list of net devices. In fact,
4580 * it is likely to be the very last entry on that list.
4581 * So this causes the notifier registry below to try and
4582 * give link-local addresses to all devices besides loopback_dev
4583 * first, then loopback_dev, which cases all the non-loopback_dev
4584 * devices to fail to get a link-local address.
4586 * So, as a temporary fix, allocate the ipv6 structure for
4587 * loopback_dev first by hand.
4588 * Longer term, all of the dependencies ipv6 has upon the loopback
4589 * device and it being up should be removed.
4591 rtnl_lock();
4592 if (!ipv6_add_dev(init_net.loopback_dev))
4593 err = -ENOMEM;
4594 rtnl_unlock();
4595 if (err)
4596 goto errlo;
4598 register_netdevice_notifier(&ipv6_dev_notf);
4600 addrconf_verify(0);
4602 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4603 if (err < 0)
4604 goto errout;
4606 /* Only the first call to __rtnl_register can fail */
4607 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4608 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4609 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4610 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4611 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4613 ipv6_addr_label_rtnl_register();
4615 return 0;
4616 errout:
4617 unregister_netdevice_notifier(&ipv6_dev_notf);
4618 errlo:
4619 unregister_pernet_subsys(&addrconf_ops);
4621 return err;
4624 void addrconf_cleanup(void)
4626 struct inet6_ifaddr *ifa;
4627 struct net_device *dev;
4628 int i;
4630 unregister_netdevice_notifier(&ipv6_dev_notf);
4631 unregister_pernet_subsys(&addrconf_ops);
4633 rtnl_lock();
4635 /* clean dev list */
4636 for_each_netdev(&init_net, dev) {
4637 if (__in6_dev_get(dev) == NULL)
4638 continue;
4639 addrconf_ifdown(dev, 1);
4641 addrconf_ifdown(init_net.loopback_dev, 2);
4644 * Check hash table.
4646 write_lock_bh(&addrconf_hash_lock);
4647 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4648 for (ifa=inet6_addr_lst[i]; ifa; ) {
4649 struct inet6_ifaddr *bifa;
4651 bifa = ifa;
4652 ifa = ifa->lst_next;
4653 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4654 /* Do not free it; something is wrong.
4655 Now we can investigate it with debugger.
4659 write_unlock_bh(&addrconf_hash_lock);
4661 del_timer(&addr_chk_timer);
4662 rtnl_unlock();