mac80211: Reset dynamic ps timer in Rx path.
[linux-2.6/mini2440.git] / net / ipv6 / addrconf.c
blobd1f77ccea28cafd75592b190bfc7afd1940b8840
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #ifdef CONFIG_SYSCTL
57 #include <linux/sysctl.h>
58 #endif
59 #include <linux/capability.h>
60 #include <linux/delay.h>
61 #include <linux/notifier.h>
62 #include <linux/string.h>
64 #include <net/net_namespace.h>
65 #include <net/sock.h>
66 #include <net/snmp.h>
68 #include <net/ipv6.h>
69 #include <net/protocol.h>
70 #include <net/ndisc.h>
71 #include <net/ip6_route.h>
72 #include <net/addrconf.h>
73 #include <net/tcp.h>
74 #include <net/ip.h>
75 #include <net/netlink.h>
76 #include <net/pkt_sched.h>
77 #include <linux/if_tunnel.h>
78 #include <linux/rtnetlink.h>
80 #ifdef CONFIG_IPV6_PRIVACY
81 #include <linux/random.h>
82 #endif
84 #include <asm/uaccess.h>
85 #include <asm/unaligned.h>
87 #include <linux/proc_fs.h>
88 #include <linux/seq_file.h>
90 /* Set to 3 to get tracing... */
91 #define ACONF_DEBUG 2
93 #if ACONF_DEBUG >= 3
94 #define ADBG(x) printk x
95 #else
96 #define ADBG(x)
97 #endif
99 #define INFINITY_LIFE_TIME 0xFFFFFFFF
100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102 #ifdef CONFIG_SYSCTL
103 static void addrconf_sysctl_register(struct inet6_dev *idev);
104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
105 #else
106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
113 #endif
115 #ifdef CONFIG_IPV6_PRIVACY
116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
118 static void ipv6_regen_rndid(unsigned long data);
120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
121 #endif
123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
124 static int ipv6_count_addresses(struct inet6_dev *idev);
127 * Configured unicast address hash table
129 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
132 static void addrconf_verify(unsigned long);
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
140 static void addrconf_bonding_change(struct net_device *dev,
141 unsigned long event);
142 static int addrconf_ifdown(struct net_device *dev, int how);
144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
145 static void addrconf_dad_timer(unsigned long data);
146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
147 static void addrconf_dad_run(struct inet6_dev *idev);
148 static void addrconf_rs_timer(unsigned long data);
149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
153 struct prefix_info *pinfo);
154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
155 struct net_device *dev);
157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
160 .forwarding = 0,
161 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
162 .mtu6 = IPV6_MIN_MTU,
163 .accept_ra = 1,
164 .accept_redirects = 1,
165 .autoconf = 1,
166 .force_mld_version = 0,
167 .dad_transmits = 1,
168 .rtr_solicits = MAX_RTR_SOLICITATIONS,
169 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
170 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
171 #ifdef CONFIG_IPV6_PRIVACY
172 .use_tempaddr = 0,
173 .temp_valid_lft = TEMP_VALID_LIFETIME,
174 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
175 .regen_max_retry = REGEN_MAX_RETRY,
176 .max_desync_factor = MAX_DESYNC_FACTOR,
177 #endif
178 .max_addresses = IPV6_MAX_ADDRESSES,
179 .accept_ra_defrtr = 1,
180 .accept_ra_pinfo = 1,
181 #ifdef CONFIG_IPV6_ROUTER_PREF
182 .accept_ra_rtr_pref = 1,
183 .rtr_probe_interval = 60 * HZ,
184 #ifdef CONFIG_IPV6_ROUTE_INFO
185 .accept_ra_rt_info_max_plen = 0,
186 #endif
187 #endif
188 .proxy_ndp = 0,
189 .accept_source_route = 0, /* we do not accept RH0 by default. */
190 .disable_ipv6 = 0,
191 .accept_dad = 1,
194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
195 .forwarding = 0,
196 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
197 .mtu6 = IPV6_MIN_MTU,
198 .accept_ra = 1,
199 .accept_redirects = 1,
200 .autoconf = 1,
201 .dad_transmits = 1,
202 .rtr_solicits = MAX_RTR_SOLICITATIONS,
203 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
204 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
205 #ifdef CONFIG_IPV6_PRIVACY
206 .use_tempaddr = 0,
207 .temp_valid_lft = TEMP_VALID_LIFETIME,
208 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
209 .regen_max_retry = REGEN_MAX_RETRY,
210 .max_desync_factor = MAX_DESYNC_FACTOR,
211 #endif
212 .max_addresses = IPV6_MAX_ADDRESSES,
213 .accept_ra_defrtr = 1,
214 .accept_ra_pinfo = 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216 .accept_ra_rtr_pref = 1,
217 .rtr_probe_interval = 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219 .accept_ra_rt_info_max_plen = 0,
220 #endif
221 #endif
222 .proxy_ndp = 0,
223 .accept_source_route = 0, /* we do not accept RH0 by default. */
224 .disable_ipv6 = 0,
225 .accept_dad = 1,
228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
234 /* Check if a valid qdisc is available */
235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
237 return !qdisc_tx_is_noop(dev);
240 /* Check if a route is valid prefix route */
241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
243 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
248 if (del_timer(&ifp->timer))
249 __in6_ifa_put(ifp);
252 enum addrconf_timer_t
254 AC_NONE,
255 AC_DAD,
256 AC_RS,
259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
260 enum addrconf_timer_t what,
261 unsigned long when)
263 if (!del_timer(&ifp->timer))
264 in6_ifa_hold(ifp);
266 switch (what) {
267 case AC_DAD:
268 ifp->timer.function = addrconf_dad_timer;
269 break;
270 case AC_RS:
271 ifp->timer.function = addrconf_rs_timer;
272 break;
273 default:;
275 ifp->timer.expires = jiffies + when;
276 add_timer(&ifp->timer);
279 static int snmp6_alloc_dev(struct inet6_dev *idev)
281 if (snmp_mib_init((void **)idev->stats.ipv6,
282 sizeof(struct ipstats_mib)) < 0)
283 goto err_ip;
284 if (snmp_mib_init((void **)idev->stats.icmpv6,
285 sizeof(struct icmpv6_mib)) < 0)
286 goto err_icmp;
287 if (snmp_mib_init((void **)idev->stats.icmpv6msg,
288 sizeof(struct icmpv6msg_mib)) < 0)
289 goto err_icmpmsg;
291 return 0;
293 err_icmpmsg:
294 snmp_mib_free((void **)idev->stats.icmpv6);
295 err_icmp:
296 snmp_mib_free((void **)idev->stats.ipv6);
297 err_ip:
298 return -ENOMEM;
301 static void snmp6_free_dev(struct inet6_dev *idev)
303 snmp_mib_free((void **)idev->stats.icmpv6msg);
304 snmp_mib_free((void **)idev->stats.icmpv6);
305 snmp_mib_free((void **)idev->stats.ipv6);
308 /* Nobody refers to this device, we may destroy it. */
310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
312 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
313 kfree(idev);
316 void in6_dev_finish_destroy(struct inet6_dev *idev)
318 struct net_device *dev = idev->dev;
320 WARN_ON(idev->addr_list != NULL);
321 WARN_ON(idev->mc_list != NULL);
323 #ifdef NET_REFCNT_DEBUG
324 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
325 #endif
326 dev_put(dev);
327 if (!idev->dead) {
328 printk("Freeing alive inet6 device %p\n", idev);
329 return;
331 snmp6_free_dev(idev);
332 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
335 EXPORT_SYMBOL(in6_dev_finish_destroy);
337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
339 struct inet6_dev *ndev;
341 ASSERT_RTNL();
343 if (dev->mtu < IPV6_MIN_MTU)
344 return NULL;
346 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
348 if (ndev == NULL)
349 return NULL;
351 rwlock_init(&ndev->lock);
352 ndev->dev = dev;
353 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
354 ndev->cnf.mtu6 = dev->mtu;
355 ndev->cnf.sysctl = NULL;
356 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
357 if (ndev->nd_parms == NULL) {
358 kfree(ndev);
359 return NULL;
361 if (ndev->cnf.forwarding)
362 dev_disable_lro(dev);
363 /* We refer to the device */
364 dev_hold(dev);
366 if (snmp6_alloc_dev(ndev) < 0) {
367 ADBG((KERN_WARNING
368 "%s(): cannot allocate memory for statistics; dev=%s.\n",
369 __func__, dev->name));
370 neigh_parms_release(&nd_tbl, ndev->nd_parms);
371 ndev->dead = 1;
372 in6_dev_finish_destroy(ndev);
373 return NULL;
376 if (snmp6_register_dev(ndev) < 0) {
377 ADBG((KERN_WARNING
378 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
379 __func__, dev->name));
380 neigh_parms_release(&nd_tbl, ndev->nd_parms);
381 ndev->dead = 1;
382 in6_dev_finish_destroy(ndev);
383 return NULL;
386 /* One reference from device. We must do this before
387 * we invoke __ipv6_regen_rndid().
389 in6_dev_hold(ndev);
391 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
392 ndev->cnf.accept_dad = -1;
394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
395 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
396 printk(KERN_INFO
397 "%s: Disabled Multicast RS\n",
398 dev->name);
399 ndev->cnf.rtr_solicits = 0;
401 #endif
403 #ifdef CONFIG_IPV6_PRIVACY
404 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
405 if ((dev->flags&IFF_LOOPBACK) ||
406 dev->type == ARPHRD_TUNNEL ||
407 dev->type == ARPHRD_TUNNEL6 ||
408 dev->type == ARPHRD_SIT ||
409 dev->type == ARPHRD_NONE) {
410 printk(KERN_INFO
411 "%s: Disabled Privacy Extensions\n",
412 dev->name);
413 ndev->cnf.use_tempaddr = -1;
414 } else {
415 in6_dev_hold(ndev);
416 ipv6_regen_rndid((unsigned long) ndev);
418 #endif
420 if (netif_running(dev) && addrconf_qdisc_ok(dev))
421 ndev->if_flags |= IF_READY;
423 ipv6_mc_init_dev(ndev);
424 ndev->tstamp = jiffies;
425 addrconf_sysctl_register(ndev);
426 /* protected by rtnl_lock */
427 rcu_assign_pointer(dev->ip6_ptr, ndev);
429 /* Join all-node multicast group */
430 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
432 return ndev;
435 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
437 struct inet6_dev *idev;
439 ASSERT_RTNL();
441 if ((idev = __in6_dev_get(dev)) == NULL) {
442 if ((idev = ipv6_add_dev(dev)) == NULL)
443 return NULL;
446 if (dev->flags&IFF_UP)
447 ipv6_mc_up(idev);
448 return idev;
451 #ifdef CONFIG_SYSCTL
452 static void dev_forward_change(struct inet6_dev *idev)
454 struct net_device *dev;
455 struct inet6_ifaddr *ifa;
457 if (!idev)
458 return;
459 dev = idev->dev;
460 if (idev->cnf.forwarding)
461 dev_disable_lro(dev);
462 if (dev && (dev->flags & IFF_MULTICAST)) {
463 if (idev->cnf.forwarding)
464 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
465 else
466 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
468 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
469 if (ifa->flags&IFA_F_TENTATIVE)
470 continue;
471 if (idev->cnf.forwarding)
472 addrconf_join_anycast(ifa);
473 else
474 addrconf_leave_anycast(ifa);
479 static void addrconf_forward_change(struct net *net, __s32 newf)
481 struct net_device *dev;
482 struct inet6_dev *idev;
484 read_lock(&dev_base_lock);
485 for_each_netdev(net, dev) {
486 rcu_read_lock();
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);
494 rcu_read_unlock();
496 read_unlock(&dev_base_lock);
499 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
501 struct net *net;
503 net = (struct net *)table->extra2;
504 if (p == &net->ipv6.devconf_dflt->forwarding)
505 return 0;
507 if (!rtnl_trylock()) {
508 /* Restore the original values before restarting */
509 *p = old;
510 return restart_syscall();
513 if (p == &net->ipv6.devconf_all->forwarding) {
514 __s32 newf = net->ipv6.devconf_all->forwarding;
515 net->ipv6.devconf_dflt->forwarding = newf;
516 addrconf_forward_change(net, newf);
517 } else if ((!*p) ^ (!old))
518 dev_forward_change((struct inet6_dev *)table->extra1);
519 rtnl_unlock();
521 if (*p)
522 rt6_purge_dflt_routers(net);
523 return 1;
525 #endif
527 /* Nobody refers to this ifaddr, destroy it */
529 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
531 WARN_ON(ifp->if_next != NULL);
532 WARN_ON(ifp->lst_next != NULL);
534 #ifdef NET_REFCNT_DEBUG
535 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
536 #endif
538 in6_dev_put(ifp->idev);
540 if (del_timer(&ifp->timer))
541 printk("Timer is still running, when freeing ifa=%p\n", ifp);
543 if (!ifp->dead) {
544 printk("Freeing alive inet6 address %p\n", ifp);
545 return;
547 dst_release(&ifp->rt->u.dst);
549 kfree(ifp);
552 static void
553 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
555 struct inet6_ifaddr *ifa, **ifap;
556 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
559 * Each device address list is sorted in order of scope -
560 * global before linklocal.
562 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
563 ifap = &ifa->if_next) {
564 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
565 break;
568 ifp->if_next = *ifap;
569 *ifap = ifp;
573 * Hash function taken from net_alias.c
575 static u8 ipv6_addr_hash(const struct in6_addr *addr)
577 __u32 word;
580 * We perform the hash function over the last 64 bits of the address
581 * This will include the IEEE address token on links that support it.
584 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
585 word ^= (word >> 16);
586 word ^= (word >> 8);
588 return ((word ^ (word >> 4)) & 0x0f);
591 /* On success it returns ifp with increased reference count */
593 static struct inet6_ifaddr *
594 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
595 int scope, u32 flags)
597 struct inet6_ifaddr *ifa = NULL;
598 struct rt6_info *rt;
599 int hash;
600 int err = 0;
601 int addr_type = ipv6_addr_type(addr);
603 if (addr_type == IPV6_ADDR_ANY ||
604 addr_type & IPV6_ADDR_MULTICAST ||
605 (!(idev->dev->flags & IFF_LOOPBACK) &&
606 addr_type & IPV6_ADDR_LOOPBACK))
607 return ERR_PTR(-EADDRNOTAVAIL);
609 rcu_read_lock_bh();
610 if (idev->dead) {
611 err = -ENODEV; /*XXX*/
612 goto out2;
615 if (idev->cnf.disable_ipv6) {
616 err = -EACCES;
617 goto out2;
620 write_lock(&addrconf_hash_lock);
622 /* Ignore adding duplicate addresses on an interface */
623 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
624 ADBG(("ipv6_add_addr: already assigned\n"));
625 err = -EEXIST;
626 goto out;
629 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
631 if (ifa == NULL) {
632 ADBG(("ipv6_add_addr: malloc failed\n"));
633 err = -ENOBUFS;
634 goto out;
637 rt = addrconf_dst_alloc(idev, addr, 0);
638 if (IS_ERR(rt)) {
639 err = PTR_ERR(rt);
640 goto out;
643 ipv6_addr_copy(&ifa->addr, addr);
645 spin_lock_init(&ifa->lock);
646 init_timer(&ifa->timer);
647 ifa->timer.data = (unsigned long) ifa;
648 ifa->scope = scope;
649 ifa->prefix_len = pfxlen;
650 ifa->flags = flags | IFA_F_TENTATIVE;
651 ifa->cstamp = ifa->tstamp = jiffies;
653 ifa->rt = rt;
656 * part one of RFC 4429, section 3.3
657 * We should not configure an address as
658 * optimistic if we do not yet know the link
659 * layer address of our nexhop router
662 if (rt->rt6i_nexthop == NULL)
663 ifa->flags &= ~IFA_F_OPTIMISTIC;
665 ifa->idev = idev;
666 in6_dev_hold(idev);
667 /* For caller */
668 in6_ifa_hold(ifa);
670 /* Add to big hash table */
671 hash = ipv6_addr_hash(addr);
673 ifa->lst_next = inet6_addr_lst[hash];
674 inet6_addr_lst[hash] = ifa;
675 in6_ifa_hold(ifa);
676 write_unlock(&addrconf_hash_lock);
678 write_lock(&idev->lock);
679 /* Add to inet6_dev unicast addr list. */
680 ipv6_link_dev_addr(idev, ifa);
682 #ifdef CONFIG_IPV6_PRIVACY
683 if (ifa->flags&IFA_F_TEMPORARY) {
684 ifa->tmp_next = idev->tempaddr_list;
685 idev->tempaddr_list = ifa;
686 in6_ifa_hold(ifa);
688 #endif
690 in6_ifa_hold(ifa);
691 write_unlock(&idev->lock);
692 out2:
693 rcu_read_unlock_bh();
695 if (likely(err == 0))
696 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
697 else {
698 kfree(ifa);
699 ifa = ERR_PTR(err);
702 return ifa;
703 out:
704 write_unlock(&addrconf_hash_lock);
705 goto out2;
708 /* This function wants to get referenced ifp and releases it before return */
710 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
712 struct inet6_ifaddr *ifa, **ifap;
713 struct inet6_dev *idev = ifp->idev;
714 int hash;
715 int deleted = 0, onlink = 0;
716 unsigned long expires = jiffies;
718 hash = ipv6_addr_hash(&ifp->addr);
720 ifp->dead = 1;
722 write_lock_bh(&addrconf_hash_lock);
723 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
724 ifap = &ifa->lst_next) {
725 if (ifa == ifp) {
726 *ifap = ifa->lst_next;
727 __in6_ifa_put(ifp);
728 ifa->lst_next = NULL;
729 break;
732 write_unlock_bh(&addrconf_hash_lock);
734 write_lock_bh(&idev->lock);
735 #ifdef CONFIG_IPV6_PRIVACY
736 if (ifp->flags&IFA_F_TEMPORARY) {
737 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
738 ifap = &ifa->tmp_next) {
739 if (ifa == ifp) {
740 *ifap = ifa->tmp_next;
741 if (ifp->ifpub) {
742 in6_ifa_put(ifp->ifpub);
743 ifp->ifpub = NULL;
745 __in6_ifa_put(ifp);
746 ifa->tmp_next = NULL;
747 break;
751 #endif
753 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
754 if (ifa == ifp) {
755 *ifap = ifa->if_next;
756 __in6_ifa_put(ifp);
757 ifa->if_next = NULL;
758 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
759 break;
760 deleted = 1;
761 continue;
762 } else if (ifp->flags & IFA_F_PERMANENT) {
763 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
764 ifp->prefix_len)) {
765 if (ifa->flags & IFA_F_PERMANENT) {
766 onlink = 1;
767 if (deleted)
768 break;
769 } else {
770 unsigned long lifetime;
772 if (!onlink)
773 onlink = -1;
775 spin_lock(&ifa->lock);
777 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
779 * Note: Because this address is
780 * not permanent, lifetime <
781 * LONG_MAX / HZ here.
783 if (time_before(expires,
784 ifa->tstamp + lifetime * HZ))
785 expires = ifa->tstamp + lifetime * HZ;
786 spin_unlock(&ifa->lock);
790 ifap = &ifa->if_next;
792 write_unlock_bh(&idev->lock);
794 addrconf_del_timer(ifp);
796 ipv6_ifa_notify(RTM_DELADDR, ifp);
798 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
801 * Purge or update corresponding prefix
803 * 1) we don't purge prefix here if address was not permanent.
804 * prefix is managed by its own lifetime.
805 * 2) if there're no addresses, delete prefix.
806 * 3) if there're still other permanent address(es),
807 * corresponding prefix is still permanent.
808 * 4) otherwise, update prefix lifetime to the
809 * longest valid lifetime among the corresponding
810 * addresses on the device.
811 * Note: subsequent RA will update lifetime.
813 * --yoshfuji
815 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
816 struct in6_addr prefix;
817 struct rt6_info *rt;
818 struct net *net = dev_net(ifp->idev->dev);
819 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
820 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
822 if (rt && addrconf_is_prefix_route(rt)) {
823 if (onlink == 0) {
824 ip6_del_rt(rt);
825 rt = NULL;
826 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
827 rt->rt6i_expires = expires;
828 rt->rt6i_flags |= RTF_EXPIRES;
831 dst_release(&rt->u.dst);
834 in6_ifa_put(ifp);
837 #ifdef CONFIG_IPV6_PRIVACY
838 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
840 struct inet6_dev *idev = ifp->idev;
841 struct in6_addr addr, *tmpaddr;
842 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
843 unsigned long regen_advance;
844 int tmp_plen;
845 int ret = 0;
846 int max_addresses;
847 u32 addr_flags;
849 write_lock(&idev->lock);
850 if (ift) {
851 spin_lock_bh(&ift->lock);
852 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
853 spin_unlock_bh(&ift->lock);
854 tmpaddr = &addr;
855 } else {
856 tmpaddr = NULL;
858 retry:
859 in6_dev_hold(idev);
860 if (idev->cnf.use_tempaddr <= 0) {
861 write_unlock(&idev->lock);
862 printk(KERN_INFO
863 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
864 in6_dev_put(idev);
865 ret = -1;
866 goto out;
868 spin_lock_bh(&ifp->lock);
869 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
870 idev->cnf.use_tempaddr = -1; /*XXX*/
871 spin_unlock_bh(&ifp->lock);
872 write_unlock(&idev->lock);
873 printk(KERN_WARNING
874 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
875 in6_dev_put(idev);
876 ret = -1;
877 goto out;
879 in6_ifa_hold(ifp);
880 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
881 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
882 spin_unlock_bh(&ifp->lock);
883 write_unlock(&idev->lock);
884 printk(KERN_WARNING
885 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
886 in6_ifa_put(ifp);
887 in6_dev_put(idev);
888 ret = -1;
889 goto out;
891 memcpy(&addr.s6_addr[8], idev->rndid, 8);
892 tmp_valid_lft = min_t(__u32,
893 ifp->valid_lft,
894 idev->cnf.temp_valid_lft);
895 tmp_prefered_lft = min_t(__u32,
896 ifp->prefered_lft,
897 idev->cnf.temp_prefered_lft - desync_factor / HZ);
898 tmp_plen = ifp->prefix_len;
899 max_addresses = idev->cnf.max_addresses;
900 tmp_cstamp = ifp->cstamp;
901 tmp_tstamp = ifp->tstamp;
902 spin_unlock_bh(&ifp->lock);
904 regen_advance = idev->cnf.regen_max_retry *
905 idev->cnf.dad_transmits *
906 idev->nd_parms->retrans_time / HZ;
907 write_unlock(&idev->lock);
909 /* A temporary address is created only if this calculated Preferred
910 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
911 * an implementation must not create a temporary address with a zero
912 * Preferred Lifetime.
914 if (tmp_prefered_lft <= regen_advance) {
915 in6_ifa_put(ifp);
916 in6_dev_put(idev);
917 ret = -1;
918 goto out;
921 addr_flags = IFA_F_TEMPORARY;
922 /* set in addrconf_prefix_rcv() */
923 if (ifp->flags & IFA_F_OPTIMISTIC)
924 addr_flags |= IFA_F_OPTIMISTIC;
926 ift = !max_addresses ||
927 ipv6_count_addresses(idev) < max_addresses ?
928 ipv6_add_addr(idev, &addr, tmp_plen,
929 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
930 addr_flags) : NULL;
931 if (!ift || IS_ERR(ift)) {
932 in6_ifa_put(ifp);
933 in6_dev_put(idev);
934 printk(KERN_INFO
935 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
936 tmpaddr = &addr;
937 write_lock(&idev->lock);
938 goto retry;
941 spin_lock_bh(&ift->lock);
942 ift->ifpub = ifp;
943 ift->valid_lft = tmp_valid_lft;
944 ift->prefered_lft = tmp_prefered_lft;
945 ift->cstamp = tmp_cstamp;
946 ift->tstamp = tmp_tstamp;
947 spin_unlock_bh(&ift->lock);
949 addrconf_dad_start(ift, 0);
950 in6_ifa_put(ift);
951 in6_dev_put(idev);
952 out:
953 return ret;
955 #endif
958 * Choose an appropriate source address (RFC3484)
960 enum {
961 IPV6_SADDR_RULE_INIT = 0,
962 IPV6_SADDR_RULE_LOCAL,
963 IPV6_SADDR_RULE_SCOPE,
964 IPV6_SADDR_RULE_PREFERRED,
965 #ifdef CONFIG_IPV6_MIP6
966 IPV6_SADDR_RULE_HOA,
967 #endif
968 IPV6_SADDR_RULE_OIF,
969 IPV6_SADDR_RULE_LABEL,
970 #ifdef CONFIG_IPV6_PRIVACY
971 IPV6_SADDR_RULE_PRIVACY,
972 #endif
973 IPV6_SADDR_RULE_ORCHID,
974 IPV6_SADDR_RULE_PREFIX,
975 IPV6_SADDR_RULE_MAX
978 struct ipv6_saddr_score {
979 int rule;
980 int addr_type;
981 struct inet6_ifaddr *ifa;
982 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
983 int scopedist;
984 int matchlen;
987 struct ipv6_saddr_dst {
988 const struct in6_addr *addr;
989 int ifindex;
990 int scope;
991 int label;
992 unsigned int prefs;
995 static inline int ipv6_saddr_preferred(int type)
997 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
998 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
999 return 1;
1000 return 0;
1003 static int ipv6_get_saddr_eval(struct net *net,
1004 struct ipv6_saddr_score *score,
1005 struct ipv6_saddr_dst *dst,
1006 int i)
1008 int ret;
1010 if (i <= score->rule) {
1011 switch (i) {
1012 case IPV6_SADDR_RULE_SCOPE:
1013 ret = score->scopedist;
1014 break;
1015 case IPV6_SADDR_RULE_PREFIX:
1016 ret = score->matchlen;
1017 break;
1018 default:
1019 ret = !!test_bit(i, score->scorebits);
1021 goto out;
1024 switch (i) {
1025 case IPV6_SADDR_RULE_INIT:
1026 /* Rule 0: remember if hiscore is not ready yet */
1027 ret = !!score->ifa;
1028 break;
1029 case IPV6_SADDR_RULE_LOCAL:
1030 /* Rule 1: Prefer same address */
1031 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1032 break;
1033 case IPV6_SADDR_RULE_SCOPE:
1034 /* Rule 2: Prefer appropriate scope
1036 * ret
1038 * -1 | d 15
1039 * ---+--+-+---> scope
1041 * | d is scope of the destination.
1042 * B-d | \
1043 * | \ <- smaller scope is better if
1044 * B-15 | \ if scope is enough for destinaion.
1045 * | ret = B - scope (-1 <= scope >= d <= 15).
1046 * d-C-1 | /
1047 * |/ <- greater is better
1048 * -C / if scope is not enough for destination.
1049 * /| ret = scope - C (-1 <= d < scope <= 15).
1051 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1052 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1053 * Assume B = 0 and we get C > 29.
1055 ret = __ipv6_addr_src_scope(score->addr_type);
1056 if (ret >= dst->scope)
1057 ret = -ret;
1058 else
1059 ret -= 128; /* 30 is enough */
1060 score->scopedist = ret;
1061 break;
1062 case IPV6_SADDR_RULE_PREFERRED:
1063 /* Rule 3: Avoid deprecated and optimistic addresses */
1064 ret = ipv6_saddr_preferred(score->addr_type) ||
1065 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1066 break;
1067 #ifdef CONFIG_IPV6_MIP6
1068 case IPV6_SADDR_RULE_HOA:
1070 /* Rule 4: Prefer home address */
1071 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1072 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1073 break;
1075 #endif
1076 case IPV6_SADDR_RULE_OIF:
1077 /* Rule 5: Prefer outgoing interface */
1078 ret = (!dst->ifindex ||
1079 dst->ifindex == score->ifa->idev->dev->ifindex);
1080 break;
1081 case IPV6_SADDR_RULE_LABEL:
1082 /* Rule 6: Prefer matching label */
1083 ret = ipv6_addr_label(net,
1084 &score->ifa->addr, score->addr_type,
1085 score->ifa->idev->dev->ifindex) == dst->label;
1086 break;
1087 #ifdef CONFIG_IPV6_PRIVACY
1088 case IPV6_SADDR_RULE_PRIVACY:
1090 /* Rule 7: Prefer public address
1091 * Note: prefer temprary address if use_tempaddr >= 2
1093 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1094 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1095 score->ifa->idev->cnf.use_tempaddr >= 2;
1096 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1097 break;
1099 #endif
1100 case IPV6_SADDR_RULE_ORCHID:
1101 /* Rule 8-: Prefer ORCHID vs ORCHID or
1102 * non-ORCHID vs non-ORCHID
1104 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1105 ipv6_addr_orchid(dst->addr));
1106 break;
1107 case IPV6_SADDR_RULE_PREFIX:
1108 /* Rule 8: Use longest matching prefix */
1109 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1110 dst->addr);
1111 break;
1112 default:
1113 ret = 0;
1116 if (ret)
1117 __set_bit(i, score->scorebits);
1118 score->rule = i;
1119 out:
1120 return ret;
1123 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1124 const struct in6_addr *daddr, unsigned int prefs,
1125 struct in6_addr *saddr)
1127 struct ipv6_saddr_score scores[2],
1128 *score = &scores[0], *hiscore = &scores[1];
1129 struct ipv6_saddr_dst dst;
1130 struct net_device *dev;
1131 int dst_type;
1133 dst_type = __ipv6_addr_type(daddr);
1134 dst.addr = daddr;
1135 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1136 dst.scope = __ipv6_addr_src_scope(dst_type);
1137 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1138 dst.prefs = prefs;
1140 hiscore->rule = -1;
1141 hiscore->ifa = NULL;
1143 read_lock(&dev_base_lock);
1144 rcu_read_lock();
1146 for_each_netdev(net, dev) {
1147 struct inet6_dev *idev;
1149 /* Candidate Source Address (section 4)
1150 * - multicast and link-local destination address,
1151 * the set of candidate source address MUST only
1152 * include addresses assigned to interfaces
1153 * belonging to the same link as the outgoing
1154 * interface.
1155 * (- For site-local destination addresses, the
1156 * set of candidate source addresses MUST only
1157 * include addresses assigned to interfaces
1158 * belonging to the same site as the outgoing
1159 * interface.)
1161 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1162 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1163 dst.ifindex && dev->ifindex != dst.ifindex)
1164 continue;
1166 idev = __in6_dev_get(dev);
1167 if (!idev)
1168 continue;
1170 read_lock_bh(&idev->lock);
1171 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1172 int i;
1175 * - Tentative Address (RFC2462 section 5.4)
1176 * - A tentative address is not considered
1177 * "assigned to an interface" in the traditional
1178 * sense, unless it is also flagged as optimistic.
1179 * - Candidate Source Address (section 4)
1180 * - In any case, anycast addresses, multicast
1181 * addresses, and the unspecified address MUST
1182 * NOT be included in a candidate set.
1184 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1185 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1186 continue;
1188 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1190 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1191 score->addr_type & IPV6_ADDR_MULTICAST)) {
1192 LIMIT_NETDEBUG(KERN_DEBUG
1193 "ADDRCONF: unspecified / multicast address "
1194 "assigned as unicast address on %s",
1195 dev->name);
1196 continue;
1199 score->rule = -1;
1200 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1202 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1203 int minihiscore, miniscore;
1205 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1206 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1208 if (minihiscore > miniscore) {
1209 if (i == IPV6_SADDR_RULE_SCOPE &&
1210 score->scopedist > 0) {
1212 * special case:
1213 * each remaining entry
1214 * has too small (not enough)
1215 * scope, because ifa entries
1216 * are sorted by their scope
1217 * values.
1219 goto try_nextdev;
1221 break;
1222 } else if (minihiscore < miniscore) {
1223 if (hiscore->ifa)
1224 in6_ifa_put(hiscore->ifa);
1226 in6_ifa_hold(score->ifa);
1228 swap(hiscore, score);
1230 /* restore our iterator */
1231 score->ifa = hiscore->ifa;
1233 break;
1237 try_nextdev:
1238 read_unlock_bh(&idev->lock);
1240 rcu_read_unlock();
1241 read_unlock(&dev_base_lock);
1243 if (!hiscore->ifa)
1244 return -EADDRNOTAVAIL;
1246 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1247 in6_ifa_put(hiscore->ifa);
1248 return 0;
1251 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1253 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1254 unsigned char banned_flags)
1256 struct inet6_dev *idev;
1257 int err = -EADDRNOTAVAIL;
1259 rcu_read_lock();
1260 if ((idev = __in6_dev_get(dev)) != NULL) {
1261 struct inet6_ifaddr *ifp;
1263 read_lock_bh(&idev->lock);
1264 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1265 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1266 ipv6_addr_copy(addr, &ifp->addr);
1267 err = 0;
1268 break;
1271 read_unlock_bh(&idev->lock);
1273 rcu_read_unlock();
1274 return err;
1277 static int ipv6_count_addresses(struct inet6_dev *idev)
1279 int cnt = 0;
1280 struct inet6_ifaddr *ifp;
1282 read_lock_bh(&idev->lock);
1283 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1284 cnt++;
1285 read_unlock_bh(&idev->lock);
1286 return cnt;
1289 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1290 struct net_device *dev, int strict)
1292 struct inet6_ifaddr * ifp;
1293 u8 hash = ipv6_addr_hash(addr);
1295 read_lock_bh(&addrconf_hash_lock);
1296 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1297 if (!net_eq(dev_net(ifp->idev->dev), net))
1298 continue;
1299 if (ipv6_addr_equal(&ifp->addr, addr) &&
1300 !(ifp->flags&IFA_F_TENTATIVE)) {
1301 if (dev == NULL || ifp->idev->dev == dev ||
1302 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1303 break;
1306 read_unlock_bh(&addrconf_hash_lock);
1307 return ifp != NULL;
1309 EXPORT_SYMBOL(ipv6_chk_addr);
1311 static
1312 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1313 struct net_device *dev)
1315 struct inet6_ifaddr * ifp;
1316 u8 hash = ipv6_addr_hash(addr);
1318 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1319 if (!net_eq(dev_net(ifp->idev->dev), net))
1320 continue;
1321 if (ipv6_addr_equal(&ifp->addr, addr)) {
1322 if (dev == NULL || ifp->idev->dev == dev)
1323 break;
1326 return ifp != NULL;
1329 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1331 struct inet6_dev *idev;
1332 struct inet6_ifaddr *ifa;
1333 int onlink;
1335 onlink = 0;
1336 rcu_read_lock();
1337 idev = __in6_dev_get(dev);
1338 if (idev) {
1339 read_lock_bh(&idev->lock);
1340 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1341 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1342 ifa->prefix_len);
1343 if (onlink)
1344 break;
1346 read_unlock_bh(&idev->lock);
1348 rcu_read_unlock();
1349 return onlink;
1352 EXPORT_SYMBOL(ipv6_chk_prefix);
1354 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1355 struct net_device *dev, int strict)
1357 struct inet6_ifaddr * ifp;
1358 u8 hash = ipv6_addr_hash(addr);
1360 read_lock_bh(&addrconf_hash_lock);
1361 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1362 if (!net_eq(dev_net(ifp->idev->dev), net))
1363 continue;
1364 if (ipv6_addr_equal(&ifp->addr, addr)) {
1365 if (dev == NULL || ifp->idev->dev == dev ||
1366 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1367 in6_ifa_hold(ifp);
1368 break;
1372 read_unlock_bh(&addrconf_hash_lock);
1374 return ifp;
1377 /* Gets referenced address, destroys ifaddr */
1379 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1381 if (ifp->flags&IFA_F_PERMANENT) {
1382 spin_lock_bh(&ifp->lock);
1383 addrconf_del_timer(ifp);
1384 ifp->flags |= IFA_F_TENTATIVE;
1385 if (dad_failed)
1386 ifp->flags |= IFA_F_DADFAILED;
1387 spin_unlock_bh(&ifp->lock);
1388 in6_ifa_put(ifp);
1389 #ifdef CONFIG_IPV6_PRIVACY
1390 } else if (ifp->flags&IFA_F_TEMPORARY) {
1391 struct inet6_ifaddr *ifpub;
1392 spin_lock_bh(&ifp->lock);
1393 ifpub = ifp->ifpub;
1394 if (ifpub) {
1395 in6_ifa_hold(ifpub);
1396 spin_unlock_bh(&ifp->lock);
1397 ipv6_create_tempaddr(ifpub, ifp);
1398 in6_ifa_put(ifpub);
1399 } else {
1400 spin_unlock_bh(&ifp->lock);
1402 ipv6_del_addr(ifp);
1403 #endif
1404 } else
1405 ipv6_del_addr(ifp);
1408 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1410 struct inet6_dev *idev = ifp->idev;
1412 if (net_ratelimit())
1413 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1414 ifp->idev->dev->name, &ifp->addr);
1416 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1417 struct in6_addr addr;
1419 addr.s6_addr32[0] = htonl(0xfe800000);
1420 addr.s6_addr32[1] = 0;
1422 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1423 ipv6_addr_equal(&ifp->addr, &addr)) {
1424 /* DAD failed for link-local based on MAC address */
1425 idev->cnf.disable_ipv6 = 1;
1427 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1428 ifp->idev->dev->name);
1432 addrconf_dad_stop(ifp, 1);
1435 /* Join to solicited addr multicast group. */
1437 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1439 struct in6_addr maddr;
1441 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1442 return;
1444 addrconf_addr_solict_mult(addr, &maddr);
1445 ipv6_dev_mc_inc(dev, &maddr);
1448 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1450 struct in6_addr maddr;
1452 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453 return;
1455 addrconf_addr_solict_mult(addr, &maddr);
1456 __ipv6_dev_mc_dec(idev, &maddr);
1459 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1461 struct in6_addr addr;
1462 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1463 if (ipv6_addr_any(&addr))
1464 return;
1465 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1468 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1470 struct in6_addr addr;
1471 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1472 if (ipv6_addr_any(&addr))
1473 return;
1474 __ipv6_dev_ac_dec(ifp->idev, &addr);
1477 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1479 if (dev->addr_len != ETH_ALEN)
1480 return -1;
1481 memcpy(eui, dev->dev_addr, 3);
1482 memcpy(eui + 5, dev->dev_addr + 3, 3);
1485 * The zSeries OSA network cards can be shared among various
1486 * OS instances, but the OSA cards have only one MAC address.
1487 * This leads to duplicate address conflicts in conjunction
1488 * with IPv6 if more than one instance uses the same card.
1490 * The driver for these cards can deliver a unique 16-bit
1491 * identifier for each instance sharing the same card. It is
1492 * placed instead of 0xFFFE in the interface identifier. The
1493 * "u" bit of the interface identifier is not inverted in this
1494 * case. Hence the resulting interface identifier has local
1495 * scope according to RFC2373.
1497 if (dev->dev_id) {
1498 eui[3] = (dev->dev_id >> 8) & 0xFF;
1499 eui[4] = dev->dev_id & 0xFF;
1500 } else {
1501 eui[3] = 0xFF;
1502 eui[4] = 0xFE;
1503 eui[0] ^= 2;
1505 return 0;
1508 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1510 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1511 if (dev->addr_len != ARCNET_ALEN)
1512 return -1;
1513 memset(eui, 0, 7);
1514 eui[7] = *(u8*)dev->dev_addr;
1515 return 0;
1518 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1520 if (dev->addr_len != INFINIBAND_ALEN)
1521 return -1;
1522 memcpy(eui, dev->dev_addr + 12, 8);
1523 eui[0] |= 2;
1524 return 0;
1527 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1529 if (addr == 0)
1530 return -1;
1531 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1532 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1533 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1534 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1535 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1536 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1537 eui[1] = 0;
1538 eui[2] = 0x5E;
1539 eui[3] = 0xFE;
1540 memcpy(eui + 4, &addr, 4);
1541 return 0;
1543 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1545 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1547 if (dev->priv_flags & IFF_ISATAP)
1548 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1549 return -1;
1552 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1554 switch (dev->type) {
1555 case ARPHRD_ETHER:
1556 case ARPHRD_FDDI:
1557 case ARPHRD_IEEE802_TR:
1558 return addrconf_ifid_eui48(eui, dev);
1559 case ARPHRD_ARCNET:
1560 return addrconf_ifid_arcnet(eui, dev);
1561 case ARPHRD_INFINIBAND:
1562 return addrconf_ifid_infiniband(eui, dev);
1563 case ARPHRD_SIT:
1564 return addrconf_ifid_sit(eui, dev);
1566 return -1;
1569 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1571 int err = -1;
1572 struct inet6_ifaddr *ifp;
1574 read_lock_bh(&idev->lock);
1575 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1576 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1577 memcpy(eui, ifp->addr.s6_addr+8, 8);
1578 err = 0;
1579 break;
1582 read_unlock_bh(&idev->lock);
1583 return err;
1586 #ifdef CONFIG_IPV6_PRIVACY
1587 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1588 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1590 regen:
1591 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1592 idev->rndid[0] &= ~0x02;
1595 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1596 * check if generated address is not inappropriate
1598 * - Reserved subnet anycast (RFC 2526)
1599 * 11111101 11....11 1xxxxxxx
1600 * - ISATAP (RFC4214) 6.1
1601 * 00-00-5E-FE-xx-xx-xx-xx
1602 * - value 0
1603 * - XXX: already assigned to an address on the device
1605 if (idev->rndid[0] == 0xfd &&
1606 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1607 (idev->rndid[7]&0x80))
1608 goto regen;
1609 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1610 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1611 goto regen;
1612 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1613 goto regen;
1616 return 0;
1619 static void ipv6_regen_rndid(unsigned long data)
1621 struct inet6_dev *idev = (struct inet6_dev *) data;
1622 unsigned long expires;
1624 rcu_read_lock_bh();
1625 write_lock_bh(&idev->lock);
1627 if (idev->dead)
1628 goto out;
1630 if (__ipv6_regen_rndid(idev) < 0)
1631 goto out;
1633 expires = jiffies +
1634 idev->cnf.temp_prefered_lft * HZ -
1635 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1636 if (time_before(expires, jiffies)) {
1637 printk(KERN_WARNING
1638 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1639 idev->dev->name);
1640 goto out;
1643 if (!mod_timer(&idev->regen_timer, expires))
1644 in6_dev_hold(idev);
1646 out:
1647 write_unlock_bh(&idev->lock);
1648 rcu_read_unlock_bh();
1649 in6_dev_put(idev);
1652 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1653 int ret = 0;
1655 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1656 ret = __ipv6_regen_rndid(idev);
1657 return ret;
1659 #endif
1662 * Add prefix route.
1665 static void
1666 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1667 unsigned long expires, u32 flags)
1669 struct fib6_config cfg = {
1670 .fc_table = RT6_TABLE_PREFIX,
1671 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1672 .fc_ifindex = dev->ifindex,
1673 .fc_expires = expires,
1674 .fc_dst_len = plen,
1675 .fc_flags = RTF_UP | flags,
1676 .fc_nlinfo.nl_net = dev_net(dev),
1677 .fc_protocol = RTPROT_KERNEL,
1680 ipv6_addr_copy(&cfg.fc_dst, pfx);
1682 /* Prevent useless cloning on PtP SIT.
1683 This thing is done here expecting that the whole
1684 class of non-broadcast devices need not cloning.
1686 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1687 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1688 cfg.fc_flags |= RTF_NONEXTHOP;
1689 #endif
1691 ip6_route_add(&cfg);
1694 /* Create "default" multicast route to the interface */
1696 static void addrconf_add_mroute(struct net_device *dev)
1698 struct fib6_config cfg = {
1699 .fc_table = RT6_TABLE_LOCAL,
1700 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1701 .fc_ifindex = dev->ifindex,
1702 .fc_dst_len = 8,
1703 .fc_flags = RTF_UP,
1704 .fc_nlinfo.nl_net = dev_net(dev),
1707 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1709 ip6_route_add(&cfg);
1712 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1713 static void sit_route_add(struct net_device *dev)
1715 struct fib6_config cfg = {
1716 .fc_table = RT6_TABLE_MAIN,
1717 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1718 .fc_ifindex = dev->ifindex,
1719 .fc_dst_len = 96,
1720 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1721 .fc_nlinfo.nl_net = dev_net(dev),
1724 /* prefix length - 96 bits "::d.d.d.d" */
1725 ip6_route_add(&cfg);
1727 #endif
1729 static void addrconf_add_lroute(struct net_device *dev)
1731 struct in6_addr addr;
1733 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1734 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1737 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1739 struct inet6_dev *idev;
1741 ASSERT_RTNL();
1743 if ((idev = ipv6_find_idev(dev)) == NULL)
1744 return NULL;
1746 /* Add default multicast route */
1747 addrconf_add_mroute(dev);
1749 /* Add link local route */
1750 addrconf_add_lroute(dev);
1751 return idev;
1754 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1756 struct prefix_info *pinfo;
1757 __u32 valid_lft;
1758 __u32 prefered_lft;
1759 int addr_type;
1760 struct inet6_dev *in6_dev;
1761 struct net *net = dev_net(dev);
1763 pinfo = (struct prefix_info *) opt;
1765 if (len < sizeof(struct prefix_info)) {
1766 ADBG(("addrconf: prefix option too short\n"));
1767 return;
1771 * Validation checks ([ADDRCONF], page 19)
1774 addr_type = ipv6_addr_type(&pinfo->prefix);
1776 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1777 return;
1779 valid_lft = ntohl(pinfo->valid);
1780 prefered_lft = ntohl(pinfo->prefered);
1782 if (prefered_lft > valid_lft) {
1783 if (net_ratelimit())
1784 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1785 return;
1788 in6_dev = in6_dev_get(dev);
1790 if (in6_dev == NULL) {
1791 if (net_ratelimit())
1792 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1793 return;
1797 * Two things going on here:
1798 * 1) Add routes for on-link prefixes
1799 * 2) Configure prefixes with the auto flag set
1802 if (pinfo->onlink) {
1803 struct rt6_info *rt;
1804 unsigned long rt_expires;
1806 /* Avoid arithmetic overflow. Really, we could
1807 * save rt_expires in seconds, likely valid_lft,
1808 * but it would require division in fib gc, that it
1809 * not good.
1811 if (HZ > USER_HZ)
1812 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1813 else
1814 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1816 if (addrconf_finite_timeout(rt_expires))
1817 rt_expires *= HZ;
1819 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1820 dev->ifindex, 1);
1822 if (rt && addrconf_is_prefix_route(rt)) {
1823 /* Autoconf prefix route */
1824 if (valid_lft == 0) {
1825 ip6_del_rt(rt);
1826 rt = NULL;
1827 } else if (addrconf_finite_timeout(rt_expires)) {
1828 /* not infinity */
1829 rt->rt6i_expires = jiffies + rt_expires;
1830 rt->rt6i_flags |= RTF_EXPIRES;
1831 } else {
1832 rt->rt6i_flags &= ~RTF_EXPIRES;
1833 rt->rt6i_expires = 0;
1835 } else if (valid_lft) {
1836 clock_t expires = 0;
1837 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1838 if (addrconf_finite_timeout(rt_expires)) {
1839 /* not infinity */
1840 flags |= RTF_EXPIRES;
1841 expires = jiffies_to_clock_t(rt_expires);
1843 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1844 dev, expires, flags);
1846 if (rt)
1847 dst_release(&rt->u.dst);
1850 /* Try to figure out our local address for this prefix */
1852 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1853 struct inet6_ifaddr * ifp;
1854 struct in6_addr addr;
1855 int create = 0, update_lft = 0;
1857 if (pinfo->prefix_len == 64) {
1858 memcpy(&addr, &pinfo->prefix, 8);
1859 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1860 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1861 in6_dev_put(in6_dev);
1862 return;
1864 goto ok;
1866 if (net_ratelimit())
1867 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1868 pinfo->prefix_len);
1869 in6_dev_put(in6_dev);
1870 return;
1874 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1876 if (ifp == NULL && valid_lft) {
1877 int max_addresses = in6_dev->cnf.max_addresses;
1878 u32 addr_flags = 0;
1880 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1881 if (in6_dev->cnf.optimistic_dad &&
1882 !net->ipv6.devconf_all->forwarding)
1883 addr_flags = IFA_F_OPTIMISTIC;
1884 #endif
1886 /* Do not allow to create too much of autoconfigured
1887 * addresses; this would be too easy way to crash kernel.
1889 if (!max_addresses ||
1890 ipv6_count_addresses(in6_dev) < max_addresses)
1891 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1892 addr_type&IPV6_ADDR_SCOPE_MASK,
1893 addr_flags);
1895 if (!ifp || IS_ERR(ifp)) {
1896 in6_dev_put(in6_dev);
1897 return;
1900 update_lft = create = 1;
1901 ifp->cstamp = jiffies;
1902 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1905 if (ifp) {
1906 int flags;
1907 unsigned long now;
1908 #ifdef CONFIG_IPV6_PRIVACY
1909 struct inet6_ifaddr *ift;
1910 #endif
1911 u32 stored_lft;
1913 /* update lifetime (RFC2462 5.5.3 e) */
1914 spin_lock(&ifp->lock);
1915 now = jiffies;
1916 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1917 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1918 else
1919 stored_lft = 0;
1920 if (!update_lft && stored_lft) {
1921 if (valid_lft > MIN_VALID_LIFETIME ||
1922 valid_lft > stored_lft)
1923 update_lft = 1;
1924 else if (stored_lft <= MIN_VALID_LIFETIME) {
1925 /* valid_lft <= stored_lft is always true */
1927 * RFC 4862 Section 5.5.3e:
1928 * "Note that the preferred lifetime of
1929 * the corresponding address is always
1930 * reset to the Preferred Lifetime in
1931 * the received Prefix Information
1932 * option, regardless of whether the
1933 * valid lifetime is also reset or
1934 * ignored."
1936 * So if the preferred lifetime in
1937 * this advertisement is different
1938 * than what we have stored, but the
1939 * valid lifetime is invalid, just
1940 * reset prefered_lft.
1942 * We must set the valid lifetime
1943 * to the stored lifetime since we'll
1944 * be updating the timestamp below,
1945 * else we'll set it back to the
1946 * minumum.
1948 if (prefered_lft != ifp->prefered_lft) {
1949 valid_lft = stored_lft;
1950 update_lft = 1;
1952 } else {
1953 valid_lft = MIN_VALID_LIFETIME;
1954 if (valid_lft < prefered_lft)
1955 prefered_lft = valid_lft;
1956 update_lft = 1;
1960 if (update_lft) {
1961 ifp->valid_lft = valid_lft;
1962 ifp->prefered_lft = prefered_lft;
1963 ifp->tstamp = now;
1964 flags = ifp->flags;
1965 ifp->flags &= ~IFA_F_DEPRECATED;
1966 spin_unlock(&ifp->lock);
1968 if (!(flags&IFA_F_TENTATIVE))
1969 ipv6_ifa_notify(0, ifp);
1970 } else
1971 spin_unlock(&ifp->lock);
1973 #ifdef CONFIG_IPV6_PRIVACY
1974 read_lock_bh(&in6_dev->lock);
1975 /* update all temporary addresses in the list */
1976 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1978 * When adjusting the lifetimes of an existing
1979 * temporary address, only lower the lifetimes.
1980 * Implementations must not increase the
1981 * lifetimes of an existing temporary address
1982 * when processing a Prefix Information Option.
1984 if (ifp != ift->ifpub)
1985 continue;
1987 spin_lock(&ift->lock);
1988 flags = ift->flags;
1989 if (ift->valid_lft > valid_lft &&
1990 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1991 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1992 if (ift->prefered_lft > prefered_lft &&
1993 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1994 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1995 spin_unlock(&ift->lock);
1996 if (!(flags&IFA_F_TENTATIVE))
1997 ipv6_ifa_notify(0, ift);
2000 if (create && in6_dev->cnf.use_tempaddr > 0) {
2002 * When a new public address is created as described in [ADDRCONF],
2003 * also create a new temporary address.
2005 read_unlock_bh(&in6_dev->lock);
2006 ipv6_create_tempaddr(ifp, NULL);
2007 } else {
2008 read_unlock_bh(&in6_dev->lock);
2010 #endif
2011 in6_ifa_put(ifp);
2012 addrconf_verify(0);
2015 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2016 in6_dev_put(in6_dev);
2020 * Set destination address.
2021 * Special case for SIT interfaces where we create a new "virtual"
2022 * device.
2024 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2026 struct in6_ifreq ireq;
2027 struct net_device *dev;
2028 int err = -EINVAL;
2030 rtnl_lock();
2032 err = -EFAULT;
2033 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2034 goto err_exit;
2036 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2038 err = -ENODEV;
2039 if (dev == NULL)
2040 goto err_exit;
2042 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2043 if (dev->type == ARPHRD_SIT) {
2044 const struct net_device_ops *ops = dev->netdev_ops;
2045 struct ifreq ifr;
2046 struct ip_tunnel_parm p;
2048 err = -EADDRNOTAVAIL;
2049 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2050 goto err_exit;
2052 memset(&p, 0, sizeof(p));
2053 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2054 p.iph.saddr = 0;
2055 p.iph.version = 4;
2056 p.iph.ihl = 5;
2057 p.iph.protocol = IPPROTO_IPV6;
2058 p.iph.ttl = 64;
2059 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2061 if (ops->ndo_do_ioctl) {
2062 mm_segment_t oldfs = get_fs();
2064 set_fs(KERNEL_DS);
2065 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2066 set_fs(oldfs);
2067 } else
2068 err = -EOPNOTSUPP;
2070 if (err == 0) {
2071 err = -ENOBUFS;
2072 dev = __dev_get_by_name(net, p.name);
2073 if (!dev)
2074 goto err_exit;
2075 err = dev_open(dev);
2078 #endif
2080 err_exit:
2081 rtnl_unlock();
2082 return err;
2086 * Manual configuration of address on an interface
2088 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2089 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2090 __u32 valid_lft)
2092 struct inet6_ifaddr *ifp;
2093 struct inet6_dev *idev;
2094 struct net_device *dev;
2095 int scope;
2096 u32 flags;
2097 clock_t expires;
2098 unsigned long timeout;
2100 ASSERT_RTNL();
2102 if (plen > 128)
2103 return -EINVAL;
2105 /* check the lifetime */
2106 if (!valid_lft || prefered_lft > valid_lft)
2107 return -EINVAL;
2109 dev = __dev_get_by_index(net, ifindex);
2110 if (!dev)
2111 return -ENODEV;
2113 if ((idev = addrconf_add_dev(dev)) == NULL)
2114 return -ENOBUFS;
2116 scope = ipv6_addr_scope(pfx);
2118 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2119 if (addrconf_finite_timeout(timeout)) {
2120 expires = jiffies_to_clock_t(timeout * HZ);
2121 valid_lft = timeout;
2122 flags = RTF_EXPIRES;
2123 } else {
2124 expires = 0;
2125 flags = 0;
2126 ifa_flags |= IFA_F_PERMANENT;
2129 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2130 if (addrconf_finite_timeout(timeout)) {
2131 if (timeout == 0)
2132 ifa_flags |= IFA_F_DEPRECATED;
2133 prefered_lft = timeout;
2136 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2138 if (!IS_ERR(ifp)) {
2139 spin_lock_bh(&ifp->lock);
2140 ifp->valid_lft = valid_lft;
2141 ifp->prefered_lft = prefered_lft;
2142 ifp->tstamp = jiffies;
2143 spin_unlock_bh(&ifp->lock);
2145 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2146 expires, flags);
2148 * Note that section 3.1 of RFC 4429 indicates
2149 * that the Optimistic flag should not be set for
2150 * manually configured addresses
2152 addrconf_dad_start(ifp, 0);
2153 in6_ifa_put(ifp);
2154 addrconf_verify(0);
2155 return 0;
2158 return PTR_ERR(ifp);
2161 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2162 unsigned int plen)
2164 struct inet6_ifaddr *ifp;
2165 struct inet6_dev *idev;
2166 struct net_device *dev;
2168 if (plen > 128)
2169 return -EINVAL;
2171 dev = __dev_get_by_index(net, ifindex);
2172 if (!dev)
2173 return -ENODEV;
2175 if ((idev = __in6_dev_get(dev)) == NULL)
2176 return -ENXIO;
2178 read_lock_bh(&idev->lock);
2179 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2180 if (ifp->prefix_len == plen &&
2181 ipv6_addr_equal(pfx, &ifp->addr)) {
2182 in6_ifa_hold(ifp);
2183 read_unlock_bh(&idev->lock);
2185 ipv6_del_addr(ifp);
2187 /* If the last address is deleted administratively,
2188 disable IPv6 on this interface.
2190 if (idev->addr_list == NULL)
2191 addrconf_ifdown(idev->dev, 1);
2192 return 0;
2195 read_unlock_bh(&idev->lock);
2196 return -EADDRNOTAVAIL;
2200 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2202 struct in6_ifreq ireq;
2203 int err;
2205 if (!capable(CAP_NET_ADMIN))
2206 return -EPERM;
2208 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2209 return -EFAULT;
2211 rtnl_lock();
2212 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2213 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2214 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2215 rtnl_unlock();
2216 return err;
2219 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2221 struct in6_ifreq ireq;
2222 int err;
2224 if (!capable(CAP_NET_ADMIN))
2225 return -EPERM;
2227 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2228 return -EFAULT;
2230 rtnl_lock();
2231 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2232 ireq.ifr6_prefixlen);
2233 rtnl_unlock();
2234 return err;
2237 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2238 int plen, int scope)
2240 struct inet6_ifaddr *ifp;
2242 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2243 if (!IS_ERR(ifp)) {
2244 spin_lock_bh(&ifp->lock);
2245 ifp->flags &= ~IFA_F_TENTATIVE;
2246 spin_unlock_bh(&ifp->lock);
2247 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2248 in6_ifa_put(ifp);
2252 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2253 static void sit_add_v4_addrs(struct inet6_dev *idev)
2255 struct in6_addr addr;
2256 struct net_device *dev;
2257 struct net *net = dev_net(idev->dev);
2258 int scope;
2260 ASSERT_RTNL();
2262 memset(&addr, 0, sizeof(struct in6_addr));
2263 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2265 if (idev->dev->flags&IFF_POINTOPOINT) {
2266 addr.s6_addr32[0] = htonl(0xfe800000);
2267 scope = IFA_LINK;
2268 } else {
2269 scope = IPV6_ADDR_COMPATv4;
2272 if (addr.s6_addr32[3]) {
2273 add_addr(idev, &addr, 128, scope);
2274 return;
2277 for_each_netdev(net, dev) {
2278 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2279 if (in_dev && (dev->flags & IFF_UP)) {
2280 struct in_ifaddr * ifa;
2282 int flag = scope;
2284 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2285 int plen;
2287 addr.s6_addr32[3] = ifa->ifa_local;
2289 if (ifa->ifa_scope == RT_SCOPE_LINK)
2290 continue;
2291 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2292 if (idev->dev->flags&IFF_POINTOPOINT)
2293 continue;
2294 flag |= IFA_HOST;
2296 if (idev->dev->flags&IFF_POINTOPOINT)
2297 plen = 64;
2298 else
2299 plen = 96;
2301 add_addr(idev, &addr, plen, flag);
2306 #endif
2308 static void init_loopback(struct net_device *dev)
2310 struct inet6_dev *idev;
2312 /* ::1 */
2314 ASSERT_RTNL();
2316 if ((idev = ipv6_find_idev(dev)) == NULL) {
2317 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2318 return;
2321 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2324 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2326 struct inet6_ifaddr * ifp;
2327 u32 addr_flags = IFA_F_PERMANENT;
2329 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2330 if (idev->cnf.optimistic_dad &&
2331 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2332 addr_flags |= IFA_F_OPTIMISTIC;
2333 #endif
2336 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2337 if (!IS_ERR(ifp)) {
2338 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2339 addrconf_dad_start(ifp, 0);
2340 in6_ifa_put(ifp);
2344 static void addrconf_dev_config(struct net_device *dev)
2346 struct in6_addr addr;
2347 struct inet6_dev * idev;
2349 ASSERT_RTNL();
2351 if ((dev->type != ARPHRD_ETHER) &&
2352 (dev->type != ARPHRD_FDDI) &&
2353 (dev->type != ARPHRD_IEEE802_TR) &&
2354 (dev->type != ARPHRD_ARCNET) &&
2355 (dev->type != ARPHRD_INFINIBAND)) {
2356 /* Alas, we support only Ethernet autoconfiguration. */
2357 return;
2360 idev = addrconf_add_dev(dev);
2361 if (idev == NULL)
2362 return;
2364 memset(&addr, 0, sizeof(struct in6_addr));
2365 addr.s6_addr32[0] = htonl(0xFE800000);
2367 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2368 addrconf_add_linklocal(idev, &addr);
2371 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2372 static void addrconf_sit_config(struct net_device *dev)
2374 struct inet6_dev *idev;
2376 ASSERT_RTNL();
2379 * Configure the tunnel with one of our IPv4
2380 * addresses... we should configure all of
2381 * our v4 addrs in the tunnel
2384 if ((idev = ipv6_find_idev(dev)) == NULL) {
2385 printk(KERN_DEBUG "init sit: add_dev failed\n");
2386 return;
2389 if (dev->priv_flags & IFF_ISATAP) {
2390 struct in6_addr addr;
2392 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2393 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2394 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2395 addrconf_add_linklocal(idev, &addr);
2396 return;
2399 sit_add_v4_addrs(idev);
2401 if (dev->flags&IFF_POINTOPOINT) {
2402 addrconf_add_mroute(dev);
2403 addrconf_add_lroute(dev);
2404 } else
2405 sit_route_add(dev);
2407 #endif
2409 static inline int
2410 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2412 struct in6_addr lladdr;
2414 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2415 addrconf_add_linklocal(idev, &lladdr);
2416 return 0;
2418 return -1;
2421 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2423 struct net_device *link_dev;
2424 struct net *net = dev_net(idev->dev);
2426 /* first try to inherit the link-local address from the link device */
2427 if (idev->dev->iflink &&
2428 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2429 if (!ipv6_inherit_linklocal(idev, link_dev))
2430 return;
2432 /* then try to inherit it from any device */
2433 for_each_netdev(net, link_dev) {
2434 if (!ipv6_inherit_linklocal(idev, link_dev))
2435 return;
2437 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2441 * Autoconfigure tunnel with a link-local address so routing protocols,
2442 * DHCPv6, MLD etc. can be run over the virtual link
2445 static void addrconf_ip6_tnl_config(struct net_device *dev)
2447 struct inet6_dev *idev;
2449 ASSERT_RTNL();
2451 if ((idev = addrconf_add_dev(dev)) == NULL) {
2452 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2453 return;
2455 ip6_tnl_add_linklocal(idev);
2458 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2459 void * data)
2461 struct net_device *dev = (struct net_device *) data;
2462 struct inet6_dev *idev = __in6_dev_get(dev);
2463 int run_pending = 0;
2464 int err;
2466 switch(event) {
2467 case NETDEV_REGISTER:
2468 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2469 idev = ipv6_add_dev(dev);
2470 if (!idev)
2471 return notifier_from_errno(-ENOMEM);
2473 break;
2474 case NETDEV_UP:
2475 case NETDEV_CHANGE:
2476 if (dev->flags & IFF_SLAVE)
2477 break;
2479 if (event == NETDEV_UP) {
2480 if (!addrconf_qdisc_ok(dev)) {
2481 /* device is not ready yet. */
2482 printk(KERN_INFO
2483 "ADDRCONF(NETDEV_UP): %s: "
2484 "link is not ready\n",
2485 dev->name);
2486 break;
2489 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2490 idev = ipv6_add_dev(dev);
2492 if (idev) {
2493 idev->if_flags |= IF_READY;
2494 run_pending = 1;
2496 } else {
2497 if (!addrconf_qdisc_ok(dev)) {
2498 /* device is still not ready. */
2499 break;
2502 if (idev) {
2503 if (idev->if_flags & IF_READY) {
2504 /* device is already configured. */
2505 break;
2507 idev->if_flags |= IF_READY;
2510 printk(KERN_INFO
2511 "ADDRCONF(NETDEV_CHANGE): %s: "
2512 "link becomes ready\n",
2513 dev->name);
2515 run_pending = 1;
2518 switch(dev->type) {
2519 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2520 case ARPHRD_SIT:
2521 addrconf_sit_config(dev);
2522 break;
2523 #endif
2524 case ARPHRD_TUNNEL6:
2525 addrconf_ip6_tnl_config(dev);
2526 break;
2527 case ARPHRD_LOOPBACK:
2528 init_loopback(dev);
2529 break;
2531 default:
2532 addrconf_dev_config(dev);
2533 break;
2535 if (idev) {
2536 if (run_pending)
2537 addrconf_dad_run(idev);
2539 /* If the MTU changed during the interface down, when the
2540 interface up, the changed MTU must be reflected in the
2541 idev as well as routers.
2543 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2544 rt6_mtu_change(dev, dev->mtu);
2545 idev->cnf.mtu6 = dev->mtu;
2547 idev->tstamp = jiffies;
2548 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2549 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2550 stop IPv6 on this interface.
2552 if (dev->mtu < IPV6_MIN_MTU)
2553 addrconf_ifdown(dev, event != NETDEV_DOWN);
2555 break;
2557 case NETDEV_CHANGEMTU:
2558 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2559 rt6_mtu_change(dev, dev->mtu);
2560 idev->cnf.mtu6 = dev->mtu;
2561 break;
2564 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2565 idev = ipv6_add_dev(dev);
2566 if (idev)
2567 break;
2570 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2572 case NETDEV_DOWN:
2573 case NETDEV_UNREGISTER:
2575 * Remove all addresses from this interface.
2577 addrconf_ifdown(dev, event != NETDEV_DOWN);
2578 break;
2580 case NETDEV_CHANGENAME:
2581 if (idev) {
2582 snmp6_unregister_dev(idev);
2583 addrconf_sysctl_unregister(idev);
2584 addrconf_sysctl_register(idev);
2585 err = snmp6_register_dev(idev);
2586 if (err)
2587 return notifier_from_errno(err);
2589 break;
2590 case NETDEV_BONDING_OLDTYPE:
2591 case NETDEV_BONDING_NEWTYPE:
2592 addrconf_bonding_change(dev, event);
2593 break;
2596 return NOTIFY_OK;
2600 * addrconf module should be notified of a device going up
2602 static struct notifier_block ipv6_dev_notf = {
2603 .notifier_call = addrconf_notify,
2604 .priority = 0
2607 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2609 struct inet6_dev *idev;
2610 ASSERT_RTNL();
2612 idev = __in6_dev_get(dev);
2614 if (event == NETDEV_BONDING_NEWTYPE)
2615 ipv6_mc_remap(idev);
2616 else if (event == NETDEV_BONDING_OLDTYPE)
2617 ipv6_mc_unmap(idev);
2620 static int addrconf_ifdown(struct net_device *dev, int how)
2622 struct inet6_dev *idev;
2623 struct inet6_ifaddr *ifa, **bifa;
2624 struct net *net = dev_net(dev);
2625 int i;
2627 ASSERT_RTNL();
2629 rt6_ifdown(net, dev);
2630 neigh_ifdown(&nd_tbl, dev);
2632 idev = __in6_dev_get(dev);
2633 if (idev == NULL)
2634 return -ENODEV;
2636 /* Step 1: remove reference to ipv6 device from parent device.
2637 Do not dev_put!
2639 if (how) {
2640 idev->dead = 1;
2642 /* protected by rtnl_lock */
2643 rcu_assign_pointer(dev->ip6_ptr, NULL);
2645 /* Step 1.5: remove snmp6 entry */
2646 snmp6_unregister_dev(idev);
2650 /* Step 2: clear hash table */
2651 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2652 bifa = &inet6_addr_lst[i];
2654 write_lock_bh(&addrconf_hash_lock);
2655 while ((ifa = *bifa) != NULL) {
2656 if (ifa->idev == idev) {
2657 *bifa = ifa->lst_next;
2658 ifa->lst_next = NULL;
2659 addrconf_del_timer(ifa);
2660 in6_ifa_put(ifa);
2661 continue;
2663 bifa = &ifa->lst_next;
2665 write_unlock_bh(&addrconf_hash_lock);
2668 write_lock_bh(&idev->lock);
2670 /* Step 3: clear flags for stateless addrconf */
2671 if (!how)
2672 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2674 /* Step 4: clear address list */
2675 #ifdef CONFIG_IPV6_PRIVACY
2676 if (how && del_timer(&idev->regen_timer))
2677 in6_dev_put(idev);
2679 /* clear tempaddr list */
2680 while ((ifa = idev->tempaddr_list) != NULL) {
2681 idev->tempaddr_list = ifa->tmp_next;
2682 ifa->tmp_next = NULL;
2683 ifa->dead = 1;
2684 write_unlock_bh(&idev->lock);
2685 spin_lock_bh(&ifa->lock);
2687 if (ifa->ifpub) {
2688 in6_ifa_put(ifa->ifpub);
2689 ifa->ifpub = NULL;
2691 spin_unlock_bh(&ifa->lock);
2692 in6_ifa_put(ifa);
2693 write_lock_bh(&idev->lock);
2695 #endif
2696 while ((ifa = idev->addr_list) != NULL) {
2697 idev->addr_list = ifa->if_next;
2698 ifa->if_next = NULL;
2699 ifa->dead = 1;
2700 addrconf_del_timer(ifa);
2701 write_unlock_bh(&idev->lock);
2703 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2704 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2705 in6_ifa_put(ifa);
2707 write_lock_bh(&idev->lock);
2709 write_unlock_bh(&idev->lock);
2711 /* Step 5: Discard multicast list */
2713 if (how)
2714 ipv6_mc_destroy_dev(idev);
2715 else
2716 ipv6_mc_down(idev);
2718 idev->tstamp = jiffies;
2720 /* Shot the device (if unregistered) */
2722 if (how) {
2723 addrconf_sysctl_unregister(idev);
2724 neigh_parms_release(&nd_tbl, idev->nd_parms);
2725 neigh_ifdown(&nd_tbl, dev);
2726 in6_dev_put(idev);
2728 return 0;
2731 static void addrconf_rs_timer(unsigned long data)
2733 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2735 if (ifp->idev->cnf.forwarding)
2736 goto out;
2738 if (ifp->idev->if_flags & IF_RA_RCVD) {
2740 * Announcement received after solicitation
2741 * was sent
2743 goto out;
2746 spin_lock(&ifp->lock);
2747 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2748 /* The wait after the last probe can be shorter */
2749 addrconf_mod_timer(ifp, AC_RS,
2750 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2751 ifp->idev->cnf.rtr_solicit_delay :
2752 ifp->idev->cnf.rtr_solicit_interval);
2753 spin_unlock(&ifp->lock);
2755 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2756 } else {
2757 spin_unlock(&ifp->lock);
2759 * Note: we do not support deprecated "all on-link"
2760 * assumption any longer.
2762 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2763 ifp->idev->dev->name);
2766 out:
2767 in6_ifa_put(ifp);
2771 * Duplicate Address Detection
2773 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2775 unsigned long rand_num;
2776 struct inet6_dev *idev = ifp->idev;
2778 if (ifp->flags & IFA_F_OPTIMISTIC)
2779 rand_num = 0;
2780 else
2781 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2783 ifp->probes = idev->cnf.dad_transmits;
2784 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2787 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2789 struct inet6_dev *idev = ifp->idev;
2790 struct net_device *dev = idev->dev;
2792 addrconf_join_solict(dev, &ifp->addr);
2794 net_srandom(ifp->addr.s6_addr32[3]);
2796 read_lock_bh(&idev->lock);
2797 if (ifp->dead)
2798 goto out;
2799 spin_lock_bh(&ifp->lock);
2801 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2802 idev->cnf.accept_dad < 1 ||
2803 !(ifp->flags&IFA_F_TENTATIVE) ||
2804 ifp->flags & IFA_F_NODAD) {
2805 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2806 spin_unlock_bh(&ifp->lock);
2807 read_unlock_bh(&idev->lock);
2809 addrconf_dad_completed(ifp);
2810 return;
2813 if (!(idev->if_flags & IF_READY)) {
2814 spin_unlock_bh(&ifp->lock);
2815 read_unlock_bh(&idev->lock);
2817 * If the device is not ready:
2818 * - keep it tentative if it is a permanent address.
2819 * - otherwise, kill it.
2821 in6_ifa_hold(ifp);
2822 addrconf_dad_stop(ifp, 0);
2823 return;
2827 * Optimistic nodes can start receiving
2828 * Frames right away
2830 if(ifp->flags & IFA_F_OPTIMISTIC)
2831 ip6_ins_rt(ifp->rt);
2833 addrconf_dad_kick(ifp);
2834 spin_unlock_bh(&ifp->lock);
2835 out:
2836 read_unlock_bh(&idev->lock);
2839 static void addrconf_dad_timer(unsigned long data)
2841 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2842 struct inet6_dev *idev = ifp->idev;
2843 struct in6_addr mcaddr;
2845 read_lock_bh(&idev->lock);
2846 if (idev->dead) {
2847 read_unlock_bh(&idev->lock);
2848 goto out;
2850 spin_lock_bh(&ifp->lock);
2851 if (ifp->probes == 0) {
2853 * DAD was successful
2856 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2857 spin_unlock_bh(&ifp->lock);
2858 read_unlock_bh(&idev->lock);
2860 addrconf_dad_completed(ifp);
2862 goto out;
2865 ifp->probes--;
2866 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2867 spin_unlock_bh(&ifp->lock);
2868 read_unlock_bh(&idev->lock);
2870 /* send a neighbour solicitation for our addr */
2871 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2872 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2873 out:
2874 in6_ifa_put(ifp);
2877 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2879 struct net_device * dev = ifp->idev->dev;
2882 * Configure the address for reception. Now it is valid.
2885 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2887 /* If added prefix is link local and forwarding is off,
2888 start sending router solicitations.
2891 if (ifp->idev->cnf.forwarding == 0 &&
2892 ifp->idev->cnf.rtr_solicits > 0 &&
2893 (dev->flags&IFF_LOOPBACK) == 0 &&
2894 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2896 * If a host as already performed a random delay
2897 * [...] as part of DAD [...] there is no need
2898 * to delay again before sending the first RS
2900 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2902 spin_lock_bh(&ifp->lock);
2903 ifp->probes = 1;
2904 ifp->idev->if_flags |= IF_RS_SENT;
2905 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2906 spin_unlock_bh(&ifp->lock);
2910 static void addrconf_dad_run(struct inet6_dev *idev) {
2911 struct inet6_ifaddr *ifp;
2913 read_lock_bh(&idev->lock);
2914 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2915 spin_lock_bh(&ifp->lock);
2916 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2917 spin_unlock_bh(&ifp->lock);
2918 continue;
2920 spin_unlock_bh(&ifp->lock);
2921 addrconf_dad_kick(ifp);
2923 read_unlock_bh(&idev->lock);
2926 #ifdef CONFIG_PROC_FS
2927 struct if6_iter_state {
2928 struct seq_net_private p;
2929 int bucket;
2932 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2934 struct inet6_ifaddr *ifa = NULL;
2935 struct if6_iter_state *state = seq->private;
2936 struct net *net = seq_file_net(seq);
2938 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2939 ifa = inet6_addr_lst[state->bucket];
2941 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2942 ifa = ifa->lst_next;
2943 if (ifa)
2944 break;
2946 return ifa;
2949 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2951 struct if6_iter_state *state = seq->private;
2952 struct net *net = seq_file_net(seq);
2954 ifa = ifa->lst_next;
2955 try_again:
2956 if (ifa) {
2957 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2958 ifa = ifa->lst_next;
2959 goto try_again;
2963 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2964 ifa = inet6_addr_lst[state->bucket];
2965 goto try_again;
2968 return ifa;
2971 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2973 struct inet6_ifaddr *ifa = if6_get_first(seq);
2975 if (ifa)
2976 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2977 --pos;
2978 return pos ? NULL : ifa;
2981 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2982 __acquires(addrconf_hash_lock)
2984 read_lock_bh(&addrconf_hash_lock);
2985 return if6_get_idx(seq, *pos);
2988 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2990 struct inet6_ifaddr *ifa;
2992 ifa = if6_get_next(seq, v);
2993 ++*pos;
2994 return ifa;
2997 static void if6_seq_stop(struct seq_file *seq, void *v)
2998 __releases(addrconf_hash_lock)
3000 read_unlock_bh(&addrconf_hash_lock);
3003 static int if6_seq_show(struct seq_file *seq, void *v)
3005 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3006 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3007 &ifp->addr,
3008 ifp->idev->dev->ifindex,
3009 ifp->prefix_len,
3010 ifp->scope,
3011 ifp->flags,
3012 ifp->idev->dev->name);
3013 return 0;
3016 static const struct seq_operations if6_seq_ops = {
3017 .start = if6_seq_start,
3018 .next = if6_seq_next,
3019 .show = if6_seq_show,
3020 .stop = if6_seq_stop,
3023 static int if6_seq_open(struct inode *inode, struct file *file)
3025 return seq_open_net(inode, file, &if6_seq_ops,
3026 sizeof(struct if6_iter_state));
3029 static const struct file_operations if6_fops = {
3030 .owner = THIS_MODULE,
3031 .open = if6_seq_open,
3032 .read = seq_read,
3033 .llseek = seq_lseek,
3034 .release = seq_release_net,
3037 static int if6_proc_net_init(struct net *net)
3039 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3040 return -ENOMEM;
3041 return 0;
3044 static void if6_proc_net_exit(struct net *net)
3046 proc_net_remove(net, "if_inet6");
3049 static struct pernet_operations if6_proc_net_ops = {
3050 .init = if6_proc_net_init,
3051 .exit = if6_proc_net_exit,
3054 int __init if6_proc_init(void)
3056 return register_pernet_subsys(&if6_proc_net_ops);
3059 void if6_proc_exit(void)
3061 unregister_pernet_subsys(&if6_proc_net_ops);
3063 #endif /* CONFIG_PROC_FS */
3065 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3066 /* Check if address is a home address configured on any interface. */
3067 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3069 int ret = 0;
3070 struct inet6_ifaddr * ifp;
3071 u8 hash = ipv6_addr_hash(addr);
3072 read_lock_bh(&addrconf_hash_lock);
3073 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3074 if (!net_eq(dev_net(ifp->idev->dev), net))
3075 continue;
3076 if (ipv6_addr_equal(&ifp->addr, addr) &&
3077 (ifp->flags & IFA_F_HOMEADDRESS)) {
3078 ret = 1;
3079 break;
3082 read_unlock_bh(&addrconf_hash_lock);
3083 return ret;
3085 #endif
3088 * Periodic address status verification
3091 static void addrconf_verify(unsigned long foo)
3093 struct inet6_ifaddr *ifp;
3094 unsigned long now, next;
3095 int i;
3097 spin_lock_bh(&addrconf_verify_lock);
3098 now = jiffies;
3099 next = now + ADDR_CHECK_FREQUENCY;
3101 del_timer(&addr_chk_timer);
3103 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3105 restart:
3106 read_lock(&addrconf_hash_lock);
3107 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3108 unsigned long age;
3109 #ifdef CONFIG_IPV6_PRIVACY
3110 unsigned long regen_advance;
3111 #endif
3113 if (ifp->flags & IFA_F_PERMANENT)
3114 continue;
3116 spin_lock(&ifp->lock);
3117 age = (now - ifp->tstamp) / HZ;
3119 #ifdef CONFIG_IPV6_PRIVACY
3120 regen_advance = ifp->idev->cnf.regen_max_retry *
3121 ifp->idev->cnf.dad_transmits *
3122 ifp->idev->nd_parms->retrans_time / HZ;
3123 #endif
3125 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3126 age >= ifp->valid_lft) {
3127 spin_unlock(&ifp->lock);
3128 in6_ifa_hold(ifp);
3129 read_unlock(&addrconf_hash_lock);
3130 ipv6_del_addr(ifp);
3131 goto restart;
3132 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3133 spin_unlock(&ifp->lock);
3134 continue;
3135 } else if (age >= ifp->prefered_lft) {
3136 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3137 int deprecate = 0;
3139 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3140 deprecate = 1;
3141 ifp->flags |= IFA_F_DEPRECATED;
3144 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3145 next = ifp->tstamp + ifp->valid_lft * HZ;
3147 spin_unlock(&ifp->lock);
3149 if (deprecate) {
3150 in6_ifa_hold(ifp);
3151 read_unlock(&addrconf_hash_lock);
3153 ipv6_ifa_notify(0, ifp);
3154 in6_ifa_put(ifp);
3155 goto restart;
3157 #ifdef CONFIG_IPV6_PRIVACY
3158 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3159 !(ifp->flags&IFA_F_TENTATIVE)) {
3160 if (age >= ifp->prefered_lft - regen_advance) {
3161 struct inet6_ifaddr *ifpub = ifp->ifpub;
3162 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3163 next = ifp->tstamp + ifp->prefered_lft * HZ;
3164 if (!ifp->regen_count && ifpub) {
3165 ifp->regen_count++;
3166 in6_ifa_hold(ifp);
3167 in6_ifa_hold(ifpub);
3168 spin_unlock(&ifp->lock);
3169 read_unlock(&addrconf_hash_lock);
3170 spin_lock(&ifpub->lock);
3171 ifpub->regen_count = 0;
3172 spin_unlock(&ifpub->lock);
3173 ipv6_create_tempaddr(ifpub, ifp);
3174 in6_ifa_put(ifpub);
3175 in6_ifa_put(ifp);
3176 goto restart;
3178 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3179 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3180 spin_unlock(&ifp->lock);
3181 #endif
3182 } else {
3183 /* ifp->prefered_lft <= ifp->valid_lft */
3184 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3185 next = ifp->tstamp + ifp->prefered_lft * HZ;
3186 spin_unlock(&ifp->lock);
3189 read_unlock(&addrconf_hash_lock);
3192 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3193 add_timer(&addr_chk_timer);
3194 spin_unlock_bh(&addrconf_verify_lock);
3197 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3199 struct in6_addr *pfx = NULL;
3201 if (addr)
3202 pfx = nla_data(addr);
3204 if (local) {
3205 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3206 pfx = NULL;
3207 else
3208 pfx = nla_data(local);
3211 return pfx;
3214 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3215 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3216 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3217 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3220 static int
3221 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3223 struct net *net = sock_net(skb->sk);
3224 struct ifaddrmsg *ifm;
3225 struct nlattr *tb[IFA_MAX+1];
3226 struct in6_addr *pfx;
3227 int err;
3229 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3230 if (err < 0)
3231 return err;
3233 ifm = nlmsg_data(nlh);
3234 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3235 if (pfx == NULL)
3236 return -EINVAL;
3238 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3241 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3242 u32 prefered_lft, u32 valid_lft)
3244 u32 flags;
3245 clock_t expires;
3246 unsigned long timeout;
3248 if (!valid_lft || (prefered_lft > valid_lft))
3249 return -EINVAL;
3251 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3252 if (addrconf_finite_timeout(timeout)) {
3253 expires = jiffies_to_clock_t(timeout * HZ);
3254 valid_lft = timeout;
3255 flags = RTF_EXPIRES;
3256 } else {
3257 expires = 0;
3258 flags = 0;
3259 ifa_flags |= IFA_F_PERMANENT;
3262 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3263 if (addrconf_finite_timeout(timeout)) {
3264 if (timeout == 0)
3265 ifa_flags |= IFA_F_DEPRECATED;
3266 prefered_lft = timeout;
3269 spin_lock_bh(&ifp->lock);
3270 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3271 ifp->tstamp = jiffies;
3272 ifp->valid_lft = valid_lft;
3273 ifp->prefered_lft = prefered_lft;
3275 spin_unlock_bh(&ifp->lock);
3276 if (!(ifp->flags&IFA_F_TENTATIVE))
3277 ipv6_ifa_notify(0, ifp);
3279 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3280 expires, flags);
3281 addrconf_verify(0);
3283 return 0;
3286 static int
3287 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3289 struct net *net = sock_net(skb->sk);
3290 struct ifaddrmsg *ifm;
3291 struct nlattr *tb[IFA_MAX+1];
3292 struct in6_addr *pfx;
3293 struct inet6_ifaddr *ifa;
3294 struct net_device *dev;
3295 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3296 u8 ifa_flags;
3297 int err;
3299 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3300 if (err < 0)
3301 return err;
3303 ifm = nlmsg_data(nlh);
3304 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3305 if (pfx == NULL)
3306 return -EINVAL;
3308 if (tb[IFA_CACHEINFO]) {
3309 struct ifa_cacheinfo *ci;
3311 ci = nla_data(tb[IFA_CACHEINFO]);
3312 valid_lft = ci->ifa_valid;
3313 preferred_lft = ci->ifa_prefered;
3314 } else {
3315 preferred_lft = INFINITY_LIFE_TIME;
3316 valid_lft = INFINITY_LIFE_TIME;
3319 dev = __dev_get_by_index(net, ifm->ifa_index);
3320 if (dev == NULL)
3321 return -ENODEV;
3323 /* We ignore other flags so far. */
3324 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3326 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3327 if (ifa == NULL) {
3329 * It would be best to check for !NLM_F_CREATE here but
3330 * userspace alreay relies on not having to provide this.
3332 return inet6_addr_add(net, ifm->ifa_index, pfx,
3333 ifm->ifa_prefixlen, ifa_flags,
3334 preferred_lft, valid_lft);
3337 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3338 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3339 err = -EEXIST;
3340 else
3341 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3343 in6_ifa_put(ifa);
3345 return err;
3348 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3349 u8 scope, int ifindex)
3351 struct ifaddrmsg *ifm;
3353 ifm = nlmsg_data(nlh);
3354 ifm->ifa_family = AF_INET6;
3355 ifm->ifa_prefixlen = prefixlen;
3356 ifm->ifa_flags = flags;
3357 ifm->ifa_scope = scope;
3358 ifm->ifa_index = ifindex;
3361 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3362 unsigned long tstamp, u32 preferred, u32 valid)
3364 struct ifa_cacheinfo ci;
3366 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3367 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3368 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3369 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3370 ci.ifa_prefered = preferred;
3371 ci.ifa_valid = valid;
3373 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3376 static inline int rt_scope(int ifa_scope)
3378 if (ifa_scope & IFA_HOST)
3379 return RT_SCOPE_HOST;
3380 else if (ifa_scope & IFA_LINK)
3381 return RT_SCOPE_LINK;
3382 else if (ifa_scope & IFA_SITE)
3383 return RT_SCOPE_SITE;
3384 else
3385 return RT_SCOPE_UNIVERSE;
3388 static inline int inet6_ifaddr_msgsize(void)
3390 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3391 + nla_total_size(16) /* IFA_ADDRESS */
3392 + nla_total_size(sizeof(struct ifa_cacheinfo));
3395 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3396 u32 pid, u32 seq, int event, unsigned int flags)
3398 struct nlmsghdr *nlh;
3399 u32 preferred, valid;
3401 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3402 if (nlh == NULL)
3403 return -EMSGSIZE;
3405 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3406 ifa->idev->dev->ifindex);
3408 if (!(ifa->flags&IFA_F_PERMANENT)) {
3409 preferred = ifa->prefered_lft;
3410 valid = ifa->valid_lft;
3411 if (preferred != INFINITY_LIFE_TIME) {
3412 long tval = (jiffies - ifa->tstamp)/HZ;
3413 if (preferred > tval)
3414 preferred -= tval;
3415 else
3416 preferred = 0;
3417 if (valid != INFINITY_LIFE_TIME)
3418 valid -= tval;
3420 } else {
3421 preferred = INFINITY_LIFE_TIME;
3422 valid = INFINITY_LIFE_TIME;
3425 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3426 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3427 nlmsg_cancel(skb, nlh);
3428 return -EMSGSIZE;
3431 return nlmsg_end(skb, nlh);
3434 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3435 u32 pid, u32 seq, int event, u16 flags)
3437 struct nlmsghdr *nlh;
3438 u8 scope = RT_SCOPE_UNIVERSE;
3439 int ifindex = ifmca->idev->dev->ifindex;
3441 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3442 scope = RT_SCOPE_SITE;
3444 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3445 if (nlh == NULL)
3446 return -EMSGSIZE;
3448 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3449 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3450 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3451 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3452 nlmsg_cancel(skb, nlh);
3453 return -EMSGSIZE;
3456 return nlmsg_end(skb, nlh);
3459 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3460 u32 pid, u32 seq, int event, unsigned int flags)
3462 struct nlmsghdr *nlh;
3463 u8 scope = RT_SCOPE_UNIVERSE;
3464 int ifindex = ifaca->aca_idev->dev->ifindex;
3466 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3467 scope = RT_SCOPE_SITE;
3469 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3470 if (nlh == NULL)
3471 return -EMSGSIZE;
3473 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3474 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3475 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3476 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3477 nlmsg_cancel(skb, nlh);
3478 return -EMSGSIZE;
3481 return nlmsg_end(skb, nlh);
3484 enum addr_type_t
3486 UNICAST_ADDR,
3487 MULTICAST_ADDR,
3488 ANYCAST_ADDR,
3491 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3492 enum addr_type_t type)
3494 int idx, ip_idx;
3495 int s_idx, s_ip_idx;
3496 int err = 1;
3497 struct net_device *dev;
3498 struct inet6_dev *idev = NULL;
3499 struct inet6_ifaddr *ifa;
3500 struct ifmcaddr6 *ifmca;
3501 struct ifacaddr6 *ifaca;
3502 struct net *net = sock_net(skb->sk);
3504 s_idx = cb->args[0];
3505 s_ip_idx = ip_idx = cb->args[1];
3507 idx = 0;
3508 for_each_netdev(net, dev) {
3509 if (idx < s_idx)
3510 goto cont;
3511 if (idx > s_idx)
3512 s_ip_idx = 0;
3513 ip_idx = 0;
3514 if ((idev = in6_dev_get(dev)) == NULL)
3515 goto cont;
3516 read_lock_bh(&idev->lock);
3517 switch (type) {
3518 case UNICAST_ADDR:
3519 /* unicast address incl. temp addr */
3520 for (ifa = idev->addr_list; ifa;
3521 ifa = ifa->if_next, ip_idx++) {
3522 if (ip_idx < s_ip_idx)
3523 continue;
3524 err = inet6_fill_ifaddr(skb, ifa,
3525 NETLINK_CB(cb->skb).pid,
3526 cb->nlh->nlmsg_seq,
3527 RTM_NEWADDR,
3528 NLM_F_MULTI);
3530 break;
3531 case MULTICAST_ADDR:
3532 /* multicast address */
3533 for (ifmca = idev->mc_list; ifmca;
3534 ifmca = ifmca->next, ip_idx++) {
3535 if (ip_idx < s_ip_idx)
3536 continue;
3537 err = inet6_fill_ifmcaddr(skb, ifmca,
3538 NETLINK_CB(cb->skb).pid,
3539 cb->nlh->nlmsg_seq,
3540 RTM_GETMULTICAST,
3541 NLM_F_MULTI);
3543 break;
3544 case ANYCAST_ADDR:
3545 /* anycast address */
3546 for (ifaca = idev->ac_list; ifaca;
3547 ifaca = ifaca->aca_next, ip_idx++) {
3548 if (ip_idx < s_ip_idx)
3549 continue;
3550 err = inet6_fill_ifacaddr(skb, ifaca,
3551 NETLINK_CB(cb->skb).pid,
3552 cb->nlh->nlmsg_seq,
3553 RTM_GETANYCAST,
3554 NLM_F_MULTI);
3556 break;
3557 default:
3558 break;
3560 read_unlock_bh(&idev->lock);
3561 in6_dev_put(idev);
3563 if (err <= 0)
3564 break;
3565 cont:
3566 idx++;
3568 cb->args[0] = idx;
3569 cb->args[1] = ip_idx;
3570 return skb->len;
3573 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3575 enum addr_type_t type = UNICAST_ADDR;
3577 return inet6_dump_addr(skb, cb, type);
3580 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3582 enum addr_type_t type = MULTICAST_ADDR;
3584 return inet6_dump_addr(skb, cb, type);
3588 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3590 enum addr_type_t type = ANYCAST_ADDR;
3592 return inet6_dump_addr(skb, cb, type);
3595 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3596 void *arg)
3598 struct net *net = sock_net(in_skb->sk);
3599 struct ifaddrmsg *ifm;
3600 struct nlattr *tb[IFA_MAX+1];
3601 struct in6_addr *addr = NULL;
3602 struct net_device *dev = NULL;
3603 struct inet6_ifaddr *ifa;
3604 struct sk_buff *skb;
3605 int err;
3607 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3608 if (err < 0)
3609 goto errout;
3611 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3612 if (addr == NULL) {
3613 err = -EINVAL;
3614 goto errout;
3617 ifm = nlmsg_data(nlh);
3618 if (ifm->ifa_index)
3619 dev = __dev_get_by_index(net, ifm->ifa_index);
3621 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3622 err = -EADDRNOTAVAIL;
3623 goto errout;
3626 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3627 err = -ENOBUFS;
3628 goto errout_ifa;
3631 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3632 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3633 if (err < 0) {
3634 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3635 WARN_ON(err == -EMSGSIZE);
3636 kfree_skb(skb);
3637 goto errout_ifa;
3639 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3640 errout_ifa:
3641 in6_ifa_put(ifa);
3642 errout:
3643 return err;
3646 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3648 struct sk_buff *skb;
3649 struct net *net = dev_net(ifa->idev->dev);
3650 int err = -ENOBUFS;
3652 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3653 if (skb == NULL)
3654 goto errout;
3656 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3657 if (err < 0) {
3658 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3659 WARN_ON(err == -EMSGSIZE);
3660 kfree_skb(skb);
3661 goto errout;
3663 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3664 return;
3665 errout:
3666 if (err < 0)
3667 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3670 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3671 __s32 *array, int bytes)
3673 BUG_ON(bytes < (DEVCONF_MAX * 4));
3675 memset(array, 0, bytes);
3676 array[DEVCONF_FORWARDING] = cnf->forwarding;
3677 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3678 array[DEVCONF_MTU6] = cnf->mtu6;
3679 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3680 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3681 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3682 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3683 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3684 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3685 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3686 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3687 #ifdef CONFIG_IPV6_PRIVACY
3688 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3689 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3690 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3691 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3692 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3693 #endif
3694 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3695 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3696 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3697 #ifdef CONFIG_IPV6_ROUTER_PREF
3698 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3699 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3700 #ifdef CONFIG_IPV6_ROUTE_INFO
3701 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3702 #endif
3703 #endif
3704 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3705 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3706 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3707 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3708 #endif
3709 #ifdef CONFIG_IPV6_MROUTE
3710 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3711 #endif
3712 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3713 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3716 static inline size_t inet6_if_nlmsg_size(void)
3718 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3719 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3720 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3721 + nla_total_size(4) /* IFLA_MTU */
3722 + nla_total_size(4) /* IFLA_LINK */
3723 + nla_total_size( /* IFLA_PROTINFO */
3724 nla_total_size(4) /* IFLA_INET6_FLAGS */
3725 + nla_total_size(sizeof(struct ifla_cacheinfo))
3726 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3727 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3728 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3732 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3733 int bytes)
3735 int i;
3736 int pad = bytes - sizeof(u64) * items;
3737 BUG_ON(pad < 0);
3739 /* Use put_unaligned() because stats may not be aligned for u64. */
3740 put_unaligned(items, &stats[0]);
3741 for (i = 1; i < items; i++)
3742 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3744 memset(&stats[items], 0, pad);
3747 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3748 int bytes)
3750 switch(attrtype) {
3751 case IFLA_INET6_STATS:
3752 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3753 break;
3754 case IFLA_INET6_ICMP6STATS:
3755 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3756 break;
3760 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3761 u32 pid, u32 seq, int event, unsigned int flags)
3763 struct net_device *dev = idev->dev;
3764 struct nlattr *nla;
3765 struct ifinfomsg *hdr;
3766 struct nlmsghdr *nlh;
3767 void *protoinfo;
3768 struct ifla_cacheinfo ci;
3770 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3771 if (nlh == NULL)
3772 return -EMSGSIZE;
3774 hdr = nlmsg_data(nlh);
3775 hdr->ifi_family = AF_INET6;
3776 hdr->__ifi_pad = 0;
3777 hdr->ifi_type = dev->type;
3778 hdr->ifi_index = dev->ifindex;
3779 hdr->ifi_flags = dev_get_flags(dev);
3780 hdr->ifi_change = 0;
3782 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3784 if (dev->addr_len)
3785 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3787 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3788 if (dev->ifindex != dev->iflink)
3789 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3791 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3792 if (protoinfo == NULL)
3793 goto nla_put_failure;
3795 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3797 ci.max_reasm_len = IPV6_MAXPLEN;
3798 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3799 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3800 ci.reachable_time = idev->nd_parms->reachable_time;
3801 ci.retrans_time = idev->nd_parms->retrans_time;
3802 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3804 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3805 if (nla == NULL)
3806 goto nla_put_failure;
3807 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3809 /* XXX - MC not implemented */
3811 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3812 if (nla == NULL)
3813 goto nla_put_failure;
3814 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3816 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3817 if (nla == NULL)
3818 goto nla_put_failure;
3819 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3821 nla_nest_end(skb, protoinfo);
3822 return nlmsg_end(skb, nlh);
3824 nla_put_failure:
3825 nlmsg_cancel(skb, nlh);
3826 return -EMSGSIZE;
3829 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3831 struct net *net = sock_net(skb->sk);
3832 int idx, err;
3833 int s_idx = cb->args[0];
3834 struct net_device *dev;
3835 struct inet6_dev *idev;
3837 read_lock(&dev_base_lock);
3838 idx = 0;
3839 for_each_netdev(net, dev) {
3840 if (idx < s_idx)
3841 goto cont;
3842 if ((idev = in6_dev_get(dev)) == NULL)
3843 goto cont;
3844 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3845 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3846 in6_dev_put(idev);
3847 if (err <= 0)
3848 break;
3849 cont:
3850 idx++;
3852 read_unlock(&dev_base_lock);
3853 cb->args[0] = idx;
3855 return skb->len;
3858 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3860 struct sk_buff *skb;
3861 struct net *net = dev_net(idev->dev);
3862 int err = -ENOBUFS;
3864 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3865 if (skb == NULL)
3866 goto errout;
3868 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3869 if (err < 0) {
3870 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3871 WARN_ON(err == -EMSGSIZE);
3872 kfree_skb(skb);
3873 goto errout;
3875 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3876 return;
3877 errout:
3878 if (err < 0)
3879 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3882 static inline size_t inet6_prefix_nlmsg_size(void)
3884 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3885 + nla_total_size(sizeof(struct in6_addr))
3886 + nla_total_size(sizeof(struct prefix_cacheinfo));
3889 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3890 struct prefix_info *pinfo, u32 pid, u32 seq,
3891 int event, unsigned int flags)
3893 struct prefixmsg *pmsg;
3894 struct nlmsghdr *nlh;
3895 struct prefix_cacheinfo ci;
3897 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3898 if (nlh == NULL)
3899 return -EMSGSIZE;
3901 pmsg = nlmsg_data(nlh);
3902 pmsg->prefix_family = AF_INET6;
3903 pmsg->prefix_pad1 = 0;
3904 pmsg->prefix_pad2 = 0;
3905 pmsg->prefix_ifindex = idev->dev->ifindex;
3906 pmsg->prefix_len = pinfo->prefix_len;
3907 pmsg->prefix_type = pinfo->type;
3908 pmsg->prefix_pad3 = 0;
3909 pmsg->prefix_flags = 0;
3910 if (pinfo->onlink)
3911 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3912 if (pinfo->autoconf)
3913 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3915 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3917 ci.preferred_time = ntohl(pinfo->prefered);
3918 ci.valid_time = ntohl(pinfo->valid);
3919 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3921 return nlmsg_end(skb, nlh);
3923 nla_put_failure:
3924 nlmsg_cancel(skb, nlh);
3925 return -EMSGSIZE;
3928 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3929 struct prefix_info *pinfo)
3931 struct sk_buff *skb;
3932 struct net *net = dev_net(idev->dev);
3933 int err = -ENOBUFS;
3935 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3936 if (skb == NULL)
3937 goto errout;
3939 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3940 if (err < 0) {
3941 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3942 WARN_ON(err == -EMSGSIZE);
3943 kfree_skb(skb);
3944 goto errout;
3946 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3947 return;
3948 errout:
3949 if (err < 0)
3950 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3953 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3955 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3957 switch (event) {
3958 case RTM_NEWADDR:
3960 * If the address was optimistic
3961 * we inserted the route at the start of
3962 * our DAD process, so we don't need
3963 * to do it again
3965 if (!(ifp->rt->rt6i_node))
3966 ip6_ins_rt(ifp->rt);
3967 if (ifp->idev->cnf.forwarding)
3968 addrconf_join_anycast(ifp);
3969 break;
3970 case RTM_DELADDR:
3971 if (ifp->idev->cnf.forwarding)
3972 addrconf_leave_anycast(ifp);
3973 addrconf_leave_solict(ifp->idev, &ifp->addr);
3974 dst_hold(&ifp->rt->u.dst);
3975 if (ip6_del_rt(ifp->rt))
3976 dst_free(&ifp->rt->u.dst);
3977 break;
3981 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3983 rcu_read_lock_bh();
3984 if (likely(ifp->idev->dead == 0))
3985 __ipv6_ifa_notify(event, ifp);
3986 rcu_read_unlock_bh();
3989 #ifdef CONFIG_SYSCTL
3991 static
3992 int addrconf_sysctl_forward(ctl_table *ctl, int write,
3993 void __user *buffer, size_t *lenp, loff_t *ppos)
3995 int *valp = ctl->data;
3996 int val = *valp;
3997 loff_t pos = *ppos;
3998 int ret;
4000 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4002 if (write)
4003 ret = addrconf_fixup_forwarding(ctl, valp, val);
4004 if (ret)
4005 *ppos = pos;
4006 return ret;
4009 static int addrconf_sysctl_forward_strategy(ctl_table *table,
4010 void __user *oldval,
4011 size_t __user *oldlenp,
4012 void __user *newval, size_t newlen)
4014 int *valp = table->data;
4015 int val = *valp;
4016 int new;
4018 if (!newval || !newlen)
4019 return 0;
4020 if (newlen != sizeof(int))
4021 return -EINVAL;
4022 if (get_user(new, (int __user *)newval))
4023 return -EFAULT;
4024 if (new == *valp)
4025 return 0;
4026 if (oldval && oldlenp) {
4027 size_t len;
4028 if (get_user(len, oldlenp))
4029 return -EFAULT;
4030 if (len) {
4031 if (len > table->maxlen)
4032 len = table->maxlen;
4033 if (copy_to_user(oldval, valp, len))
4034 return -EFAULT;
4035 if (put_user(len, oldlenp))
4036 return -EFAULT;
4040 *valp = new;
4041 return addrconf_fixup_forwarding(table, valp, val);
4044 static void dev_disable_change(struct inet6_dev *idev)
4046 if (!idev || !idev->dev)
4047 return;
4049 if (idev->cnf.disable_ipv6)
4050 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4051 else
4052 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4055 static void addrconf_disable_change(struct net *net, __s32 newf)
4057 struct net_device *dev;
4058 struct inet6_dev *idev;
4060 read_lock(&dev_base_lock);
4061 for_each_netdev(net, dev) {
4062 rcu_read_lock();
4063 idev = __in6_dev_get(dev);
4064 if (idev) {
4065 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4066 idev->cnf.disable_ipv6 = newf;
4067 if (changed)
4068 dev_disable_change(idev);
4070 rcu_read_unlock();
4072 read_unlock(&dev_base_lock);
4075 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4077 struct net *net;
4079 net = (struct net *)table->extra2;
4081 if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4082 return 0;
4084 if (!rtnl_trylock()) {
4085 /* Restore the original values before restarting */
4086 *p = old;
4087 return restart_syscall();
4090 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4091 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4092 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4093 addrconf_disable_change(net, newf);
4094 } else if ((!*p) ^ (!old))
4095 dev_disable_change((struct inet6_dev *)table->extra1);
4097 rtnl_unlock();
4098 return 0;
4101 static
4102 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4103 void __user *buffer, size_t *lenp, loff_t *ppos)
4105 int *valp = ctl->data;
4106 int val = *valp;
4107 loff_t pos = *ppos;
4108 int ret;
4110 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4112 if (write)
4113 ret = addrconf_disable_ipv6(ctl, valp, val);
4114 if (ret)
4115 *ppos = pos;
4116 return ret;
4119 static struct addrconf_sysctl_table
4121 struct ctl_table_header *sysctl_header;
4122 ctl_table addrconf_vars[DEVCONF_MAX+1];
4123 char *dev_name;
4124 } addrconf_sysctl __read_mostly = {
4125 .sysctl_header = NULL,
4126 .addrconf_vars = {
4128 .ctl_name = NET_IPV6_FORWARDING,
4129 .procname = "forwarding",
4130 .data = &ipv6_devconf.forwarding,
4131 .maxlen = sizeof(int),
4132 .mode = 0644,
4133 .proc_handler = addrconf_sysctl_forward,
4134 .strategy = addrconf_sysctl_forward_strategy,
4137 .ctl_name = NET_IPV6_HOP_LIMIT,
4138 .procname = "hop_limit",
4139 .data = &ipv6_devconf.hop_limit,
4140 .maxlen = sizeof(int),
4141 .mode = 0644,
4142 .proc_handler = proc_dointvec,
4145 .ctl_name = NET_IPV6_MTU,
4146 .procname = "mtu",
4147 .data = &ipv6_devconf.mtu6,
4148 .maxlen = sizeof(int),
4149 .mode = 0644,
4150 .proc_handler = proc_dointvec,
4153 .ctl_name = NET_IPV6_ACCEPT_RA,
4154 .procname = "accept_ra",
4155 .data = &ipv6_devconf.accept_ra,
4156 .maxlen = sizeof(int),
4157 .mode = 0644,
4158 .proc_handler = proc_dointvec,
4161 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4162 .procname = "accept_redirects",
4163 .data = &ipv6_devconf.accept_redirects,
4164 .maxlen = sizeof(int),
4165 .mode = 0644,
4166 .proc_handler = proc_dointvec,
4169 .ctl_name = NET_IPV6_AUTOCONF,
4170 .procname = "autoconf",
4171 .data = &ipv6_devconf.autoconf,
4172 .maxlen = sizeof(int),
4173 .mode = 0644,
4174 .proc_handler = proc_dointvec,
4177 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4178 .procname = "dad_transmits",
4179 .data = &ipv6_devconf.dad_transmits,
4180 .maxlen = sizeof(int),
4181 .mode = 0644,
4182 .proc_handler = proc_dointvec,
4185 .ctl_name = NET_IPV6_RTR_SOLICITS,
4186 .procname = "router_solicitations",
4187 .data = &ipv6_devconf.rtr_solicits,
4188 .maxlen = sizeof(int),
4189 .mode = 0644,
4190 .proc_handler = proc_dointvec,
4193 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4194 .procname = "router_solicitation_interval",
4195 .data = &ipv6_devconf.rtr_solicit_interval,
4196 .maxlen = sizeof(int),
4197 .mode = 0644,
4198 .proc_handler = proc_dointvec_jiffies,
4199 .strategy = sysctl_jiffies,
4202 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4203 .procname = "router_solicitation_delay",
4204 .data = &ipv6_devconf.rtr_solicit_delay,
4205 .maxlen = sizeof(int),
4206 .mode = 0644,
4207 .proc_handler = proc_dointvec_jiffies,
4208 .strategy = sysctl_jiffies,
4211 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4212 .procname = "force_mld_version",
4213 .data = &ipv6_devconf.force_mld_version,
4214 .maxlen = sizeof(int),
4215 .mode = 0644,
4216 .proc_handler = proc_dointvec,
4218 #ifdef CONFIG_IPV6_PRIVACY
4220 .ctl_name = NET_IPV6_USE_TEMPADDR,
4221 .procname = "use_tempaddr",
4222 .data = &ipv6_devconf.use_tempaddr,
4223 .maxlen = sizeof(int),
4224 .mode = 0644,
4225 .proc_handler = proc_dointvec,
4228 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4229 .procname = "temp_valid_lft",
4230 .data = &ipv6_devconf.temp_valid_lft,
4231 .maxlen = sizeof(int),
4232 .mode = 0644,
4233 .proc_handler = proc_dointvec,
4236 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4237 .procname = "temp_prefered_lft",
4238 .data = &ipv6_devconf.temp_prefered_lft,
4239 .maxlen = sizeof(int),
4240 .mode = 0644,
4241 .proc_handler = proc_dointvec,
4244 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4245 .procname = "regen_max_retry",
4246 .data = &ipv6_devconf.regen_max_retry,
4247 .maxlen = sizeof(int),
4248 .mode = 0644,
4249 .proc_handler = proc_dointvec,
4252 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4253 .procname = "max_desync_factor",
4254 .data = &ipv6_devconf.max_desync_factor,
4255 .maxlen = sizeof(int),
4256 .mode = 0644,
4257 .proc_handler = proc_dointvec,
4259 #endif
4261 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4262 .procname = "max_addresses",
4263 .data = &ipv6_devconf.max_addresses,
4264 .maxlen = sizeof(int),
4265 .mode = 0644,
4266 .proc_handler = proc_dointvec,
4269 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4270 .procname = "accept_ra_defrtr",
4271 .data = &ipv6_devconf.accept_ra_defrtr,
4272 .maxlen = sizeof(int),
4273 .mode = 0644,
4274 .proc_handler = proc_dointvec,
4277 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4278 .procname = "accept_ra_pinfo",
4279 .data = &ipv6_devconf.accept_ra_pinfo,
4280 .maxlen = sizeof(int),
4281 .mode = 0644,
4282 .proc_handler = proc_dointvec,
4284 #ifdef CONFIG_IPV6_ROUTER_PREF
4286 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4287 .procname = "accept_ra_rtr_pref",
4288 .data = &ipv6_devconf.accept_ra_rtr_pref,
4289 .maxlen = sizeof(int),
4290 .mode = 0644,
4291 .proc_handler = proc_dointvec,
4294 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4295 .procname = "router_probe_interval",
4296 .data = &ipv6_devconf.rtr_probe_interval,
4297 .maxlen = sizeof(int),
4298 .mode = 0644,
4299 .proc_handler = proc_dointvec_jiffies,
4300 .strategy = sysctl_jiffies,
4302 #ifdef CONFIG_IPV6_ROUTE_INFO
4304 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4305 .procname = "accept_ra_rt_info_max_plen",
4306 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4307 .maxlen = sizeof(int),
4308 .mode = 0644,
4309 .proc_handler = proc_dointvec,
4311 #endif
4312 #endif
4314 .ctl_name = NET_IPV6_PROXY_NDP,
4315 .procname = "proxy_ndp",
4316 .data = &ipv6_devconf.proxy_ndp,
4317 .maxlen = sizeof(int),
4318 .mode = 0644,
4319 .proc_handler = proc_dointvec,
4322 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4323 .procname = "accept_source_route",
4324 .data = &ipv6_devconf.accept_source_route,
4325 .maxlen = sizeof(int),
4326 .mode = 0644,
4327 .proc_handler = proc_dointvec,
4329 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4331 .ctl_name = CTL_UNNUMBERED,
4332 .procname = "optimistic_dad",
4333 .data = &ipv6_devconf.optimistic_dad,
4334 .maxlen = sizeof(int),
4335 .mode = 0644,
4336 .proc_handler = proc_dointvec,
4339 #endif
4340 #ifdef CONFIG_IPV6_MROUTE
4342 .ctl_name = CTL_UNNUMBERED,
4343 .procname = "mc_forwarding",
4344 .data = &ipv6_devconf.mc_forwarding,
4345 .maxlen = sizeof(int),
4346 .mode = 0444,
4347 .proc_handler = proc_dointvec,
4349 #endif
4351 .ctl_name = CTL_UNNUMBERED,
4352 .procname = "disable_ipv6",
4353 .data = &ipv6_devconf.disable_ipv6,
4354 .maxlen = sizeof(int),
4355 .mode = 0644,
4356 .proc_handler = addrconf_sysctl_disable,
4357 .strategy = sysctl_intvec,
4360 .ctl_name = CTL_UNNUMBERED,
4361 .procname = "accept_dad",
4362 .data = &ipv6_devconf.accept_dad,
4363 .maxlen = sizeof(int),
4364 .mode = 0644,
4365 .proc_handler = proc_dointvec,
4368 .ctl_name = 0, /* sentinel */
4373 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4374 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4376 int i;
4377 struct addrconf_sysctl_table *t;
4379 #define ADDRCONF_CTL_PATH_DEV 3
4381 struct ctl_path addrconf_ctl_path[] = {
4382 { .procname = "net", .ctl_name = CTL_NET, },
4383 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4384 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4385 { /* to be set */ },
4386 { },
4390 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4391 if (t == NULL)
4392 goto out;
4394 for (i=0; t->addrconf_vars[i].data; i++) {
4395 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4396 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4397 t->addrconf_vars[i].extra2 = net;
4401 * Make a copy of dev_name, because '.procname' is regarded as const
4402 * by sysctl and we wouldn't want anyone to change it under our feet
4403 * (see SIOCSIFNAME).
4405 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4406 if (!t->dev_name)
4407 goto free;
4409 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4410 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4412 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4413 t->addrconf_vars);
4414 if (t->sysctl_header == NULL)
4415 goto free_procname;
4417 p->sysctl = t;
4418 return 0;
4420 free_procname:
4421 kfree(t->dev_name);
4422 free:
4423 kfree(t);
4424 out:
4425 return -ENOBUFS;
4428 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4430 struct addrconf_sysctl_table *t;
4432 if (p->sysctl == NULL)
4433 return;
4435 t = p->sysctl;
4436 p->sysctl = NULL;
4437 unregister_sysctl_table(t->sysctl_header);
4438 kfree(t->dev_name);
4439 kfree(t);
4442 static void addrconf_sysctl_register(struct inet6_dev *idev)
4444 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4445 NET_IPV6_NEIGH, "ipv6",
4446 &ndisc_ifinfo_sysctl_change,
4447 ndisc_ifinfo_sysctl_strategy);
4448 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4449 idev->dev->ifindex, idev, &idev->cnf);
4452 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4454 __addrconf_sysctl_unregister(&idev->cnf);
4455 neigh_sysctl_unregister(idev->nd_parms);
4459 #endif
4461 static int addrconf_init_net(struct net *net)
4463 int err;
4464 struct ipv6_devconf *all, *dflt;
4466 err = -ENOMEM;
4467 all = &ipv6_devconf;
4468 dflt = &ipv6_devconf_dflt;
4470 if (net != &init_net) {
4471 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4472 if (all == NULL)
4473 goto err_alloc_all;
4475 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4476 if (dflt == NULL)
4477 goto err_alloc_dflt;
4478 } else {
4479 /* these will be inherited by all namespaces */
4480 dflt->autoconf = ipv6_defaults.autoconf;
4481 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4484 net->ipv6.devconf_all = all;
4485 net->ipv6.devconf_dflt = dflt;
4487 #ifdef CONFIG_SYSCTL
4488 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4489 NULL, all);
4490 if (err < 0)
4491 goto err_reg_all;
4493 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4494 NULL, dflt);
4495 if (err < 0)
4496 goto err_reg_dflt;
4497 #endif
4498 return 0;
4500 #ifdef CONFIG_SYSCTL
4501 err_reg_dflt:
4502 __addrconf_sysctl_unregister(all);
4503 err_reg_all:
4504 kfree(dflt);
4505 #endif
4506 err_alloc_dflt:
4507 kfree(all);
4508 err_alloc_all:
4509 return err;
4512 static void addrconf_exit_net(struct net *net)
4514 #ifdef CONFIG_SYSCTL
4515 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4516 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4517 #endif
4518 if (net != &init_net) {
4519 kfree(net->ipv6.devconf_dflt);
4520 kfree(net->ipv6.devconf_all);
4524 static struct pernet_operations addrconf_ops = {
4525 .init = addrconf_init_net,
4526 .exit = addrconf_exit_net,
4530 * Device notifier
4533 int register_inet6addr_notifier(struct notifier_block *nb)
4535 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4538 EXPORT_SYMBOL(register_inet6addr_notifier);
4540 int unregister_inet6addr_notifier(struct notifier_block *nb)
4542 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4545 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4548 * Init / cleanup code
4551 int __init addrconf_init(void)
4553 int err;
4555 if ((err = ipv6_addr_label_init()) < 0) {
4556 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4557 err);
4558 return err;
4561 register_pernet_subsys(&addrconf_ops);
4563 /* The addrconf netdev notifier requires that loopback_dev
4564 * has it's ipv6 private information allocated and setup
4565 * before it can bring up and give link-local addresses
4566 * to other devices which are up.
4568 * Unfortunately, loopback_dev is not necessarily the first
4569 * entry in the global dev_base list of net devices. In fact,
4570 * it is likely to be the very last entry on that list.
4571 * So this causes the notifier registry below to try and
4572 * give link-local addresses to all devices besides loopback_dev
4573 * first, then loopback_dev, which cases all the non-loopback_dev
4574 * devices to fail to get a link-local address.
4576 * So, as a temporary fix, allocate the ipv6 structure for
4577 * loopback_dev first by hand.
4578 * Longer term, all of the dependencies ipv6 has upon the loopback
4579 * device and it being up should be removed.
4581 rtnl_lock();
4582 if (!ipv6_add_dev(init_net.loopback_dev))
4583 err = -ENOMEM;
4584 rtnl_unlock();
4585 if (err)
4586 goto errlo;
4588 register_netdevice_notifier(&ipv6_dev_notf);
4590 addrconf_verify(0);
4592 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4593 if (err < 0)
4594 goto errout;
4596 /* Only the first call to __rtnl_register can fail */
4597 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4598 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4599 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4600 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4601 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4603 ipv6_addr_label_rtnl_register();
4605 return 0;
4606 errout:
4607 unregister_netdevice_notifier(&ipv6_dev_notf);
4608 errlo:
4609 unregister_pernet_subsys(&addrconf_ops);
4611 return err;
4614 void addrconf_cleanup(void)
4616 struct inet6_ifaddr *ifa;
4617 struct net_device *dev;
4618 int i;
4620 unregister_netdevice_notifier(&ipv6_dev_notf);
4621 unregister_pernet_subsys(&addrconf_ops);
4623 rtnl_lock();
4625 /* clean dev list */
4626 for_each_netdev(&init_net, dev) {
4627 if (__in6_dev_get(dev) == NULL)
4628 continue;
4629 addrconf_ifdown(dev, 1);
4631 addrconf_ifdown(init_net.loopback_dev, 2);
4634 * Check hash table.
4636 write_lock_bh(&addrconf_hash_lock);
4637 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4638 for (ifa=inet6_addr_lst[i]; ifa; ) {
4639 struct inet6_ifaddr *bifa;
4641 bifa = ifa;
4642 ifa = ifa->lst_next;
4643 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4644 /* Do not free it; something is wrong.
4645 Now we can investigate it with debugger.
4649 write_unlock_bh(&addrconf_hash_lock);
4651 del_timer(&addr_chk_timer);
4652 rtnl_unlock();