ipv6 addrconf: Allow infinite prefix lifetime.
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / addrconf.c
blob3a835578fd1c8adae7d99f1573a16989d0c8d369
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 * $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
18 * Changes:
20 * Janos Farkas : delete timer on ifdown
21 * <chexum@bankinf.banki.hu>
22 * Andi Kleen : kill double kfree on module
23 * unload.
24 * Maciej W. Rozycki : FDDI support
25 * sekiya@USAGI : Don't send too many RS
26 * packets.
27 * yoshfuji@USAGI : Fixed interval between DAD
28 * packets.
29 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
30 * address validation timer.
31 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
32 * support.
33 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
34 * address on a same interface.
35 * YOSHIFUJI Hideaki @USAGI : ARCnet support
36 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
37 * seq_file.
38 * YOSHIFUJI Hideaki @USAGI : improved source address
39 * selection; consider scope,
40 * status etc.
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
100 #define INFINITY_LIFE_TIME 0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
114 #endif
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
124 static int ipv6_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 int addrconf_ifdown(struct net_device *dev, int how);
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153 struct net_device *dev);
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
157 struct ipv6_devconf ipv6_devconf __read_mostly = {
158 .forwarding = 0,
159 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
160 .mtu6 = IPV6_MIN_MTU,
161 .accept_ra = 1,
162 .accept_redirects = 1,
163 .autoconf = 1,
164 .force_mld_version = 0,
165 .dad_transmits = 1,
166 .rtr_solicits = MAX_RTR_SOLICITATIONS,
167 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
168 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170 .use_tempaddr = 0,
171 .temp_valid_lft = TEMP_VALID_LIFETIME,
172 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
173 .regen_max_retry = REGEN_MAX_RETRY,
174 .max_desync_factor = MAX_DESYNC_FACTOR,
175 #endif
176 .max_addresses = IPV6_MAX_ADDRESSES,
177 .accept_ra_defrtr = 1,
178 .accept_ra_pinfo = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180 .accept_ra_rtr_pref = 1,
181 .rtr_probe_interval = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183 .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186 .proxy_ndp = 0,
187 .accept_source_route = 0, /* we do not accept RH0 by default. */
190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
191 .forwarding = 0,
192 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
193 .mtu6 = IPV6_MIN_MTU,
194 .accept_ra = 1,
195 .accept_redirects = 1,
196 .autoconf = 1,
197 .dad_transmits = 1,
198 .rtr_solicits = MAX_RTR_SOLICITATIONS,
199 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
200 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
201 #ifdef CONFIG_IPV6_PRIVACY
202 .use_tempaddr = 0,
203 .temp_valid_lft = TEMP_VALID_LIFETIME,
204 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
205 .regen_max_retry = REGEN_MAX_RETRY,
206 .max_desync_factor = MAX_DESYNC_FACTOR,
207 #endif
208 .max_addresses = IPV6_MAX_ADDRESSES,
209 .accept_ra_defrtr = 1,
210 .accept_ra_pinfo = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212 .accept_ra_rtr_pref = 1,
213 .rtr_probe_interval = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215 .accept_ra_rt_info_max_plen = 0,
216 #endif
217 #endif
218 .proxy_ndp = 0,
219 .accept_source_route = 0, /* we do not accept RH0 by default. */
222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
225 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
226 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
228 /* Check if a valid qdisc is available */
229 static inline int addrconf_qdisc_ok(struct net_device *dev)
231 return (dev->qdisc != &noop_qdisc);
234 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
236 if (del_timer(&ifp->timer))
237 __in6_ifa_put(ifp);
240 enum addrconf_timer_t
242 AC_NONE,
243 AC_DAD,
244 AC_RS,
247 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
248 enum addrconf_timer_t what,
249 unsigned long when)
251 if (!del_timer(&ifp->timer))
252 in6_ifa_hold(ifp);
254 switch (what) {
255 case AC_DAD:
256 ifp->timer.function = addrconf_dad_timer;
257 break;
258 case AC_RS:
259 ifp->timer.function = addrconf_rs_timer;
260 break;
261 default:;
263 ifp->timer.expires = jiffies + when;
264 add_timer(&ifp->timer);
267 static int snmp6_alloc_dev(struct inet6_dev *idev)
269 if (snmp_mib_init((void **)idev->stats.ipv6,
270 sizeof(struct ipstats_mib)) < 0)
271 goto err_ip;
272 if (snmp_mib_init((void **)idev->stats.icmpv6,
273 sizeof(struct icmpv6_mib)) < 0)
274 goto err_icmp;
275 if (snmp_mib_init((void **)idev->stats.icmpv6msg,
276 sizeof(struct icmpv6msg_mib)) < 0)
277 goto err_icmpmsg;
279 return 0;
281 err_icmpmsg:
282 snmp_mib_free((void **)idev->stats.icmpv6);
283 err_icmp:
284 snmp_mib_free((void **)idev->stats.ipv6);
285 err_ip:
286 return -ENOMEM;
289 static void snmp6_free_dev(struct inet6_dev *idev)
291 snmp_mib_free((void **)idev->stats.icmpv6msg);
292 snmp_mib_free((void **)idev->stats.icmpv6);
293 snmp_mib_free((void **)idev->stats.ipv6);
296 /* Nobody refers to this device, we may destroy it. */
298 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
300 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
301 kfree(idev);
304 void in6_dev_finish_destroy(struct inet6_dev *idev)
306 struct net_device *dev = idev->dev;
307 BUG_TRAP(idev->addr_list==NULL);
308 BUG_TRAP(idev->mc_list==NULL);
309 #ifdef NET_REFCNT_DEBUG
310 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
311 #endif
312 dev_put(dev);
313 if (!idev->dead) {
314 printk("Freeing alive inet6 device %p\n", idev);
315 return;
317 snmp6_free_dev(idev);
318 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
321 EXPORT_SYMBOL(in6_dev_finish_destroy);
323 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
325 struct inet6_dev *ndev;
327 ASSERT_RTNL();
329 if (dev->mtu < IPV6_MIN_MTU)
330 return NULL;
332 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
334 if (ndev == NULL)
335 return NULL;
337 rwlock_init(&ndev->lock);
338 ndev->dev = dev;
339 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
340 ndev->cnf.mtu6 = dev->mtu;
341 ndev->cnf.sysctl = NULL;
342 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
343 if (ndev->nd_parms == NULL) {
344 kfree(ndev);
345 return NULL;
347 /* We refer to the device */
348 dev_hold(dev);
350 if (snmp6_alloc_dev(ndev) < 0) {
351 ADBG((KERN_WARNING
352 "%s(): cannot allocate memory for statistics; dev=%s.\n",
353 __func__, dev->name));
354 neigh_parms_release(&nd_tbl, ndev->nd_parms);
355 ndev->dead = 1;
356 in6_dev_finish_destroy(ndev);
357 return NULL;
360 if (snmp6_register_dev(ndev) < 0) {
361 ADBG((KERN_WARNING
362 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
363 __func__, dev->name));
364 neigh_parms_release(&nd_tbl, ndev->nd_parms);
365 ndev->dead = 1;
366 in6_dev_finish_destroy(ndev);
367 return NULL;
370 /* One reference from device. We must do this before
371 * we invoke __ipv6_regen_rndid().
373 in6_dev_hold(ndev);
375 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
376 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
377 printk(KERN_INFO
378 "%s: Disabled Multicast RS\n",
379 dev->name);
380 ndev->cnf.rtr_solicits = 0;
382 #endif
384 #ifdef CONFIG_IPV6_PRIVACY
385 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
386 if ((dev->flags&IFF_LOOPBACK) ||
387 dev->type == ARPHRD_TUNNEL ||
388 dev->type == ARPHRD_TUNNEL6 ||
389 dev->type == ARPHRD_SIT ||
390 dev->type == ARPHRD_NONE) {
391 printk(KERN_INFO
392 "%s: Disabled Privacy Extensions\n",
393 dev->name);
394 ndev->cnf.use_tempaddr = -1;
395 } else {
396 in6_dev_hold(ndev);
397 ipv6_regen_rndid((unsigned long) ndev);
399 #endif
401 if (netif_running(dev) && addrconf_qdisc_ok(dev))
402 ndev->if_flags |= IF_READY;
404 ipv6_mc_init_dev(ndev);
405 ndev->tstamp = jiffies;
406 addrconf_sysctl_register(ndev);
407 /* protected by rtnl_lock */
408 rcu_assign_pointer(dev->ip6_ptr, ndev);
410 /* Join all-node multicast group */
411 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
413 return ndev;
416 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
418 struct inet6_dev *idev;
420 ASSERT_RTNL();
422 if ((idev = __in6_dev_get(dev)) == NULL) {
423 if ((idev = ipv6_add_dev(dev)) == NULL)
424 return NULL;
427 if (dev->flags&IFF_UP)
428 ipv6_mc_up(idev);
429 return idev;
432 #ifdef CONFIG_SYSCTL
433 static void dev_forward_change(struct inet6_dev *idev)
435 struct net_device *dev;
436 struct inet6_ifaddr *ifa;
438 if (!idev)
439 return;
440 dev = idev->dev;
441 if (dev && (dev->flags & IFF_MULTICAST)) {
442 if (idev->cnf.forwarding)
443 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
444 else
445 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
447 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
448 if (ifa->flags&IFA_F_TENTATIVE)
449 continue;
450 if (idev->cnf.forwarding)
451 addrconf_join_anycast(ifa);
452 else
453 addrconf_leave_anycast(ifa);
458 static void addrconf_forward_change(struct net *net, __s32 newf)
460 struct net_device *dev;
461 struct inet6_dev *idev;
463 read_lock(&dev_base_lock);
464 for_each_netdev(net, dev) {
465 rcu_read_lock();
466 idev = __in6_dev_get(dev);
467 if (idev) {
468 int changed = (!idev->cnf.forwarding) ^ (!newf);
469 idev->cnf.forwarding = newf;
470 if (changed)
471 dev_forward_change(idev);
473 rcu_read_unlock();
475 read_unlock(&dev_base_lock);
478 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
480 struct net *net;
482 net = (struct net *)table->extra2;
483 if (p == &net->ipv6.devconf_dflt->forwarding)
484 return;
486 if (p == &net->ipv6.devconf_all->forwarding) {
487 __s32 newf = net->ipv6.devconf_all->forwarding;
488 net->ipv6.devconf_dflt->forwarding = newf;
489 addrconf_forward_change(net, newf);
490 } else if ((!*p) ^ (!old))
491 dev_forward_change((struct inet6_dev *)table->extra1);
493 if (*p)
494 rt6_purge_dflt_routers(net);
496 #endif
498 /* Nobody refers to this ifaddr, destroy it */
500 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
502 BUG_TRAP(ifp->if_next==NULL);
503 BUG_TRAP(ifp->lst_next==NULL);
504 #ifdef NET_REFCNT_DEBUG
505 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
506 #endif
508 in6_dev_put(ifp->idev);
510 if (del_timer(&ifp->timer))
511 printk("Timer is still running, when freeing ifa=%p\n", ifp);
513 if (!ifp->dead) {
514 printk("Freeing alive inet6 address %p\n", ifp);
515 return;
517 dst_release(&ifp->rt->u.dst);
519 kfree(ifp);
522 static void
523 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
525 struct inet6_ifaddr *ifa, **ifap;
526 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
529 * Each device address list is sorted in order of scope -
530 * global before linklocal.
532 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
533 ifap = &ifa->if_next) {
534 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
535 break;
538 ifp->if_next = *ifap;
539 *ifap = ifp;
543 * Hash function taken from net_alias.c
545 static u8 ipv6_addr_hash(const struct in6_addr *addr)
547 __u32 word;
550 * We perform the hash function over the last 64 bits of the address
551 * This will include the IEEE address token on links that support it.
554 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
555 word ^= (word >> 16);
556 word ^= (word >> 8);
558 return ((word ^ (word >> 4)) & 0x0f);
561 /* On success it returns ifp with increased reference count */
563 static struct inet6_ifaddr *
564 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
565 int scope, u32 flags)
567 struct inet6_ifaddr *ifa = NULL;
568 struct rt6_info *rt;
569 int hash;
570 int err = 0;
572 rcu_read_lock_bh();
573 if (idev->dead) {
574 err = -ENODEV; /*XXX*/
575 goto out2;
578 write_lock(&addrconf_hash_lock);
580 /* Ignore adding duplicate addresses on an interface */
581 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
582 ADBG(("ipv6_add_addr: already assigned\n"));
583 err = -EEXIST;
584 goto out;
587 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
589 if (ifa == NULL) {
590 ADBG(("ipv6_add_addr: malloc failed\n"));
591 err = -ENOBUFS;
592 goto out;
595 rt = addrconf_dst_alloc(idev, addr, 0);
596 if (IS_ERR(rt)) {
597 err = PTR_ERR(rt);
598 goto out;
601 ipv6_addr_copy(&ifa->addr, addr);
603 spin_lock_init(&ifa->lock);
604 init_timer(&ifa->timer);
605 ifa->timer.data = (unsigned long) ifa;
606 ifa->scope = scope;
607 ifa->prefix_len = pfxlen;
608 ifa->flags = flags | IFA_F_TENTATIVE;
609 ifa->cstamp = ifa->tstamp = jiffies;
611 ifa->rt = rt;
614 * part one of RFC 4429, section 3.3
615 * We should not configure an address as
616 * optimistic if we do not yet know the link
617 * layer address of our nexhop router
620 if (rt->rt6i_nexthop == NULL)
621 ifa->flags &= ~IFA_F_OPTIMISTIC;
623 ifa->idev = idev;
624 in6_dev_hold(idev);
625 /* For caller */
626 in6_ifa_hold(ifa);
628 /* Add to big hash table */
629 hash = ipv6_addr_hash(addr);
631 ifa->lst_next = inet6_addr_lst[hash];
632 inet6_addr_lst[hash] = ifa;
633 in6_ifa_hold(ifa);
634 write_unlock(&addrconf_hash_lock);
636 write_lock(&idev->lock);
637 /* Add to inet6_dev unicast addr list. */
638 ipv6_link_dev_addr(idev, ifa);
640 #ifdef CONFIG_IPV6_PRIVACY
641 if (ifa->flags&IFA_F_TEMPORARY) {
642 ifa->tmp_next = idev->tempaddr_list;
643 idev->tempaddr_list = ifa;
644 in6_ifa_hold(ifa);
646 #endif
648 in6_ifa_hold(ifa);
649 write_unlock(&idev->lock);
650 out2:
651 rcu_read_unlock_bh();
653 if (likely(err == 0))
654 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
655 else {
656 kfree(ifa);
657 ifa = ERR_PTR(err);
660 return ifa;
661 out:
662 write_unlock(&addrconf_hash_lock);
663 goto out2;
666 /* This function wants to get referenced ifp and releases it before return */
668 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
670 struct inet6_ifaddr *ifa, **ifap;
671 struct inet6_dev *idev = ifp->idev;
672 int hash;
673 int deleted = 0, onlink = 0;
674 unsigned long expires = jiffies;
676 hash = ipv6_addr_hash(&ifp->addr);
678 ifp->dead = 1;
680 write_lock_bh(&addrconf_hash_lock);
681 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
682 ifap = &ifa->lst_next) {
683 if (ifa == ifp) {
684 *ifap = ifa->lst_next;
685 __in6_ifa_put(ifp);
686 ifa->lst_next = NULL;
687 break;
690 write_unlock_bh(&addrconf_hash_lock);
692 write_lock_bh(&idev->lock);
693 #ifdef CONFIG_IPV6_PRIVACY
694 if (ifp->flags&IFA_F_TEMPORARY) {
695 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
696 ifap = &ifa->tmp_next) {
697 if (ifa == ifp) {
698 *ifap = ifa->tmp_next;
699 if (ifp->ifpub) {
700 in6_ifa_put(ifp->ifpub);
701 ifp->ifpub = NULL;
703 __in6_ifa_put(ifp);
704 ifa->tmp_next = NULL;
705 break;
709 #endif
711 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
712 if (ifa == ifp) {
713 *ifap = ifa->if_next;
714 __in6_ifa_put(ifp);
715 ifa->if_next = NULL;
716 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
717 break;
718 deleted = 1;
719 continue;
720 } else if (ifp->flags & IFA_F_PERMANENT) {
721 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
722 ifp->prefix_len)) {
723 if (ifa->flags & IFA_F_PERMANENT) {
724 onlink = 1;
725 if (deleted)
726 break;
727 } else {
728 unsigned long lifetime;
730 if (!onlink)
731 onlink = -1;
733 spin_lock(&ifa->lock);
734 lifetime = min_t(unsigned long,
735 ifa->valid_lft, 0x7fffffffUL/HZ);
736 if (time_before(expires,
737 ifa->tstamp + lifetime * HZ))
738 expires = ifa->tstamp + lifetime * HZ;
739 spin_unlock(&ifa->lock);
743 ifap = &ifa->if_next;
745 write_unlock_bh(&idev->lock);
747 ipv6_ifa_notify(RTM_DELADDR, ifp);
749 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
751 addrconf_del_timer(ifp);
754 * Purge or update corresponding prefix
756 * 1) we don't purge prefix here if address was not permanent.
757 * prefix is managed by its own lifetime.
758 * 2) if there're no addresses, delete prefix.
759 * 3) if there're still other permanent address(es),
760 * corresponding prefix is still permanent.
761 * 4) otherwise, update prefix lifetime to the
762 * longest valid lifetime among the corresponding
763 * addresses on the device.
764 * Note: subsequent RA will update lifetime.
766 * --yoshfuji
768 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
769 struct in6_addr prefix;
770 struct rt6_info *rt;
771 struct net *net = dev_net(ifp->idev->dev);
772 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
773 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
775 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
776 if (onlink == 0) {
777 ip6_del_rt(rt);
778 rt = NULL;
779 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
780 rt->rt6i_expires = expires;
781 rt->rt6i_flags |= RTF_EXPIRES;
784 dst_release(&rt->u.dst);
787 in6_ifa_put(ifp);
790 #ifdef CONFIG_IPV6_PRIVACY
791 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
793 struct inet6_dev *idev = ifp->idev;
794 struct in6_addr addr, *tmpaddr;
795 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
796 unsigned long regen_advance;
797 int tmp_plen;
798 int ret = 0;
799 int max_addresses;
800 u32 addr_flags;
802 write_lock(&idev->lock);
803 if (ift) {
804 spin_lock_bh(&ift->lock);
805 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
806 spin_unlock_bh(&ift->lock);
807 tmpaddr = &addr;
808 } else {
809 tmpaddr = NULL;
811 retry:
812 in6_dev_hold(idev);
813 if (idev->cnf.use_tempaddr <= 0) {
814 write_unlock(&idev->lock);
815 printk(KERN_INFO
816 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
817 in6_dev_put(idev);
818 ret = -1;
819 goto out;
821 spin_lock_bh(&ifp->lock);
822 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
823 idev->cnf.use_tempaddr = -1; /*XXX*/
824 spin_unlock_bh(&ifp->lock);
825 write_unlock(&idev->lock);
826 printk(KERN_WARNING
827 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
828 in6_dev_put(idev);
829 ret = -1;
830 goto out;
832 in6_ifa_hold(ifp);
833 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
834 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
835 spin_unlock_bh(&ifp->lock);
836 write_unlock(&idev->lock);
837 printk(KERN_WARNING
838 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
839 in6_ifa_put(ifp);
840 in6_dev_put(idev);
841 ret = -1;
842 goto out;
844 memcpy(&addr.s6_addr[8], idev->rndid, 8);
845 tmp_valid_lft = min_t(__u32,
846 ifp->valid_lft,
847 idev->cnf.temp_valid_lft);
848 tmp_prefered_lft = min_t(__u32,
849 ifp->prefered_lft,
850 idev->cnf.temp_prefered_lft - desync_factor / HZ);
851 tmp_plen = ifp->prefix_len;
852 max_addresses = idev->cnf.max_addresses;
853 tmp_cstamp = ifp->cstamp;
854 tmp_tstamp = ifp->tstamp;
855 spin_unlock_bh(&ifp->lock);
857 regen_advance = idev->cnf.regen_max_retry *
858 idev->cnf.dad_transmits *
859 idev->nd_parms->retrans_time / HZ;
860 write_unlock(&idev->lock);
862 /* A temporary address is created only if this calculated Preferred
863 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
864 * an implementation must not create a temporary address with a zero
865 * Preferred Lifetime.
867 if (tmp_prefered_lft <= regen_advance) {
868 in6_ifa_put(ifp);
869 in6_dev_put(idev);
870 ret = -1;
871 goto out;
874 addr_flags = IFA_F_TEMPORARY;
875 /* set in addrconf_prefix_rcv() */
876 if (ifp->flags & IFA_F_OPTIMISTIC)
877 addr_flags |= IFA_F_OPTIMISTIC;
879 ift = !max_addresses ||
880 ipv6_count_addresses(idev) < max_addresses ?
881 ipv6_add_addr(idev, &addr, tmp_plen,
882 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
883 addr_flags) : NULL;
884 if (!ift || IS_ERR(ift)) {
885 in6_ifa_put(ifp);
886 in6_dev_put(idev);
887 printk(KERN_INFO
888 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
889 tmpaddr = &addr;
890 write_lock(&idev->lock);
891 goto retry;
894 spin_lock_bh(&ift->lock);
895 ift->ifpub = ifp;
896 ift->valid_lft = tmp_valid_lft;
897 ift->prefered_lft = tmp_prefered_lft;
898 ift->cstamp = tmp_cstamp;
899 ift->tstamp = tmp_tstamp;
900 spin_unlock_bh(&ift->lock);
902 addrconf_dad_start(ift, 0);
903 in6_ifa_put(ift);
904 in6_dev_put(idev);
905 out:
906 return ret;
908 #endif
911 * Choose an appropriate source address (RFC3484)
913 enum {
914 IPV6_SADDR_RULE_INIT = 0,
915 IPV6_SADDR_RULE_LOCAL,
916 IPV6_SADDR_RULE_SCOPE,
917 IPV6_SADDR_RULE_PREFERRED,
918 #ifdef CONFIG_IPV6_MIP6
919 IPV6_SADDR_RULE_HOA,
920 #endif
921 IPV6_SADDR_RULE_OIF,
922 IPV6_SADDR_RULE_LABEL,
923 #ifdef CONFIG_IPV6_PRIVACY
924 IPV6_SADDR_RULE_PRIVACY,
925 #endif
926 IPV6_SADDR_RULE_ORCHID,
927 IPV6_SADDR_RULE_PREFIX,
928 IPV6_SADDR_RULE_MAX
931 struct ipv6_saddr_score {
932 int rule;
933 int addr_type;
934 struct inet6_ifaddr *ifa;
935 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
936 int scopedist;
937 int matchlen;
940 struct ipv6_saddr_dst {
941 const struct in6_addr *addr;
942 int ifindex;
943 int scope;
944 int label;
945 unsigned int prefs;
948 static inline int ipv6_saddr_preferred(int type)
950 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
951 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
952 return 1;
953 return 0;
956 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
957 struct ipv6_saddr_dst *dst,
958 int i)
960 int ret;
962 if (i <= score->rule) {
963 switch (i) {
964 case IPV6_SADDR_RULE_SCOPE:
965 ret = score->scopedist;
966 break;
967 case IPV6_SADDR_RULE_PREFIX:
968 ret = score->matchlen;
969 break;
970 default:
971 ret = !!test_bit(i, score->scorebits);
973 goto out;
976 switch (i) {
977 case IPV6_SADDR_RULE_INIT:
978 /* Rule 0: remember if hiscore is not ready yet */
979 ret = !!score->ifa;
980 break;
981 case IPV6_SADDR_RULE_LOCAL:
982 /* Rule 1: Prefer same address */
983 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
984 break;
985 case IPV6_SADDR_RULE_SCOPE:
986 /* Rule 2: Prefer appropriate scope
988 * ret
990 * -1 | d 15
991 * ---+--+-+---> scope
993 * | d is scope of the destination.
994 * B-d | \
995 * | \ <- smaller scope is better if
996 * B-15 | \ if scope is enough for destinaion.
997 * | ret = B - scope (-1 <= scope >= d <= 15).
998 * d-C-1 | /
999 * |/ <- greater is better
1000 * -C / if scope is not enough for destination.
1001 * /| ret = scope - C (-1 <= d < scope <= 15).
1003 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1004 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1005 * Assume B = 0 and we get C > 29.
1007 ret = __ipv6_addr_src_scope(score->addr_type);
1008 if (ret >= dst->scope)
1009 ret = -ret;
1010 else
1011 ret -= 128; /* 30 is enough */
1012 score->scopedist = ret;
1013 break;
1014 case IPV6_SADDR_RULE_PREFERRED:
1015 /* Rule 3: Avoid deprecated and optimistic addresses */
1016 ret = ipv6_saddr_preferred(score->addr_type) ||
1017 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1018 break;
1019 #ifdef CONFIG_IPV6_MIP6
1020 case IPV6_SADDR_RULE_HOA:
1022 /* Rule 4: Prefer home address */
1023 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1024 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1025 break;
1027 #endif
1028 case IPV6_SADDR_RULE_OIF:
1029 /* Rule 5: Prefer outgoing interface */
1030 ret = (!dst->ifindex ||
1031 dst->ifindex == score->ifa->idev->dev->ifindex);
1032 break;
1033 case IPV6_SADDR_RULE_LABEL:
1034 /* Rule 6: Prefer matching label */
1035 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
1036 score->ifa->idev->dev->ifindex) == dst->label;
1037 break;
1038 #ifdef CONFIG_IPV6_PRIVACY
1039 case IPV6_SADDR_RULE_PRIVACY:
1041 /* Rule 7: Prefer public address
1042 * Note: prefer temprary address if use_tempaddr >= 2
1044 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1045 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1046 score->ifa->idev->cnf.use_tempaddr >= 2;
1047 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1048 break;
1050 #endif
1051 case IPV6_SADDR_RULE_ORCHID:
1052 /* Rule 8-: Prefer ORCHID vs ORCHID or
1053 * non-ORCHID vs non-ORCHID
1055 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1056 ipv6_addr_orchid(dst->addr));
1057 break;
1058 case IPV6_SADDR_RULE_PREFIX:
1059 /* Rule 8: Use longest matching prefix */
1060 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1061 dst->addr);
1062 break;
1063 default:
1064 ret = 0;
1067 if (ret)
1068 __set_bit(i, score->scorebits);
1069 score->rule = i;
1070 out:
1071 return ret;
1074 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1075 const struct in6_addr *daddr, unsigned int prefs,
1076 struct in6_addr *saddr)
1078 struct ipv6_saddr_score scores[2],
1079 *score = &scores[0], *hiscore = &scores[1];
1080 struct net *net = dev_net(dst_dev);
1081 struct ipv6_saddr_dst dst;
1082 struct net_device *dev;
1083 int dst_type;
1085 dst_type = __ipv6_addr_type(daddr);
1086 dst.addr = daddr;
1087 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1088 dst.scope = __ipv6_addr_src_scope(dst_type);
1089 dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1090 dst.prefs = prefs;
1092 hiscore->rule = -1;
1093 hiscore->ifa = NULL;
1095 read_lock(&dev_base_lock);
1096 rcu_read_lock();
1098 for_each_netdev(net, dev) {
1099 struct inet6_dev *idev;
1101 /* Candidate Source Address (section 4)
1102 * - multicast and link-local destination address,
1103 * the set of candidate source address MUST only
1104 * include addresses assigned to interfaces
1105 * belonging to the same link as the outgoing
1106 * interface.
1107 * (- For site-local destination addresses, the
1108 * set of candidate source addresses MUST only
1109 * include addresses assigned to interfaces
1110 * belonging to the same site as the outgoing
1111 * interface.)
1113 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1114 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1115 dst.ifindex && dev->ifindex != dst.ifindex)
1116 continue;
1118 idev = __in6_dev_get(dev);
1119 if (!idev)
1120 continue;
1122 read_lock_bh(&idev->lock);
1123 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1124 int i;
1127 * - Tentative Address (RFC2462 section 5.4)
1128 * - A tentative address is not considered
1129 * "assigned to an interface" in the traditional
1130 * sense, unless it is also flagged as optimistic.
1131 * - Candidate Source Address (section 4)
1132 * - In any case, anycast addresses, multicast
1133 * addresses, and the unspecified address MUST
1134 * NOT be included in a candidate set.
1136 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1137 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1138 continue;
1140 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1142 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1143 score->addr_type & IPV6_ADDR_MULTICAST)) {
1144 LIMIT_NETDEBUG(KERN_DEBUG
1145 "ADDRCONF: unspecified / multicast address "
1146 "assigned as unicast address on %s",
1147 dev->name);
1148 continue;
1151 score->rule = -1;
1152 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1154 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1155 int minihiscore, miniscore;
1157 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1158 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1160 if (minihiscore > miniscore) {
1161 if (i == IPV6_SADDR_RULE_SCOPE &&
1162 score->scopedist > 0) {
1164 * special case:
1165 * each remaining entry
1166 * has too small (not enough)
1167 * scope, because ifa entries
1168 * are sorted by their scope
1169 * values.
1171 goto try_nextdev;
1173 break;
1174 } else if (minihiscore < miniscore) {
1175 struct ipv6_saddr_score *tmp;
1177 if (hiscore->ifa)
1178 in6_ifa_put(hiscore->ifa);
1180 in6_ifa_hold(score->ifa);
1182 tmp = hiscore;
1183 hiscore = score;
1184 score = tmp;
1186 /* restore our iterator */
1187 score->ifa = hiscore->ifa;
1189 break;
1193 try_nextdev:
1194 read_unlock_bh(&idev->lock);
1196 rcu_read_unlock();
1197 read_unlock(&dev_base_lock);
1199 if (!hiscore->ifa)
1200 return -EADDRNOTAVAIL;
1202 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1203 in6_ifa_put(hiscore->ifa);
1204 return 0;
1207 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1209 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1210 unsigned char banned_flags)
1212 struct inet6_dev *idev;
1213 int err = -EADDRNOTAVAIL;
1215 rcu_read_lock();
1216 if ((idev = __in6_dev_get(dev)) != NULL) {
1217 struct inet6_ifaddr *ifp;
1219 read_lock_bh(&idev->lock);
1220 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1221 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1222 ipv6_addr_copy(addr, &ifp->addr);
1223 err = 0;
1224 break;
1227 read_unlock_bh(&idev->lock);
1229 rcu_read_unlock();
1230 return err;
1233 static int ipv6_count_addresses(struct inet6_dev *idev)
1235 int cnt = 0;
1236 struct inet6_ifaddr *ifp;
1238 read_lock_bh(&idev->lock);
1239 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1240 cnt++;
1241 read_unlock_bh(&idev->lock);
1242 return cnt;
1245 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1246 struct net_device *dev, int strict)
1248 struct inet6_ifaddr * ifp;
1249 u8 hash = ipv6_addr_hash(addr);
1251 read_lock_bh(&addrconf_hash_lock);
1252 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1253 if (!net_eq(dev_net(ifp->idev->dev), net))
1254 continue;
1255 if (ipv6_addr_equal(&ifp->addr, addr) &&
1256 !(ifp->flags&IFA_F_TENTATIVE)) {
1257 if (dev == NULL || ifp->idev->dev == dev ||
1258 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1259 break;
1262 read_unlock_bh(&addrconf_hash_lock);
1263 return ifp != NULL;
1265 EXPORT_SYMBOL(ipv6_chk_addr);
1267 static
1268 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1269 struct net_device *dev)
1271 struct inet6_ifaddr * ifp;
1272 u8 hash = ipv6_addr_hash(addr);
1274 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1275 if (!net_eq(dev_net(ifp->idev->dev), net))
1276 continue;
1277 if (ipv6_addr_equal(&ifp->addr, addr)) {
1278 if (dev == NULL || ifp->idev->dev == dev)
1279 break;
1282 return ifp != NULL;
1285 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1287 struct inet6_dev *idev;
1288 struct inet6_ifaddr *ifa;
1289 int onlink;
1291 onlink = 0;
1292 rcu_read_lock();
1293 idev = __in6_dev_get(dev);
1294 if (idev) {
1295 read_lock_bh(&idev->lock);
1296 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1297 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1298 ifa->prefix_len);
1299 if (onlink)
1300 break;
1302 read_unlock_bh(&idev->lock);
1304 rcu_read_unlock();
1305 return onlink;
1308 EXPORT_SYMBOL(ipv6_chk_prefix);
1310 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1311 struct net_device *dev, int strict)
1313 struct inet6_ifaddr * ifp;
1314 u8 hash = ipv6_addr_hash(addr);
1316 read_lock_bh(&addrconf_hash_lock);
1317 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1318 if (!net_eq(dev_net(ifp->idev->dev), net))
1319 continue;
1320 if (ipv6_addr_equal(&ifp->addr, addr)) {
1321 if (dev == NULL || ifp->idev->dev == dev ||
1322 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1323 in6_ifa_hold(ifp);
1324 break;
1328 read_unlock_bh(&addrconf_hash_lock);
1330 return ifp;
1333 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1335 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1336 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1337 __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1338 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1339 int sk_ipv6only = ipv6_only_sock(sk);
1340 int sk2_ipv6only = inet_v6_ipv6only(sk2);
1341 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1342 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1344 if (!sk2_rcv_saddr && !sk_ipv6only)
1345 return 1;
1347 if (addr_type2 == IPV6_ADDR_ANY &&
1348 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1349 return 1;
1351 if (addr_type == IPV6_ADDR_ANY &&
1352 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1353 return 1;
1355 if (sk2_rcv_saddr6 &&
1356 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1357 return 1;
1359 if (addr_type == IPV6_ADDR_MAPPED &&
1360 !sk2_ipv6only &&
1361 (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1362 return 1;
1364 return 0;
1367 /* Gets referenced address, destroys ifaddr */
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1371 if (ifp->flags&IFA_F_PERMANENT) {
1372 spin_lock_bh(&ifp->lock);
1373 addrconf_del_timer(ifp);
1374 ifp->flags |= IFA_F_TENTATIVE;
1375 spin_unlock_bh(&ifp->lock);
1376 in6_ifa_put(ifp);
1377 #ifdef CONFIG_IPV6_PRIVACY
1378 } else if (ifp->flags&IFA_F_TEMPORARY) {
1379 struct inet6_ifaddr *ifpub;
1380 spin_lock_bh(&ifp->lock);
1381 ifpub = ifp->ifpub;
1382 if (ifpub) {
1383 in6_ifa_hold(ifpub);
1384 spin_unlock_bh(&ifp->lock);
1385 ipv6_create_tempaddr(ifpub, ifp);
1386 in6_ifa_put(ifpub);
1387 } else {
1388 spin_unlock_bh(&ifp->lock);
1390 ipv6_del_addr(ifp);
1391 #endif
1392 } else
1393 ipv6_del_addr(ifp);
1396 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1398 if (net_ratelimit())
1399 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1400 addrconf_dad_stop(ifp);
1403 /* Join to solicited addr multicast group. */
1405 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1407 struct in6_addr maddr;
1409 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1410 return;
1412 addrconf_addr_solict_mult(addr, &maddr);
1413 ipv6_dev_mc_inc(dev, &maddr);
1416 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1418 struct in6_addr maddr;
1420 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1421 return;
1423 addrconf_addr_solict_mult(addr, &maddr);
1424 __ipv6_dev_mc_dec(idev, &maddr);
1427 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1429 struct in6_addr addr;
1430 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1431 if (ipv6_addr_any(&addr))
1432 return;
1433 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1436 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1438 struct in6_addr addr;
1439 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1440 if (ipv6_addr_any(&addr))
1441 return;
1442 __ipv6_dev_ac_dec(ifp->idev, &addr);
1445 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1447 if (dev->addr_len != ETH_ALEN)
1448 return -1;
1449 memcpy(eui, dev->dev_addr, 3);
1450 memcpy(eui + 5, dev->dev_addr + 3, 3);
1453 * The zSeries OSA network cards can be shared among various
1454 * OS instances, but the OSA cards have only one MAC address.
1455 * This leads to duplicate address conflicts in conjunction
1456 * with IPv6 if more than one instance uses the same card.
1458 * The driver for these cards can deliver a unique 16-bit
1459 * identifier for each instance sharing the same card. It is
1460 * placed instead of 0xFFFE in the interface identifier. The
1461 * "u" bit of the interface identifier is not inverted in this
1462 * case. Hence the resulting interface identifier has local
1463 * scope according to RFC2373.
1465 if (dev->dev_id) {
1466 eui[3] = (dev->dev_id >> 8) & 0xFF;
1467 eui[4] = dev->dev_id & 0xFF;
1468 } else {
1469 eui[3] = 0xFF;
1470 eui[4] = 0xFE;
1471 eui[0] ^= 2;
1473 return 0;
1476 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1478 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1479 if (dev->addr_len != ARCNET_ALEN)
1480 return -1;
1481 memset(eui, 0, 7);
1482 eui[7] = *(u8*)dev->dev_addr;
1483 return 0;
1486 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1488 if (dev->addr_len != INFINIBAND_ALEN)
1489 return -1;
1490 memcpy(eui, dev->dev_addr + 12, 8);
1491 eui[0] |= 2;
1492 return 0;
1495 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1497 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1498 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1499 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1500 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1501 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1502 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1503 eui[1] = 0;
1504 eui[2] = 0x5E;
1505 eui[3] = 0xFE;
1506 memcpy(eui + 4, &addr, 4);
1507 return 0;
1509 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1511 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1513 if (dev->priv_flags & IFF_ISATAP)
1514 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1515 return -1;
1518 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1520 switch (dev->type) {
1521 case ARPHRD_ETHER:
1522 case ARPHRD_FDDI:
1523 case ARPHRD_IEEE802_TR:
1524 return addrconf_ifid_eui48(eui, dev);
1525 case ARPHRD_ARCNET:
1526 return addrconf_ifid_arcnet(eui, dev);
1527 case ARPHRD_INFINIBAND:
1528 return addrconf_ifid_infiniband(eui, dev);
1529 case ARPHRD_SIT:
1530 return addrconf_ifid_sit(eui, dev);
1532 return -1;
1535 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1537 int err = -1;
1538 struct inet6_ifaddr *ifp;
1540 read_lock_bh(&idev->lock);
1541 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1542 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1543 memcpy(eui, ifp->addr.s6_addr+8, 8);
1544 err = 0;
1545 break;
1548 read_unlock_bh(&idev->lock);
1549 return err;
1552 #ifdef CONFIG_IPV6_PRIVACY
1553 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1554 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1556 regen:
1557 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1558 idev->rndid[0] &= ~0x02;
1561 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1562 * check if generated address is not inappropriate
1564 * - Reserved subnet anycast (RFC 2526)
1565 * 11111101 11....11 1xxxxxxx
1566 * - ISATAP (RFC4214) 6.1
1567 * 00-00-5E-FE-xx-xx-xx-xx
1568 * - value 0
1569 * - XXX: already assigned to an address on the device
1571 if (idev->rndid[0] == 0xfd &&
1572 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1573 (idev->rndid[7]&0x80))
1574 goto regen;
1575 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1576 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1577 goto regen;
1578 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1579 goto regen;
1582 return 0;
1585 static void ipv6_regen_rndid(unsigned long data)
1587 struct inet6_dev *idev = (struct inet6_dev *) data;
1588 unsigned long expires;
1590 rcu_read_lock_bh();
1591 write_lock_bh(&idev->lock);
1593 if (idev->dead)
1594 goto out;
1596 if (__ipv6_regen_rndid(idev) < 0)
1597 goto out;
1599 expires = jiffies +
1600 idev->cnf.temp_prefered_lft * HZ -
1601 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1602 if (time_before(expires, jiffies)) {
1603 printk(KERN_WARNING
1604 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1605 idev->dev->name);
1606 goto out;
1609 if (!mod_timer(&idev->regen_timer, expires))
1610 in6_dev_hold(idev);
1612 out:
1613 write_unlock_bh(&idev->lock);
1614 rcu_read_unlock_bh();
1615 in6_dev_put(idev);
1618 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1619 int ret = 0;
1621 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1622 ret = __ipv6_regen_rndid(idev);
1623 return ret;
1625 #endif
1628 * Add prefix route.
1631 static void
1632 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1633 unsigned long expires, u32 flags)
1635 struct fib6_config cfg = {
1636 .fc_table = RT6_TABLE_PREFIX,
1637 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1638 .fc_ifindex = dev->ifindex,
1639 .fc_expires = expires,
1640 .fc_dst_len = plen,
1641 .fc_flags = RTF_UP | flags,
1642 .fc_nlinfo.nl_net = dev_net(dev),
1645 ipv6_addr_copy(&cfg.fc_dst, pfx);
1647 /* Prevent useless cloning on PtP SIT.
1648 This thing is done here expecting that the whole
1649 class of non-broadcast devices need not cloning.
1651 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1652 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1653 cfg.fc_flags |= RTF_NONEXTHOP;
1654 #endif
1656 ip6_route_add(&cfg);
1659 /* Create "default" multicast route to the interface */
1661 static void addrconf_add_mroute(struct net_device *dev)
1663 struct fib6_config cfg = {
1664 .fc_table = RT6_TABLE_LOCAL,
1665 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1666 .fc_ifindex = dev->ifindex,
1667 .fc_dst_len = 8,
1668 .fc_flags = RTF_UP,
1669 .fc_nlinfo.nl_net = dev_net(dev),
1672 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1674 ip6_route_add(&cfg);
1677 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1678 static void sit_route_add(struct net_device *dev)
1680 struct fib6_config cfg = {
1681 .fc_table = RT6_TABLE_MAIN,
1682 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1683 .fc_ifindex = dev->ifindex,
1684 .fc_dst_len = 96,
1685 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1686 .fc_nlinfo.nl_net = dev_net(dev),
1689 /* prefix length - 96 bits "::d.d.d.d" */
1690 ip6_route_add(&cfg);
1692 #endif
1694 static void addrconf_add_lroute(struct net_device *dev)
1696 struct in6_addr addr;
1698 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1699 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1702 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1704 struct inet6_dev *idev;
1706 ASSERT_RTNL();
1708 if ((idev = ipv6_find_idev(dev)) == NULL)
1709 return NULL;
1711 /* Add default multicast route */
1712 addrconf_add_mroute(dev);
1714 /* Add link local route */
1715 addrconf_add_lroute(dev);
1716 return idev;
1719 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1721 struct prefix_info *pinfo;
1722 __u32 valid_lft;
1723 __u32 prefered_lft;
1724 int addr_type;
1725 unsigned long rt_expires;
1726 struct inet6_dev *in6_dev;
1728 pinfo = (struct prefix_info *) opt;
1730 if (len < sizeof(struct prefix_info)) {
1731 ADBG(("addrconf: prefix option too short\n"));
1732 return;
1736 * Validation checks ([ADDRCONF], page 19)
1739 addr_type = ipv6_addr_type(&pinfo->prefix);
1741 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1742 return;
1744 valid_lft = ntohl(pinfo->valid);
1745 prefered_lft = ntohl(pinfo->prefered);
1747 if (prefered_lft > valid_lft) {
1748 if (net_ratelimit())
1749 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1750 return;
1753 in6_dev = in6_dev_get(dev);
1755 if (in6_dev == NULL) {
1756 if (net_ratelimit())
1757 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1758 return;
1762 * Two things going on here:
1763 * 1) Add routes for on-link prefixes
1764 * 2) Configure prefixes with the auto flag set
1767 if (valid_lft == INFINITY_LIFE_TIME)
1768 rt_expires = ~0UL;
1769 else if (valid_lft >= 0x7FFFFFFF/HZ) {
1770 /* Avoid arithmetic overflow. Really, we could
1771 * save rt_expires in seconds, likely valid_lft,
1772 * but it would require division in fib gc, that it
1773 * not good.
1775 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1776 } else
1777 rt_expires = valid_lft * HZ;
1780 * We convert this (in jiffies) to clock_t later.
1781 * Avoid arithmetic overflow there as well.
1782 * Overflow can happen only if HZ < USER_HZ.
1784 if (HZ < USER_HZ && ~rt_expires && rt_expires > 0x7FFFFFFF / USER_HZ)
1785 rt_expires = 0x7FFFFFFF / USER_HZ;
1787 if (pinfo->onlink) {
1788 struct rt6_info *rt;
1789 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1790 dev->ifindex, 1);
1792 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1793 /* Autoconf prefix route */
1794 if (valid_lft == 0) {
1795 ip6_del_rt(rt);
1796 rt = NULL;
1797 } else if (~rt_expires) {
1798 /* not infinity */
1799 rt->rt6i_expires = jiffies + rt_expires;
1800 rt->rt6i_flags |= RTF_EXPIRES;
1801 } else {
1802 rt->rt6i_flags &= ~RTF_EXPIRES;
1803 rt->rt6i_expires = 0;
1805 } else if (valid_lft) {
1806 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1807 clock_t expires = 0;
1808 if (~rt_expires) {
1809 /* not infinity */
1810 flags |= RTF_EXPIRES;
1811 expires = jiffies_to_clock_t(rt_expires);
1813 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1814 dev, expires, flags);
1816 if (rt)
1817 dst_release(&rt->u.dst);
1820 /* Try to figure out our local address for this prefix */
1822 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1823 struct inet6_ifaddr * ifp;
1824 struct in6_addr addr;
1825 int create = 0, update_lft = 0;
1827 if (pinfo->prefix_len == 64) {
1828 memcpy(&addr, &pinfo->prefix, 8);
1829 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1830 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1831 in6_dev_put(in6_dev);
1832 return;
1834 goto ok;
1836 if (net_ratelimit())
1837 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1838 pinfo->prefix_len);
1839 in6_dev_put(in6_dev);
1840 return;
1844 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1846 if (ifp == NULL && valid_lft) {
1847 int max_addresses = in6_dev->cnf.max_addresses;
1848 u32 addr_flags = 0;
1850 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1851 if (in6_dev->cnf.optimistic_dad &&
1852 !ipv6_devconf.forwarding)
1853 addr_flags = IFA_F_OPTIMISTIC;
1854 #endif
1856 /* Do not allow to create too much of autoconfigured
1857 * addresses; this would be too easy way to crash kernel.
1859 if (!max_addresses ||
1860 ipv6_count_addresses(in6_dev) < max_addresses)
1861 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1862 addr_type&IPV6_ADDR_SCOPE_MASK,
1863 addr_flags);
1865 if (!ifp || IS_ERR(ifp)) {
1866 in6_dev_put(in6_dev);
1867 return;
1870 update_lft = create = 1;
1871 ifp->cstamp = jiffies;
1872 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1875 if (ifp) {
1876 int flags;
1877 unsigned long now;
1878 #ifdef CONFIG_IPV6_PRIVACY
1879 struct inet6_ifaddr *ift;
1880 #endif
1881 u32 stored_lft;
1883 /* update lifetime (RFC2462 5.5.3 e) */
1884 spin_lock(&ifp->lock);
1885 now = jiffies;
1886 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1887 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1888 else
1889 stored_lft = 0;
1890 if (!update_lft && stored_lft) {
1891 if (valid_lft > MIN_VALID_LIFETIME ||
1892 valid_lft > stored_lft)
1893 update_lft = 1;
1894 else if (stored_lft <= MIN_VALID_LIFETIME) {
1895 /* valid_lft <= stored_lft is always true */
1896 /* XXX: IPsec */
1897 update_lft = 0;
1898 } else {
1899 valid_lft = MIN_VALID_LIFETIME;
1900 if (valid_lft < prefered_lft)
1901 prefered_lft = valid_lft;
1902 update_lft = 1;
1906 if (update_lft) {
1907 ifp->valid_lft = valid_lft;
1908 ifp->prefered_lft = prefered_lft;
1909 ifp->tstamp = now;
1910 flags = ifp->flags;
1911 ifp->flags &= ~IFA_F_DEPRECATED;
1912 spin_unlock(&ifp->lock);
1914 if (!(flags&IFA_F_TENTATIVE))
1915 ipv6_ifa_notify(0, ifp);
1916 } else
1917 spin_unlock(&ifp->lock);
1919 #ifdef CONFIG_IPV6_PRIVACY
1920 read_lock_bh(&in6_dev->lock);
1921 /* update all temporary addresses in the list */
1922 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1924 * When adjusting the lifetimes of an existing
1925 * temporary address, only lower the lifetimes.
1926 * Implementations must not increase the
1927 * lifetimes of an existing temporary address
1928 * when processing a Prefix Information Option.
1930 if (ifp != ift->ifpub)
1931 continue;
1933 spin_lock(&ift->lock);
1934 flags = ift->flags;
1935 if (ift->valid_lft > valid_lft &&
1936 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1937 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1938 if (ift->prefered_lft > prefered_lft &&
1939 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1940 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1941 spin_unlock(&ift->lock);
1942 if (!(flags&IFA_F_TENTATIVE))
1943 ipv6_ifa_notify(0, ift);
1946 if (create && in6_dev->cnf.use_tempaddr > 0) {
1948 * When a new public address is created as described in [ADDRCONF],
1949 * also create a new temporary address.
1951 read_unlock_bh(&in6_dev->lock);
1952 ipv6_create_tempaddr(ifp, NULL);
1953 } else {
1954 read_unlock_bh(&in6_dev->lock);
1956 #endif
1957 in6_ifa_put(ifp);
1958 addrconf_verify(0);
1961 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1962 in6_dev_put(in6_dev);
1966 * Set destination address.
1967 * Special case for SIT interfaces where we create a new "virtual"
1968 * device.
1970 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1972 struct in6_ifreq ireq;
1973 struct net_device *dev;
1974 int err = -EINVAL;
1976 rtnl_lock();
1978 err = -EFAULT;
1979 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1980 goto err_exit;
1982 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1984 err = -ENODEV;
1985 if (dev == NULL)
1986 goto err_exit;
1988 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1989 if (dev->type == ARPHRD_SIT) {
1990 struct ifreq ifr;
1991 mm_segment_t oldfs;
1992 struct ip_tunnel_parm p;
1994 err = -EADDRNOTAVAIL;
1995 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1996 goto err_exit;
1998 memset(&p, 0, sizeof(p));
1999 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2000 p.iph.saddr = 0;
2001 p.iph.version = 4;
2002 p.iph.ihl = 5;
2003 p.iph.protocol = IPPROTO_IPV6;
2004 p.iph.ttl = 64;
2005 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2007 oldfs = get_fs(); set_fs(KERNEL_DS);
2008 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2009 set_fs(oldfs);
2011 if (err == 0) {
2012 err = -ENOBUFS;
2013 dev = __dev_get_by_name(net, p.name);
2014 if (!dev)
2015 goto err_exit;
2016 err = dev_open(dev);
2019 #endif
2021 err_exit:
2022 rtnl_unlock();
2023 return err;
2027 * Manual configuration of address on an interface
2029 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2030 int plen, __u8 ifa_flags, __u32 prefered_lft,
2031 __u32 valid_lft)
2033 struct inet6_ifaddr *ifp;
2034 struct inet6_dev *idev;
2035 struct net_device *dev;
2036 int scope;
2037 u32 flags;
2038 clock_t expires;
2040 ASSERT_RTNL();
2042 /* check the lifetime */
2043 if (!valid_lft || prefered_lft > valid_lft)
2044 return -EINVAL;
2046 dev = __dev_get_by_index(net, ifindex);
2047 if (!dev)
2048 return -ENODEV;
2050 if ((idev = addrconf_add_dev(dev)) == NULL)
2051 return -ENOBUFS;
2053 scope = ipv6_addr_scope(pfx);
2055 if (valid_lft == INFINITY_LIFE_TIME) {
2056 ifa_flags |= IFA_F_PERMANENT;
2057 flags = 0;
2058 expires = 0;
2059 } else {
2060 if (valid_lft >= 0x7FFFFFFF/HZ)
2061 valid_lft = 0x7FFFFFFF/HZ;
2062 flags = RTF_EXPIRES;
2063 expires = jiffies_to_clock_t(valid_lft * HZ);
2066 if (prefered_lft == 0)
2067 ifa_flags |= IFA_F_DEPRECATED;
2068 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2069 (prefered_lft != INFINITY_LIFE_TIME))
2070 prefered_lft = 0x7FFFFFFF/HZ;
2072 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2074 if (!IS_ERR(ifp)) {
2075 spin_lock_bh(&ifp->lock);
2076 ifp->valid_lft = valid_lft;
2077 ifp->prefered_lft = prefered_lft;
2078 ifp->tstamp = jiffies;
2079 spin_unlock_bh(&ifp->lock);
2081 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2082 expires, flags);
2084 * Note that section 3.1 of RFC 4429 indicates
2085 * that the Optimistic flag should not be set for
2086 * manually configured addresses
2088 addrconf_dad_start(ifp, 0);
2089 in6_ifa_put(ifp);
2090 addrconf_verify(0);
2091 return 0;
2094 return PTR_ERR(ifp);
2097 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2098 int plen)
2100 struct inet6_ifaddr *ifp;
2101 struct inet6_dev *idev;
2102 struct net_device *dev;
2104 dev = __dev_get_by_index(net, ifindex);
2105 if (!dev)
2106 return -ENODEV;
2108 if ((idev = __in6_dev_get(dev)) == NULL)
2109 return -ENXIO;
2111 read_lock_bh(&idev->lock);
2112 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2113 if (ifp->prefix_len == plen &&
2114 ipv6_addr_equal(pfx, &ifp->addr)) {
2115 in6_ifa_hold(ifp);
2116 read_unlock_bh(&idev->lock);
2118 ipv6_del_addr(ifp);
2120 /* If the last address is deleted administratively,
2121 disable IPv6 on this interface.
2123 if (idev->addr_list == NULL)
2124 addrconf_ifdown(idev->dev, 1);
2125 return 0;
2128 read_unlock_bh(&idev->lock);
2129 return -EADDRNOTAVAIL;
2133 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2135 struct in6_ifreq ireq;
2136 int err;
2138 if (!capable(CAP_NET_ADMIN))
2139 return -EPERM;
2141 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2142 return -EFAULT;
2144 rtnl_lock();
2145 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2146 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2147 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2148 rtnl_unlock();
2149 return err;
2152 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2154 struct in6_ifreq ireq;
2155 int err;
2157 if (!capable(CAP_NET_ADMIN))
2158 return -EPERM;
2160 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2161 return -EFAULT;
2163 rtnl_lock();
2164 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2165 ireq.ifr6_prefixlen);
2166 rtnl_unlock();
2167 return err;
2170 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2171 static void sit_add_v4_addrs(struct inet6_dev *idev)
2173 struct inet6_ifaddr * ifp;
2174 struct in6_addr addr;
2175 struct net_device *dev;
2176 struct net *net = dev_net(idev->dev);
2177 int scope;
2179 ASSERT_RTNL();
2181 memset(&addr, 0, sizeof(struct in6_addr));
2182 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2184 if (idev->dev->flags&IFF_POINTOPOINT) {
2185 addr.s6_addr32[0] = htonl(0xfe800000);
2186 scope = IFA_LINK;
2187 } else {
2188 scope = IPV6_ADDR_COMPATv4;
2191 if (addr.s6_addr32[3]) {
2192 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2193 if (!IS_ERR(ifp)) {
2194 spin_lock_bh(&ifp->lock);
2195 ifp->flags &= ~IFA_F_TENTATIVE;
2196 spin_unlock_bh(&ifp->lock);
2197 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2198 in6_ifa_put(ifp);
2200 return;
2203 for_each_netdev(net, dev) {
2204 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2205 if (in_dev && (dev->flags & IFF_UP)) {
2206 struct in_ifaddr * ifa;
2208 int flag = scope;
2210 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2211 int plen;
2213 addr.s6_addr32[3] = ifa->ifa_local;
2215 if (ifa->ifa_scope == RT_SCOPE_LINK)
2216 continue;
2217 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2218 if (idev->dev->flags&IFF_POINTOPOINT)
2219 continue;
2220 flag |= IFA_HOST;
2222 if (idev->dev->flags&IFF_POINTOPOINT)
2223 plen = 64;
2224 else
2225 plen = 96;
2227 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2228 IFA_F_PERMANENT);
2229 if (!IS_ERR(ifp)) {
2230 spin_lock_bh(&ifp->lock);
2231 ifp->flags &= ~IFA_F_TENTATIVE;
2232 spin_unlock_bh(&ifp->lock);
2233 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2234 in6_ifa_put(ifp);
2240 #endif
2242 static void init_loopback(struct net_device *dev)
2244 struct inet6_dev *idev;
2245 struct inet6_ifaddr * ifp;
2247 /* ::1 */
2249 ASSERT_RTNL();
2251 if ((idev = ipv6_find_idev(dev)) == NULL) {
2252 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2253 return;
2256 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2257 if (!IS_ERR(ifp)) {
2258 spin_lock_bh(&ifp->lock);
2259 ifp->flags &= ~IFA_F_TENTATIVE;
2260 spin_unlock_bh(&ifp->lock);
2261 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2262 in6_ifa_put(ifp);
2266 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2268 struct inet6_ifaddr * ifp;
2269 u32 addr_flags = IFA_F_PERMANENT;
2271 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2272 if (idev->cnf.optimistic_dad &&
2273 !ipv6_devconf.forwarding)
2274 addr_flags |= IFA_F_OPTIMISTIC;
2275 #endif
2278 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2279 if (!IS_ERR(ifp)) {
2280 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2281 addrconf_dad_start(ifp, 0);
2282 in6_ifa_put(ifp);
2286 static void addrconf_dev_config(struct net_device *dev)
2288 struct in6_addr addr;
2289 struct inet6_dev * idev;
2291 ASSERT_RTNL();
2293 if ((dev->type != ARPHRD_ETHER) &&
2294 (dev->type != ARPHRD_FDDI) &&
2295 (dev->type != ARPHRD_IEEE802_TR) &&
2296 (dev->type != ARPHRD_ARCNET) &&
2297 (dev->type != ARPHRD_INFINIBAND)) {
2298 /* Alas, we support only Ethernet autoconfiguration. */
2299 return;
2302 idev = addrconf_add_dev(dev);
2303 if (idev == NULL)
2304 return;
2306 memset(&addr, 0, sizeof(struct in6_addr));
2307 addr.s6_addr32[0] = htonl(0xFE800000);
2309 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2310 addrconf_add_linklocal(idev, &addr);
2313 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2314 static void addrconf_sit_config(struct net_device *dev)
2316 struct inet6_dev *idev;
2318 ASSERT_RTNL();
2321 * Configure the tunnel with one of our IPv4
2322 * addresses... we should configure all of
2323 * our v4 addrs in the tunnel
2326 if ((idev = ipv6_find_idev(dev)) == NULL) {
2327 printk(KERN_DEBUG "init sit: add_dev failed\n");
2328 return;
2331 if (dev->priv_flags & IFF_ISATAP) {
2332 struct in6_addr addr;
2334 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2335 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2336 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2337 addrconf_add_linklocal(idev, &addr);
2338 return;
2341 sit_add_v4_addrs(idev);
2343 if (dev->flags&IFF_POINTOPOINT) {
2344 addrconf_add_mroute(dev);
2345 addrconf_add_lroute(dev);
2346 } else
2347 sit_route_add(dev);
2349 #endif
2351 static inline int
2352 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2354 struct in6_addr lladdr;
2356 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2357 addrconf_add_linklocal(idev, &lladdr);
2358 return 0;
2360 return -1;
2363 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2365 struct net_device *link_dev;
2366 struct net *net = dev_net(idev->dev);
2368 /* first try to inherit the link-local address from the link device */
2369 if (idev->dev->iflink &&
2370 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2371 if (!ipv6_inherit_linklocal(idev, link_dev))
2372 return;
2374 /* then try to inherit it from any device */
2375 for_each_netdev(net, link_dev) {
2376 if (!ipv6_inherit_linklocal(idev, link_dev))
2377 return;
2379 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2383 * Autoconfigure tunnel with a link-local address so routing protocols,
2384 * DHCPv6, MLD etc. can be run over the virtual link
2387 static void addrconf_ip6_tnl_config(struct net_device *dev)
2389 struct inet6_dev *idev;
2391 ASSERT_RTNL();
2393 if ((idev = addrconf_add_dev(dev)) == NULL) {
2394 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2395 return;
2397 ip6_tnl_add_linklocal(idev);
2400 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2401 void * data)
2403 struct net_device *dev = (struct net_device *) data;
2404 struct inet6_dev *idev = __in6_dev_get(dev);
2405 int run_pending = 0;
2406 int err;
2408 switch(event) {
2409 case NETDEV_REGISTER:
2410 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2411 idev = ipv6_add_dev(dev);
2412 if (!idev)
2413 return notifier_from_errno(-ENOMEM);
2415 break;
2416 case NETDEV_UP:
2417 case NETDEV_CHANGE:
2418 if (dev->flags & IFF_SLAVE)
2419 break;
2421 if (event == NETDEV_UP) {
2422 if (!addrconf_qdisc_ok(dev)) {
2423 /* device is not ready yet. */
2424 printk(KERN_INFO
2425 "ADDRCONF(NETDEV_UP): %s: "
2426 "link is not ready\n",
2427 dev->name);
2428 break;
2431 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2432 idev = ipv6_add_dev(dev);
2434 if (idev)
2435 idev->if_flags |= IF_READY;
2436 } else {
2437 if (!addrconf_qdisc_ok(dev)) {
2438 /* device is still not ready. */
2439 break;
2442 if (idev) {
2443 if (idev->if_flags & IF_READY) {
2444 /* device is already configured. */
2445 break;
2447 idev->if_flags |= IF_READY;
2450 printk(KERN_INFO
2451 "ADDRCONF(NETDEV_CHANGE): %s: "
2452 "link becomes ready\n",
2453 dev->name);
2455 run_pending = 1;
2458 switch(dev->type) {
2459 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2460 case ARPHRD_SIT:
2461 addrconf_sit_config(dev);
2462 break;
2463 #endif
2464 case ARPHRD_TUNNEL6:
2465 addrconf_ip6_tnl_config(dev);
2466 break;
2467 case ARPHRD_LOOPBACK:
2468 init_loopback(dev);
2469 break;
2471 default:
2472 addrconf_dev_config(dev);
2473 break;
2475 if (idev) {
2476 if (run_pending)
2477 addrconf_dad_run(idev);
2479 /* If the MTU changed during the interface down, when the
2480 interface up, the changed MTU must be reflected in the
2481 idev as well as routers.
2483 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2484 rt6_mtu_change(dev, dev->mtu);
2485 idev->cnf.mtu6 = dev->mtu;
2487 idev->tstamp = jiffies;
2488 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2489 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2490 stop IPv6 on this interface.
2492 if (dev->mtu < IPV6_MIN_MTU)
2493 addrconf_ifdown(dev, event != NETDEV_DOWN);
2495 break;
2497 case NETDEV_CHANGEMTU:
2498 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2499 rt6_mtu_change(dev, dev->mtu);
2500 idev->cnf.mtu6 = dev->mtu;
2501 break;
2504 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2505 idev = ipv6_add_dev(dev);
2506 if (idev)
2507 break;
2510 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2512 case NETDEV_DOWN:
2513 case NETDEV_UNREGISTER:
2515 * Remove all addresses from this interface.
2517 addrconf_ifdown(dev, event != NETDEV_DOWN);
2518 break;
2520 case NETDEV_CHANGENAME:
2521 if (idev) {
2522 snmp6_unregister_dev(idev);
2523 addrconf_sysctl_unregister(idev);
2524 addrconf_sysctl_register(idev);
2525 err = snmp6_register_dev(idev);
2526 if (err)
2527 return notifier_from_errno(err);
2529 break;
2532 return NOTIFY_OK;
2536 * addrconf module should be notified of a device going up
2538 static struct notifier_block ipv6_dev_notf = {
2539 .notifier_call = addrconf_notify,
2540 .priority = 0
2543 static int addrconf_ifdown(struct net_device *dev, int how)
2545 struct inet6_dev *idev;
2546 struct inet6_ifaddr *ifa, **bifa;
2547 struct net *net = dev_net(dev);
2548 int i;
2550 ASSERT_RTNL();
2552 if ((dev->flags & IFF_LOOPBACK) && how == 1)
2553 how = 0;
2555 rt6_ifdown(net, dev);
2556 neigh_ifdown(&nd_tbl, dev);
2558 idev = __in6_dev_get(dev);
2559 if (idev == NULL)
2560 return -ENODEV;
2562 /* Step 1: remove reference to ipv6 device from parent device.
2563 Do not dev_put!
2565 if (how) {
2566 idev->dead = 1;
2568 /* protected by rtnl_lock */
2569 rcu_assign_pointer(dev->ip6_ptr, NULL);
2571 /* Step 1.5: remove snmp6 entry */
2572 snmp6_unregister_dev(idev);
2576 /* Step 2: clear hash table */
2577 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2578 bifa = &inet6_addr_lst[i];
2580 write_lock_bh(&addrconf_hash_lock);
2581 while ((ifa = *bifa) != NULL) {
2582 if (ifa->idev == idev) {
2583 *bifa = ifa->lst_next;
2584 ifa->lst_next = NULL;
2585 addrconf_del_timer(ifa);
2586 in6_ifa_put(ifa);
2587 continue;
2589 bifa = &ifa->lst_next;
2591 write_unlock_bh(&addrconf_hash_lock);
2594 write_lock_bh(&idev->lock);
2596 /* Step 3: clear flags for stateless addrconf */
2597 if (!how)
2598 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2600 /* Step 4: clear address list */
2601 #ifdef CONFIG_IPV6_PRIVACY
2602 if (how && del_timer(&idev->regen_timer))
2603 in6_dev_put(idev);
2605 /* clear tempaddr list */
2606 while ((ifa = idev->tempaddr_list) != NULL) {
2607 idev->tempaddr_list = ifa->tmp_next;
2608 ifa->tmp_next = NULL;
2609 ifa->dead = 1;
2610 write_unlock_bh(&idev->lock);
2611 spin_lock_bh(&ifa->lock);
2613 if (ifa->ifpub) {
2614 in6_ifa_put(ifa->ifpub);
2615 ifa->ifpub = NULL;
2617 spin_unlock_bh(&ifa->lock);
2618 in6_ifa_put(ifa);
2619 write_lock_bh(&idev->lock);
2621 #endif
2622 while ((ifa = idev->addr_list) != NULL) {
2623 idev->addr_list = ifa->if_next;
2624 ifa->if_next = NULL;
2625 ifa->dead = 1;
2626 addrconf_del_timer(ifa);
2627 write_unlock_bh(&idev->lock);
2629 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2630 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2631 in6_ifa_put(ifa);
2633 write_lock_bh(&idev->lock);
2635 write_unlock_bh(&idev->lock);
2637 /* Step 5: Discard multicast list */
2639 if (how)
2640 ipv6_mc_destroy_dev(idev);
2641 else
2642 ipv6_mc_down(idev);
2644 idev->tstamp = jiffies;
2646 /* Shot the device (if unregistered) */
2648 if (how) {
2649 addrconf_sysctl_unregister(idev);
2650 neigh_parms_release(&nd_tbl, idev->nd_parms);
2651 neigh_ifdown(&nd_tbl, dev);
2652 in6_dev_put(idev);
2654 return 0;
2657 static void addrconf_rs_timer(unsigned long data)
2659 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2661 if (ifp->idev->cnf.forwarding)
2662 goto out;
2664 if (ifp->idev->if_flags & IF_RA_RCVD) {
2666 * Announcement received after solicitation
2667 * was sent
2669 goto out;
2672 spin_lock(&ifp->lock);
2673 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2674 /* The wait after the last probe can be shorter */
2675 addrconf_mod_timer(ifp, AC_RS,
2676 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2677 ifp->idev->cnf.rtr_solicit_delay :
2678 ifp->idev->cnf.rtr_solicit_interval);
2679 spin_unlock(&ifp->lock);
2681 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2682 } else {
2683 spin_unlock(&ifp->lock);
2685 * Note: we do not support deprecated "all on-link"
2686 * assumption any longer.
2688 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2689 ifp->idev->dev->name);
2692 out:
2693 in6_ifa_put(ifp);
2697 * Duplicate Address Detection
2699 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2701 unsigned long rand_num;
2702 struct inet6_dev *idev = ifp->idev;
2704 if (ifp->flags & IFA_F_OPTIMISTIC)
2705 rand_num = 0;
2706 else
2707 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2709 ifp->probes = idev->cnf.dad_transmits;
2710 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2713 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2715 struct inet6_dev *idev = ifp->idev;
2716 struct net_device *dev = idev->dev;
2718 addrconf_join_solict(dev, &ifp->addr);
2720 net_srandom(ifp->addr.s6_addr32[3]);
2722 read_lock_bh(&idev->lock);
2723 if (ifp->dead)
2724 goto out;
2725 spin_lock_bh(&ifp->lock);
2727 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2728 !(ifp->flags&IFA_F_TENTATIVE) ||
2729 ifp->flags & IFA_F_NODAD) {
2730 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2731 spin_unlock_bh(&ifp->lock);
2732 read_unlock_bh(&idev->lock);
2734 addrconf_dad_completed(ifp);
2735 return;
2738 if (!(idev->if_flags & IF_READY)) {
2739 spin_unlock_bh(&ifp->lock);
2740 read_unlock_bh(&idev->lock);
2742 * If the defice is not ready:
2743 * - keep it tentative if it is a permanent address.
2744 * - otherwise, kill it.
2746 in6_ifa_hold(ifp);
2747 addrconf_dad_stop(ifp);
2748 return;
2752 * Optimistic nodes can start receiving
2753 * Frames right away
2755 if(ifp->flags & IFA_F_OPTIMISTIC)
2756 ip6_ins_rt(ifp->rt);
2758 addrconf_dad_kick(ifp);
2759 spin_unlock_bh(&ifp->lock);
2760 out:
2761 read_unlock_bh(&idev->lock);
2764 static void addrconf_dad_timer(unsigned long data)
2766 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2767 struct inet6_dev *idev = ifp->idev;
2768 struct in6_addr mcaddr;
2770 read_lock_bh(&idev->lock);
2771 if (idev->dead) {
2772 read_unlock_bh(&idev->lock);
2773 goto out;
2775 spin_lock_bh(&ifp->lock);
2776 if (ifp->probes == 0) {
2778 * DAD was successful
2781 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2782 spin_unlock_bh(&ifp->lock);
2783 read_unlock_bh(&idev->lock);
2785 addrconf_dad_completed(ifp);
2787 goto out;
2790 ifp->probes--;
2791 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2792 spin_unlock_bh(&ifp->lock);
2793 read_unlock_bh(&idev->lock);
2795 /* send a neighbour solicitation for our addr */
2796 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2797 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2798 out:
2799 in6_ifa_put(ifp);
2802 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2804 struct net_device * dev = ifp->idev->dev;
2807 * Configure the address for reception. Now it is valid.
2810 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2812 /* If added prefix is link local and forwarding is off,
2813 start sending router solicitations.
2816 if (ifp->idev->cnf.forwarding == 0 &&
2817 ifp->idev->cnf.rtr_solicits > 0 &&
2818 (dev->flags&IFF_LOOPBACK) == 0 &&
2819 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2821 * If a host as already performed a random delay
2822 * [...] as part of DAD [...] there is no need
2823 * to delay again before sending the first RS
2825 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2827 spin_lock_bh(&ifp->lock);
2828 ifp->probes = 1;
2829 ifp->idev->if_flags |= IF_RS_SENT;
2830 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2831 spin_unlock_bh(&ifp->lock);
2835 static void addrconf_dad_run(struct inet6_dev *idev) {
2836 struct inet6_ifaddr *ifp;
2838 read_lock_bh(&idev->lock);
2839 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2840 spin_lock_bh(&ifp->lock);
2841 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2842 spin_unlock_bh(&ifp->lock);
2843 continue;
2845 spin_unlock_bh(&ifp->lock);
2846 addrconf_dad_kick(ifp);
2848 read_unlock_bh(&idev->lock);
2851 #ifdef CONFIG_PROC_FS
2852 struct if6_iter_state {
2853 struct seq_net_private p;
2854 int bucket;
2857 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2859 struct inet6_ifaddr *ifa = NULL;
2860 struct if6_iter_state *state = seq->private;
2861 struct net *net = seq_file_net(seq);
2863 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2864 ifa = inet6_addr_lst[state->bucket];
2866 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2867 ifa = ifa->lst_next;
2868 if (ifa)
2869 break;
2871 return ifa;
2874 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2876 struct if6_iter_state *state = seq->private;
2877 struct net *net = seq_file_net(seq);
2879 ifa = ifa->lst_next;
2880 try_again:
2881 if (ifa) {
2882 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2883 ifa = ifa->lst_next;
2884 goto try_again;
2888 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2889 ifa = inet6_addr_lst[state->bucket];
2890 goto try_again;
2893 return ifa;
2896 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2898 struct inet6_ifaddr *ifa = if6_get_first(seq);
2900 if (ifa)
2901 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2902 --pos;
2903 return pos ? NULL : ifa;
2906 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2907 __acquires(addrconf_hash_lock)
2909 read_lock_bh(&addrconf_hash_lock);
2910 return if6_get_idx(seq, *pos);
2913 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2915 struct inet6_ifaddr *ifa;
2917 ifa = if6_get_next(seq, v);
2918 ++*pos;
2919 return ifa;
2922 static void if6_seq_stop(struct seq_file *seq, void *v)
2923 __releases(addrconf_hash_lock)
2925 read_unlock_bh(&addrconf_hash_lock);
2928 static int if6_seq_show(struct seq_file *seq, void *v)
2930 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2931 seq_printf(seq,
2932 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2933 NIP6(ifp->addr),
2934 ifp->idev->dev->ifindex,
2935 ifp->prefix_len,
2936 ifp->scope,
2937 ifp->flags,
2938 ifp->idev->dev->name);
2939 return 0;
2942 static const struct seq_operations if6_seq_ops = {
2943 .start = if6_seq_start,
2944 .next = if6_seq_next,
2945 .show = if6_seq_show,
2946 .stop = if6_seq_stop,
2949 static int if6_seq_open(struct inode *inode, struct file *file)
2951 return seq_open_net(inode, file, &if6_seq_ops,
2952 sizeof(struct if6_iter_state));
2955 static const struct file_operations if6_fops = {
2956 .owner = THIS_MODULE,
2957 .open = if6_seq_open,
2958 .read = seq_read,
2959 .llseek = seq_lseek,
2960 .release = seq_release_net,
2963 static int if6_proc_net_init(struct net *net)
2965 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2966 return -ENOMEM;
2967 return 0;
2970 static void if6_proc_net_exit(struct net *net)
2972 proc_net_remove(net, "if_inet6");
2975 static struct pernet_operations if6_proc_net_ops = {
2976 .init = if6_proc_net_init,
2977 .exit = if6_proc_net_exit,
2980 int __init if6_proc_init(void)
2982 return register_pernet_subsys(&if6_proc_net_ops);
2985 void if6_proc_exit(void)
2987 unregister_pernet_subsys(&if6_proc_net_ops);
2989 #endif /* CONFIG_PROC_FS */
2991 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2992 /* Check if address is a home address configured on any interface. */
2993 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2995 int ret = 0;
2996 struct inet6_ifaddr * ifp;
2997 u8 hash = ipv6_addr_hash(addr);
2998 read_lock_bh(&addrconf_hash_lock);
2999 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3000 if (!net_eq(dev_net(ifp->idev->dev), net))
3001 continue;
3002 if (ipv6_addr_equal(&ifp->addr, addr) &&
3003 (ifp->flags & IFA_F_HOMEADDRESS)) {
3004 ret = 1;
3005 break;
3008 read_unlock_bh(&addrconf_hash_lock);
3009 return ret;
3011 #endif
3014 * Periodic address status verification
3017 static void addrconf_verify(unsigned long foo)
3019 struct inet6_ifaddr *ifp;
3020 unsigned long now, next;
3021 int i;
3023 spin_lock_bh(&addrconf_verify_lock);
3024 now = jiffies;
3025 next = now + ADDR_CHECK_FREQUENCY;
3027 del_timer(&addr_chk_timer);
3029 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3031 restart:
3032 read_lock(&addrconf_hash_lock);
3033 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3034 unsigned long age;
3035 #ifdef CONFIG_IPV6_PRIVACY
3036 unsigned long regen_advance;
3037 #endif
3039 if (ifp->flags & IFA_F_PERMANENT)
3040 continue;
3042 spin_lock(&ifp->lock);
3043 age = (now - ifp->tstamp) / HZ;
3045 #ifdef CONFIG_IPV6_PRIVACY
3046 regen_advance = ifp->idev->cnf.regen_max_retry *
3047 ifp->idev->cnf.dad_transmits *
3048 ifp->idev->nd_parms->retrans_time / HZ;
3049 #endif
3051 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3052 age >= ifp->valid_lft) {
3053 spin_unlock(&ifp->lock);
3054 in6_ifa_hold(ifp);
3055 read_unlock(&addrconf_hash_lock);
3056 ipv6_del_addr(ifp);
3057 goto restart;
3058 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3059 spin_unlock(&ifp->lock);
3060 continue;
3061 } else if (age >= ifp->prefered_lft) {
3062 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3063 int deprecate = 0;
3065 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3066 deprecate = 1;
3067 ifp->flags |= IFA_F_DEPRECATED;
3070 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3071 next = ifp->tstamp + ifp->valid_lft * HZ;
3073 spin_unlock(&ifp->lock);
3075 if (deprecate) {
3076 in6_ifa_hold(ifp);
3077 read_unlock(&addrconf_hash_lock);
3079 ipv6_ifa_notify(0, ifp);
3080 in6_ifa_put(ifp);
3081 goto restart;
3083 #ifdef CONFIG_IPV6_PRIVACY
3084 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3085 !(ifp->flags&IFA_F_TENTATIVE)) {
3086 if (age >= ifp->prefered_lft - regen_advance) {
3087 struct inet6_ifaddr *ifpub = ifp->ifpub;
3088 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3089 next = ifp->tstamp + ifp->prefered_lft * HZ;
3090 if (!ifp->regen_count && ifpub) {
3091 ifp->regen_count++;
3092 in6_ifa_hold(ifp);
3093 in6_ifa_hold(ifpub);
3094 spin_unlock(&ifp->lock);
3095 read_unlock(&addrconf_hash_lock);
3096 spin_lock(&ifpub->lock);
3097 ifpub->regen_count = 0;
3098 spin_unlock(&ifpub->lock);
3099 ipv6_create_tempaddr(ifpub, ifp);
3100 in6_ifa_put(ifpub);
3101 in6_ifa_put(ifp);
3102 goto restart;
3104 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3105 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3106 spin_unlock(&ifp->lock);
3107 #endif
3108 } else {
3109 /* ifp->prefered_lft <= ifp->valid_lft */
3110 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3111 next = ifp->tstamp + ifp->prefered_lft * HZ;
3112 spin_unlock(&ifp->lock);
3115 read_unlock(&addrconf_hash_lock);
3118 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3119 add_timer(&addr_chk_timer);
3120 spin_unlock_bh(&addrconf_verify_lock);
3123 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3125 struct in6_addr *pfx = NULL;
3127 if (addr)
3128 pfx = nla_data(addr);
3130 if (local) {
3131 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3132 pfx = NULL;
3133 else
3134 pfx = nla_data(local);
3137 return pfx;
3140 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3141 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3142 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3143 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3146 static int
3147 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3149 struct net *net = sock_net(skb->sk);
3150 struct ifaddrmsg *ifm;
3151 struct nlattr *tb[IFA_MAX+1];
3152 struct in6_addr *pfx;
3153 int err;
3155 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3156 if (err < 0)
3157 return err;
3159 ifm = nlmsg_data(nlh);
3160 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3161 if (pfx == NULL)
3162 return -EINVAL;
3164 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3167 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3168 u32 prefered_lft, u32 valid_lft)
3170 u32 flags;
3171 clock_t expires;
3173 if (!valid_lft || (prefered_lft > valid_lft))
3174 return -EINVAL;
3176 if (valid_lft == INFINITY_LIFE_TIME) {
3177 ifa_flags |= IFA_F_PERMANENT;
3178 flags = 0;
3179 expires = 0;
3180 } else {
3181 if (valid_lft >= 0x7FFFFFFF/HZ)
3182 valid_lft = 0x7FFFFFFF/HZ;
3183 flags = RTF_EXPIRES;
3184 expires = jiffies_to_clock_t(valid_lft * HZ);
3187 if (prefered_lft == 0)
3188 ifa_flags |= IFA_F_DEPRECATED;
3189 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3190 (prefered_lft != INFINITY_LIFE_TIME))
3191 prefered_lft = 0x7FFFFFFF/HZ;
3193 spin_lock_bh(&ifp->lock);
3194 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3195 ifp->tstamp = jiffies;
3196 ifp->valid_lft = valid_lft;
3197 ifp->prefered_lft = prefered_lft;
3199 spin_unlock_bh(&ifp->lock);
3200 if (!(ifp->flags&IFA_F_TENTATIVE))
3201 ipv6_ifa_notify(0, ifp);
3203 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3204 expires, flags);
3205 addrconf_verify(0);
3207 return 0;
3210 static int
3211 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3213 struct net *net = sock_net(skb->sk);
3214 struct ifaddrmsg *ifm;
3215 struct nlattr *tb[IFA_MAX+1];
3216 struct in6_addr *pfx;
3217 struct inet6_ifaddr *ifa;
3218 struct net_device *dev;
3219 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3220 u8 ifa_flags;
3221 int err;
3223 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3224 if (err < 0)
3225 return err;
3227 ifm = nlmsg_data(nlh);
3228 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3229 if (pfx == NULL)
3230 return -EINVAL;
3232 if (tb[IFA_CACHEINFO]) {
3233 struct ifa_cacheinfo *ci;
3235 ci = nla_data(tb[IFA_CACHEINFO]);
3236 valid_lft = ci->ifa_valid;
3237 preferred_lft = ci->ifa_prefered;
3238 } else {
3239 preferred_lft = INFINITY_LIFE_TIME;
3240 valid_lft = INFINITY_LIFE_TIME;
3243 dev = __dev_get_by_index(net, ifm->ifa_index);
3244 if (dev == NULL)
3245 return -ENODEV;
3247 /* We ignore other flags so far. */
3248 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3250 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3251 if (ifa == NULL) {
3253 * It would be best to check for !NLM_F_CREATE here but
3254 * userspace alreay relies on not having to provide this.
3256 return inet6_addr_add(net, ifm->ifa_index, pfx,
3257 ifm->ifa_prefixlen, ifa_flags,
3258 preferred_lft, valid_lft);
3261 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3262 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3263 err = -EEXIST;
3264 else
3265 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3267 in6_ifa_put(ifa);
3269 return err;
3272 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3273 u8 scope, int ifindex)
3275 struct ifaddrmsg *ifm;
3277 ifm = nlmsg_data(nlh);
3278 ifm->ifa_family = AF_INET6;
3279 ifm->ifa_prefixlen = prefixlen;
3280 ifm->ifa_flags = flags;
3281 ifm->ifa_scope = scope;
3282 ifm->ifa_index = ifindex;
3285 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3286 unsigned long tstamp, u32 preferred, u32 valid)
3288 struct ifa_cacheinfo ci;
3290 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3291 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3292 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3293 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3294 ci.ifa_prefered = preferred;
3295 ci.ifa_valid = valid;
3297 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3300 static inline int rt_scope(int ifa_scope)
3302 if (ifa_scope & IFA_HOST)
3303 return RT_SCOPE_HOST;
3304 else if (ifa_scope & IFA_LINK)
3305 return RT_SCOPE_LINK;
3306 else if (ifa_scope & IFA_SITE)
3307 return RT_SCOPE_SITE;
3308 else
3309 return RT_SCOPE_UNIVERSE;
3312 static inline int inet6_ifaddr_msgsize(void)
3314 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3315 + nla_total_size(16) /* IFA_ADDRESS */
3316 + nla_total_size(sizeof(struct ifa_cacheinfo));
3319 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3320 u32 pid, u32 seq, int event, unsigned int flags)
3322 struct nlmsghdr *nlh;
3323 u32 preferred, valid;
3325 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3326 if (nlh == NULL)
3327 return -EMSGSIZE;
3329 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3330 ifa->idev->dev->ifindex);
3332 if (!(ifa->flags&IFA_F_PERMANENT)) {
3333 preferred = ifa->prefered_lft;
3334 valid = ifa->valid_lft;
3335 if (preferred != INFINITY_LIFE_TIME) {
3336 long tval = (jiffies - ifa->tstamp)/HZ;
3337 preferred -= tval;
3338 if (valid != INFINITY_LIFE_TIME)
3339 valid -= tval;
3341 } else {
3342 preferred = INFINITY_LIFE_TIME;
3343 valid = INFINITY_LIFE_TIME;
3346 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3347 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3348 nlmsg_cancel(skb, nlh);
3349 return -EMSGSIZE;
3352 return nlmsg_end(skb, nlh);
3355 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3356 u32 pid, u32 seq, int event, u16 flags)
3358 struct nlmsghdr *nlh;
3359 u8 scope = RT_SCOPE_UNIVERSE;
3360 int ifindex = ifmca->idev->dev->ifindex;
3362 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3363 scope = RT_SCOPE_SITE;
3365 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3366 if (nlh == NULL)
3367 return -EMSGSIZE;
3369 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3370 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3371 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3372 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3373 nlmsg_cancel(skb, nlh);
3374 return -EMSGSIZE;
3377 return nlmsg_end(skb, nlh);
3380 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3381 u32 pid, u32 seq, int event, unsigned int flags)
3383 struct nlmsghdr *nlh;
3384 u8 scope = RT_SCOPE_UNIVERSE;
3385 int ifindex = ifaca->aca_idev->dev->ifindex;
3387 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3388 scope = RT_SCOPE_SITE;
3390 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3391 if (nlh == NULL)
3392 return -EMSGSIZE;
3394 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3395 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3396 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3397 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3398 nlmsg_cancel(skb, nlh);
3399 return -EMSGSIZE;
3402 return nlmsg_end(skb, nlh);
3405 enum addr_type_t
3407 UNICAST_ADDR,
3408 MULTICAST_ADDR,
3409 ANYCAST_ADDR,
3412 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3413 enum addr_type_t type)
3415 int idx, ip_idx;
3416 int s_idx, s_ip_idx;
3417 int err = 1;
3418 struct net_device *dev;
3419 struct inet6_dev *idev = NULL;
3420 struct inet6_ifaddr *ifa;
3421 struct ifmcaddr6 *ifmca;
3422 struct ifacaddr6 *ifaca;
3423 struct net *net = sock_net(skb->sk);
3425 s_idx = cb->args[0];
3426 s_ip_idx = ip_idx = cb->args[1];
3428 idx = 0;
3429 for_each_netdev(net, dev) {
3430 if (idx < s_idx)
3431 goto cont;
3432 if (idx > s_idx)
3433 s_ip_idx = 0;
3434 ip_idx = 0;
3435 if ((idev = in6_dev_get(dev)) == NULL)
3436 goto cont;
3437 read_lock_bh(&idev->lock);
3438 switch (type) {
3439 case UNICAST_ADDR:
3440 /* unicast address incl. temp addr */
3441 for (ifa = idev->addr_list; ifa;
3442 ifa = ifa->if_next, ip_idx++) {
3443 if (ip_idx < s_ip_idx)
3444 continue;
3445 err = inet6_fill_ifaddr(skb, ifa,
3446 NETLINK_CB(cb->skb).pid,
3447 cb->nlh->nlmsg_seq,
3448 RTM_NEWADDR,
3449 NLM_F_MULTI);
3451 break;
3452 case MULTICAST_ADDR:
3453 /* multicast address */
3454 for (ifmca = idev->mc_list; ifmca;
3455 ifmca = ifmca->next, ip_idx++) {
3456 if (ip_idx < s_ip_idx)
3457 continue;
3458 err = inet6_fill_ifmcaddr(skb, ifmca,
3459 NETLINK_CB(cb->skb).pid,
3460 cb->nlh->nlmsg_seq,
3461 RTM_GETMULTICAST,
3462 NLM_F_MULTI);
3464 break;
3465 case ANYCAST_ADDR:
3466 /* anycast address */
3467 for (ifaca = idev->ac_list; ifaca;
3468 ifaca = ifaca->aca_next, ip_idx++) {
3469 if (ip_idx < s_ip_idx)
3470 continue;
3471 err = inet6_fill_ifacaddr(skb, ifaca,
3472 NETLINK_CB(cb->skb).pid,
3473 cb->nlh->nlmsg_seq,
3474 RTM_GETANYCAST,
3475 NLM_F_MULTI);
3477 break;
3478 default:
3479 break;
3481 read_unlock_bh(&idev->lock);
3482 in6_dev_put(idev);
3484 if (err <= 0)
3485 break;
3486 cont:
3487 idx++;
3489 cb->args[0] = idx;
3490 cb->args[1] = ip_idx;
3491 return skb->len;
3494 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3496 enum addr_type_t type = UNICAST_ADDR;
3498 return inet6_dump_addr(skb, cb, type);
3501 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3503 enum addr_type_t type = MULTICAST_ADDR;
3505 return inet6_dump_addr(skb, cb, type);
3509 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3511 enum addr_type_t type = ANYCAST_ADDR;
3513 return inet6_dump_addr(skb, cb, type);
3516 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3517 void *arg)
3519 struct net *net = sock_net(in_skb->sk);
3520 struct ifaddrmsg *ifm;
3521 struct nlattr *tb[IFA_MAX+1];
3522 struct in6_addr *addr = NULL;
3523 struct net_device *dev = NULL;
3524 struct inet6_ifaddr *ifa;
3525 struct sk_buff *skb;
3526 int err;
3528 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3529 if (err < 0)
3530 goto errout;
3532 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3533 if (addr == NULL) {
3534 err = -EINVAL;
3535 goto errout;
3538 ifm = nlmsg_data(nlh);
3539 if (ifm->ifa_index)
3540 dev = __dev_get_by_index(net, ifm->ifa_index);
3542 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3543 err = -EADDRNOTAVAIL;
3544 goto errout;
3547 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3548 err = -ENOBUFS;
3549 goto errout_ifa;
3552 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3553 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3554 if (err < 0) {
3555 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3556 WARN_ON(err == -EMSGSIZE);
3557 kfree_skb(skb);
3558 goto errout_ifa;
3560 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3561 errout_ifa:
3562 in6_ifa_put(ifa);
3563 errout:
3564 return err;
3567 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3569 struct sk_buff *skb;
3570 struct net *net = dev_net(ifa->idev->dev);
3571 int err = -ENOBUFS;
3573 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3574 if (skb == NULL)
3575 goto errout;
3577 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3578 if (err < 0) {
3579 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3580 WARN_ON(err == -EMSGSIZE);
3581 kfree_skb(skb);
3582 goto errout;
3584 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3585 errout:
3586 if (err < 0)
3587 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3590 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3591 __s32 *array, int bytes)
3593 BUG_ON(bytes < (DEVCONF_MAX * 4));
3595 memset(array, 0, bytes);
3596 array[DEVCONF_FORWARDING] = cnf->forwarding;
3597 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3598 array[DEVCONF_MTU6] = cnf->mtu6;
3599 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3600 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3601 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3602 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3603 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3604 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3605 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3606 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3607 #ifdef CONFIG_IPV6_PRIVACY
3608 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3609 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3610 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3611 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3612 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3613 #endif
3614 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3615 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3616 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3617 #ifdef CONFIG_IPV6_ROUTER_PREF
3618 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3619 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3620 #ifdef CONFIG_IPV6_ROUTE_INFO
3621 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3622 #endif
3623 #endif
3624 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3625 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3626 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3627 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3628 #endif
3629 #ifdef CONFIG_IPV6_MROUTE
3630 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3631 #endif
3634 static inline size_t inet6_if_nlmsg_size(void)
3636 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3637 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3638 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3639 + nla_total_size(4) /* IFLA_MTU */
3640 + nla_total_size(4) /* IFLA_LINK */
3641 + nla_total_size( /* IFLA_PROTINFO */
3642 nla_total_size(4) /* IFLA_INET6_FLAGS */
3643 + nla_total_size(sizeof(struct ifla_cacheinfo))
3644 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3645 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3646 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3650 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3651 int bytes)
3653 int i;
3654 int pad = bytes - sizeof(u64) * items;
3655 BUG_ON(pad < 0);
3657 /* Use put_unaligned() because stats may not be aligned for u64. */
3658 put_unaligned(items, &stats[0]);
3659 for (i = 1; i < items; i++)
3660 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3662 memset(&stats[items], 0, pad);
3665 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3666 int bytes)
3668 switch(attrtype) {
3669 case IFLA_INET6_STATS:
3670 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3671 break;
3672 case IFLA_INET6_ICMP6STATS:
3673 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3674 break;
3678 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3679 u32 pid, u32 seq, int event, unsigned int flags)
3681 struct net_device *dev = idev->dev;
3682 struct nlattr *nla;
3683 struct ifinfomsg *hdr;
3684 struct nlmsghdr *nlh;
3685 void *protoinfo;
3686 struct ifla_cacheinfo ci;
3688 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3689 if (nlh == NULL)
3690 return -EMSGSIZE;
3692 hdr = nlmsg_data(nlh);
3693 hdr->ifi_family = AF_INET6;
3694 hdr->__ifi_pad = 0;
3695 hdr->ifi_type = dev->type;
3696 hdr->ifi_index = dev->ifindex;
3697 hdr->ifi_flags = dev_get_flags(dev);
3698 hdr->ifi_change = 0;
3700 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3702 if (dev->addr_len)
3703 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3705 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3706 if (dev->ifindex != dev->iflink)
3707 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3709 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3710 if (protoinfo == NULL)
3711 goto nla_put_failure;
3713 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3715 ci.max_reasm_len = IPV6_MAXPLEN;
3716 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3717 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3718 ci.reachable_time = idev->nd_parms->reachable_time;
3719 ci.retrans_time = idev->nd_parms->retrans_time;
3720 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3722 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3723 if (nla == NULL)
3724 goto nla_put_failure;
3725 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3727 /* XXX - MC not implemented */
3729 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3730 if (nla == NULL)
3731 goto nla_put_failure;
3732 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3734 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3735 if (nla == NULL)
3736 goto nla_put_failure;
3737 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3739 nla_nest_end(skb, protoinfo);
3740 return nlmsg_end(skb, nlh);
3742 nla_put_failure:
3743 nlmsg_cancel(skb, nlh);
3744 return -EMSGSIZE;
3747 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3749 struct net *net = sock_net(skb->sk);
3750 int idx, err;
3751 int s_idx = cb->args[0];
3752 struct net_device *dev;
3753 struct inet6_dev *idev;
3755 read_lock(&dev_base_lock);
3756 idx = 0;
3757 for_each_netdev(net, dev) {
3758 if (idx < s_idx)
3759 goto cont;
3760 if ((idev = in6_dev_get(dev)) == NULL)
3761 goto cont;
3762 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3763 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3764 in6_dev_put(idev);
3765 if (err <= 0)
3766 break;
3767 cont:
3768 idx++;
3770 read_unlock(&dev_base_lock);
3771 cb->args[0] = idx;
3773 return skb->len;
3776 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3778 struct sk_buff *skb;
3779 struct net *net = dev_net(idev->dev);
3780 int err = -ENOBUFS;
3782 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3783 if (skb == NULL)
3784 goto errout;
3786 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3787 if (err < 0) {
3788 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3789 WARN_ON(err == -EMSGSIZE);
3790 kfree_skb(skb);
3791 goto errout;
3793 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3794 errout:
3795 if (err < 0)
3796 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3799 static inline size_t inet6_prefix_nlmsg_size(void)
3801 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3802 + nla_total_size(sizeof(struct in6_addr))
3803 + nla_total_size(sizeof(struct prefix_cacheinfo));
3806 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3807 struct prefix_info *pinfo, u32 pid, u32 seq,
3808 int event, unsigned int flags)
3810 struct prefixmsg *pmsg;
3811 struct nlmsghdr *nlh;
3812 struct prefix_cacheinfo ci;
3814 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3815 if (nlh == NULL)
3816 return -EMSGSIZE;
3818 pmsg = nlmsg_data(nlh);
3819 pmsg->prefix_family = AF_INET6;
3820 pmsg->prefix_pad1 = 0;
3821 pmsg->prefix_pad2 = 0;
3822 pmsg->prefix_ifindex = idev->dev->ifindex;
3823 pmsg->prefix_len = pinfo->prefix_len;
3824 pmsg->prefix_type = pinfo->type;
3825 pmsg->prefix_pad3 = 0;
3826 pmsg->prefix_flags = 0;
3827 if (pinfo->onlink)
3828 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3829 if (pinfo->autoconf)
3830 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3832 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3834 ci.preferred_time = ntohl(pinfo->prefered);
3835 ci.valid_time = ntohl(pinfo->valid);
3836 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3838 return nlmsg_end(skb, nlh);
3840 nla_put_failure:
3841 nlmsg_cancel(skb, nlh);
3842 return -EMSGSIZE;
3845 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3846 struct prefix_info *pinfo)
3848 struct sk_buff *skb;
3849 struct net *net = dev_net(idev->dev);
3850 int err = -ENOBUFS;
3852 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3853 if (skb == NULL)
3854 goto errout;
3856 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3857 if (err < 0) {
3858 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3859 WARN_ON(err == -EMSGSIZE);
3860 kfree_skb(skb);
3861 goto errout;
3863 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3864 errout:
3865 if (err < 0)
3866 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3869 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3871 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3873 switch (event) {
3874 case RTM_NEWADDR:
3876 * If the address was optimistic
3877 * we inserted the route at the start of
3878 * our DAD process, so we don't need
3879 * to do it again
3881 if (!(ifp->rt->rt6i_node))
3882 ip6_ins_rt(ifp->rt);
3883 if (ifp->idev->cnf.forwarding)
3884 addrconf_join_anycast(ifp);
3885 break;
3886 case RTM_DELADDR:
3887 if (ifp->idev->cnf.forwarding)
3888 addrconf_leave_anycast(ifp);
3889 addrconf_leave_solict(ifp->idev, &ifp->addr);
3890 dst_hold(&ifp->rt->u.dst);
3891 if (ip6_del_rt(ifp->rt))
3892 dst_free(&ifp->rt->u.dst);
3893 break;
3897 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3899 rcu_read_lock_bh();
3900 if (likely(ifp->idev->dead == 0))
3901 __ipv6_ifa_notify(event, ifp);
3902 rcu_read_unlock_bh();
3905 #ifdef CONFIG_SYSCTL
3907 static
3908 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3909 void __user *buffer, size_t *lenp, loff_t *ppos)
3911 int *valp = ctl->data;
3912 int val = *valp;
3913 int ret;
3915 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3917 if (write)
3918 addrconf_fixup_forwarding(ctl, valp, val);
3919 return ret;
3922 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3923 int __user *name, int nlen,
3924 void __user *oldval,
3925 size_t __user *oldlenp,
3926 void __user *newval, size_t newlen)
3928 int *valp = table->data;
3929 int val = *valp;
3930 int new;
3932 if (!newval || !newlen)
3933 return 0;
3934 if (newlen != sizeof(int))
3935 return -EINVAL;
3936 if (get_user(new, (int __user *)newval))
3937 return -EFAULT;
3938 if (new == *valp)
3939 return 0;
3940 if (oldval && oldlenp) {
3941 size_t len;
3942 if (get_user(len, oldlenp))
3943 return -EFAULT;
3944 if (len) {
3945 if (len > table->maxlen)
3946 len = table->maxlen;
3947 if (copy_to_user(oldval, valp, len))
3948 return -EFAULT;
3949 if (put_user(len, oldlenp))
3950 return -EFAULT;
3954 *valp = new;
3955 addrconf_fixup_forwarding(table, valp, val);
3956 return 1;
3959 static struct addrconf_sysctl_table
3961 struct ctl_table_header *sysctl_header;
3962 ctl_table addrconf_vars[DEVCONF_MAX+1];
3963 char *dev_name;
3964 } addrconf_sysctl __read_mostly = {
3965 .sysctl_header = NULL,
3966 .addrconf_vars = {
3968 .ctl_name = NET_IPV6_FORWARDING,
3969 .procname = "forwarding",
3970 .data = &ipv6_devconf.forwarding,
3971 .maxlen = sizeof(int),
3972 .mode = 0644,
3973 .proc_handler = &addrconf_sysctl_forward,
3974 .strategy = &addrconf_sysctl_forward_strategy,
3977 .ctl_name = NET_IPV6_HOP_LIMIT,
3978 .procname = "hop_limit",
3979 .data = &ipv6_devconf.hop_limit,
3980 .maxlen = sizeof(int),
3981 .mode = 0644,
3982 .proc_handler = proc_dointvec,
3985 .ctl_name = NET_IPV6_MTU,
3986 .procname = "mtu",
3987 .data = &ipv6_devconf.mtu6,
3988 .maxlen = sizeof(int),
3989 .mode = 0644,
3990 .proc_handler = &proc_dointvec,
3993 .ctl_name = NET_IPV6_ACCEPT_RA,
3994 .procname = "accept_ra",
3995 .data = &ipv6_devconf.accept_ra,
3996 .maxlen = sizeof(int),
3997 .mode = 0644,
3998 .proc_handler = &proc_dointvec,
4001 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4002 .procname = "accept_redirects",
4003 .data = &ipv6_devconf.accept_redirects,
4004 .maxlen = sizeof(int),
4005 .mode = 0644,
4006 .proc_handler = &proc_dointvec,
4009 .ctl_name = NET_IPV6_AUTOCONF,
4010 .procname = "autoconf",
4011 .data = &ipv6_devconf.autoconf,
4012 .maxlen = sizeof(int),
4013 .mode = 0644,
4014 .proc_handler = &proc_dointvec,
4017 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4018 .procname = "dad_transmits",
4019 .data = &ipv6_devconf.dad_transmits,
4020 .maxlen = sizeof(int),
4021 .mode = 0644,
4022 .proc_handler = &proc_dointvec,
4025 .ctl_name = NET_IPV6_RTR_SOLICITS,
4026 .procname = "router_solicitations",
4027 .data = &ipv6_devconf.rtr_solicits,
4028 .maxlen = sizeof(int),
4029 .mode = 0644,
4030 .proc_handler = &proc_dointvec,
4033 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4034 .procname = "router_solicitation_interval",
4035 .data = &ipv6_devconf.rtr_solicit_interval,
4036 .maxlen = sizeof(int),
4037 .mode = 0644,
4038 .proc_handler = &proc_dointvec_jiffies,
4039 .strategy = &sysctl_jiffies,
4042 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4043 .procname = "router_solicitation_delay",
4044 .data = &ipv6_devconf.rtr_solicit_delay,
4045 .maxlen = sizeof(int),
4046 .mode = 0644,
4047 .proc_handler = &proc_dointvec_jiffies,
4048 .strategy = &sysctl_jiffies,
4051 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4052 .procname = "force_mld_version",
4053 .data = &ipv6_devconf.force_mld_version,
4054 .maxlen = sizeof(int),
4055 .mode = 0644,
4056 .proc_handler = &proc_dointvec,
4058 #ifdef CONFIG_IPV6_PRIVACY
4060 .ctl_name = NET_IPV6_USE_TEMPADDR,
4061 .procname = "use_tempaddr",
4062 .data = &ipv6_devconf.use_tempaddr,
4063 .maxlen = sizeof(int),
4064 .mode = 0644,
4065 .proc_handler = &proc_dointvec,
4068 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4069 .procname = "temp_valid_lft",
4070 .data = &ipv6_devconf.temp_valid_lft,
4071 .maxlen = sizeof(int),
4072 .mode = 0644,
4073 .proc_handler = &proc_dointvec,
4076 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4077 .procname = "temp_prefered_lft",
4078 .data = &ipv6_devconf.temp_prefered_lft,
4079 .maxlen = sizeof(int),
4080 .mode = 0644,
4081 .proc_handler = &proc_dointvec,
4084 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4085 .procname = "regen_max_retry",
4086 .data = &ipv6_devconf.regen_max_retry,
4087 .maxlen = sizeof(int),
4088 .mode = 0644,
4089 .proc_handler = &proc_dointvec,
4092 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4093 .procname = "max_desync_factor",
4094 .data = &ipv6_devconf.max_desync_factor,
4095 .maxlen = sizeof(int),
4096 .mode = 0644,
4097 .proc_handler = &proc_dointvec,
4099 #endif
4101 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4102 .procname = "max_addresses",
4103 .data = &ipv6_devconf.max_addresses,
4104 .maxlen = sizeof(int),
4105 .mode = 0644,
4106 .proc_handler = &proc_dointvec,
4109 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4110 .procname = "accept_ra_defrtr",
4111 .data = &ipv6_devconf.accept_ra_defrtr,
4112 .maxlen = sizeof(int),
4113 .mode = 0644,
4114 .proc_handler = &proc_dointvec,
4117 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4118 .procname = "accept_ra_pinfo",
4119 .data = &ipv6_devconf.accept_ra_pinfo,
4120 .maxlen = sizeof(int),
4121 .mode = 0644,
4122 .proc_handler = &proc_dointvec,
4124 #ifdef CONFIG_IPV6_ROUTER_PREF
4126 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4127 .procname = "accept_ra_rtr_pref",
4128 .data = &ipv6_devconf.accept_ra_rtr_pref,
4129 .maxlen = sizeof(int),
4130 .mode = 0644,
4131 .proc_handler = &proc_dointvec,
4134 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4135 .procname = "router_probe_interval",
4136 .data = &ipv6_devconf.rtr_probe_interval,
4137 .maxlen = sizeof(int),
4138 .mode = 0644,
4139 .proc_handler = &proc_dointvec_jiffies,
4140 .strategy = &sysctl_jiffies,
4142 #ifdef CONFIG_IPV6_ROUTE_INFO
4144 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4145 .procname = "accept_ra_rt_info_max_plen",
4146 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4147 .maxlen = sizeof(int),
4148 .mode = 0644,
4149 .proc_handler = &proc_dointvec,
4151 #endif
4152 #endif
4154 .ctl_name = NET_IPV6_PROXY_NDP,
4155 .procname = "proxy_ndp",
4156 .data = &ipv6_devconf.proxy_ndp,
4157 .maxlen = sizeof(int),
4158 .mode = 0644,
4159 .proc_handler = &proc_dointvec,
4162 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4163 .procname = "accept_source_route",
4164 .data = &ipv6_devconf.accept_source_route,
4165 .maxlen = sizeof(int),
4166 .mode = 0644,
4167 .proc_handler = &proc_dointvec,
4169 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4171 .ctl_name = CTL_UNNUMBERED,
4172 .procname = "optimistic_dad",
4173 .data = &ipv6_devconf.optimistic_dad,
4174 .maxlen = sizeof(int),
4175 .mode = 0644,
4176 .proc_handler = &proc_dointvec,
4179 #endif
4180 #ifdef CONFIG_IPV6_MROUTE
4182 .ctl_name = CTL_UNNUMBERED,
4183 .procname = "mc_forwarding",
4184 .data = &ipv6_devconf.mc_forwarding,
4185 .maxlen = sizeof(int),
4186 .mode = 0644,
4187 .proc_handler = &proc_dointvec,
4189 #endif
4191 .ctl_name = 0, /* sentinel */
4196 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4197 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4199 int i;
4200 struct addrconf_sysctl_table *t;
4202 #define ADDRCONF_CTL_PATH_DEV 3
4204 struct ctl_path addrconf_ctl_path[] = {
4205 { .procname = "net", .ctl_name = CTL_NET, },
4206 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4207 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4208 { /* to be set */ },
4209 { },
4213 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4214 if (t == NULL)
4215 goto out;
4217 for (i=0; t->addrconf_vars[i].data; i++) {
4218 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4219 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4220 t->addrconf_vars[i].extra2 = net;
4224 * Make a copy of dev_name, because '.procname' is regarded as const
4225 * by sysctl and we wouldn't want anyone to change it under our feet
4226 * (see SIOCSIFNAME).
4228 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4229 if (!t->dev_name)
4230 goto free;
4232 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4233 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4235 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4236 t->addrconf_vars);
4237 if (t->sysctl_header == NULL)
4238 goto free_procname;
4240 p->sysctl = t;
4241 return 0;
4243 free_procname:
4244 kfree(t->dev_name);
4245 free:
4246 kfree(t);
4247 out:
4248 return -ENOBUFS;
4251 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4253 struct addrconf_sysctl_table *t;
4255 if (p->sysctl == NULL)
4256 return;
4258 t = p->sysctl;
4259 p->sysctl = NULL;
4260 unregister_sysctl_table(t->sysctl_header);
4261 kfree(t->dev_name);
4262 kfree(t);
4265 static void addrconf_sysctl_register(struct inet6_dev *idev)
4267 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4268 NET_IPV6_NEIGH, "ipv6",
4269 &ndisc_ifinfo_sysctl_change,
4270 ndisc_ifinfo_sysctl_strategy);
4271 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4272 idev->dev->ifindex, idev, &idev->cnf);
4275 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4277 __addrconf_sysctl_unregister(&idev->cnf);
4278 neigh_sysctl_unregister(idev->nd_parms);
4282 #endif
4284 static int addrconf_init_net(struct net *net)
4286 int err;
4287 struct ipv6_devconf *all, *dflt;
4289 err = -ENOMEM;
4290 all = &ipv6_devconf;
4291 dflt = &ipv6_devconf_dflt;
4293 if (net != &init_net) {
4294 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4295 if (all == NULL)
4296 goto err_alloc_all;
4298 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4299 if (dflt == NULL)
4300 goto err_alloc_dflt;
4303 net->ipv6.devconf_all = all;
4304 net->ipv6.devconf_dflt = dflt;
4306 #ifdef CONFIG_SYSCTL
4307 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4308 NULL, all);
4309 if (err < 0)
4310 goto err_reg_all;
4312 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4313 NULL, dflt);
4314 if (err < 0)
4315 goto err_reg_dflt;
4316 #endif
4317 return 0;
4319 #ifdef CONFIG_SYSCTL
4320 err_reg_dflt:
4321 __addrconf_sysctl_unregister(all);
4322 err_reg_all:
4323 kfree(dflt);
4324 #endif
4325 err_alloc_dflt:
4326 kfree(all);
4327 err_alloc_all:
4328 return err;
4331 static void addrconf_exit_net(struct net *net)
4333 #ifdef CONFIG_SYSCTL
4334 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4335 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4336 #endif
4337 if (net != &init_net) {
4338 kfree(net->ipv6.devconf_dflt);
4339 kfree(net->ipv6.devconf_all);
4343 static struct pernet_operations addrconf_ops = {
4344 .init = addrconf_init_net,
4345 .exit = addrconf_exit_net,
4349 * Device notifier
4352 int register_inet6addr_notifier(struct notifier_block *nb)
4354 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4357 EXPORT_SYMBOL(register_inet6addr_notifier);
4359 int unregister_inet6addr_notifier(struct notifier_block *nb)
4361 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4364 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4366 static void addrconf_net_exit(struct net *net)
4368 struct net_device *dev;
4370 rtnl_lock();
4371 /* clean dev list */
4372 for_each_netdev(net, dev) {
4373 if (__in6_dev_get(dev) == NULL)
4374 continue;
4375 addrconf_ifdown(dev, 1);
4377 addrconf_ifdown(net->loopback_dev, 2);
4378 rtnl_unlock();
4381 static struct pernet_operations addrconf_net_ops = {
4382 .exit = addrconf_net_exit,
4386 * Init / cleanup code
4389 int __init addrconf_init(void)
4391 int err;
4393 if ((err = ipv6_addr_label_init()) < 0) {
4394 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4395 err);
4396 return err;
4399 register_pernet_subsys(&addrconf_ops);
4401 /* The addrconf netdev notifier requires that loopback_dev
4402 * has it's ipv6 private information allocated and setup
4403 * before it can bring up and give link-local addresses
4404 * to other devices which are up.
4406 * Unfortunately, loopback_dev is not necessarily the first
4407 * entry in the global dev_base list of net devices. In fact,
4408 * it is likely to be the very last entry on that list.
4409 * So this causes the notifier registry below to try and
4410 * give link-local addresses to all devices besides loopback_dev
4411 * first, then loopback_dev, which cases all the non-loopback_dev
4412 * devices to fail to get a link-local address.
4414 * So, as a temporary fix, allocate the ipv6 structure for
4415 * loopback_dev first by hand.
4416 * Longer term, all of the dependencies ipv6 has upon the loopback
4417 * device and it being up should be removed.
4419 rtnl_lock();
4420 if (!ipv6_add_dev(init_net.loopback_dev))
4421 err = -ENOMEM;
4422 rtnl_unlock();
4423 if (err)
4424 goto errlo;
4426 err = register_pernet_device(&addrconf_net_ops);
4427 if (err)
4428 return err;
4430 register_netdevice_notifier(&ipv6_dev_notf);
4432 addrconf_verify(0);
4434 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4435 if (err < 0)
4436 goto errout;
4438 /* Only the first call to __rtnl_register can fail */
4439 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4440 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4441 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4442 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4443 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4445 ipv6_addr_label_rtnl_register();
4447 return 0;
4448 errout:
4449 unregister_netdevice_notifier(&ipv6_dev_notf);
4450 errlo:
4451 unregister_pernet_subsys(&addrconf_ops);
4453 return err;
4456 void addrconf_cleanup(void)
4458 struct inet6_ifaddr *ifa;
4459 int i;
4461 unregister_netdevice_notifier(&ipv6_dev_notf);
4462 unregister_pernet_device(&addrconf_net_ops);
4464 unregister_pernet_subsys(&addrconf_ops);
4466 rtnl_lock();
4469 * Check hash table.
4471 write_lock_bh(&addrconf_hash_lock);
4472 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4473 for (ifa=inet6_addr_lst[i]; ifa; ) {
4474 struct inet6_ifaddr *bifa;
4476 bifa = ifa;
4477 ifa = ifa->lst_next;
4478 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4479 /* Do not free it; something is wrong.
4480 Now we can investigate it with debugger.
4484 write_unlock_bh(&addrconf_hash_lock);
4486 del_timer(&addr_chk_timer);
4487 rtnl_unlock();
4489 unregister_pernet_subsys(&addrconf_net_ops);