[IPV6]: Fix source address selection for ORCHID addresses
[linux-2.6] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
10  *
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.
15  */
16
17 /*
18  *      Changes:
19  *
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.
41  */
42
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>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
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>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99
100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102
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)
108 {
109 }
110
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 {
113 }
114 #endif
115
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);
120
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
123
124 static int ipv6_count_addresses(struct inet6_dev *idev);
125
126 /*
127  *      Configured unicast address hash table
128  */
129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
131
132 static void addrconf_verify(unsigned long);
133
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
136
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139
140 static int addrconf_ifdown(struct net_device *dev, int how);
141
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);
149
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);
154
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156
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. */
188 };
189
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. */
220 };
221
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
226 /* Check if a valid qdisc is available */
227 static inline int addrconf_qdisc_ok(struct net_device *dev)
228 {
229         return (dev->qdisc != &noop_qdisc);
230 }
231
232 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
233 {
234         if (del_timer(&ifp->timer))
235                 __in6_ifa_put(ifp);
236 }
237
238 enum addrconf_timer_t
239 {
240         AC_NONE,
241         AC_DAD,
242         AC_RS,
243 };
244
245 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
246                                enum addrconf_timer_t what,
247                                unsigned long when)
248 {
249         if (!del_timer(&ifp->timer))
250                 in6_ifa_hold(ifp);
251
252         switch (what) {
253         case AC_DAD:
254                 ifp->timer.function = addrconf_dad_timer;
255                 break;
256         case AC_RS:
257                 ifp->timer.function = addrconf_rs_timer;
258                 break;
259         default:;
260         }
261         ifp->timer.expires = jiffies + when;
262         add_timer(&ifp->timer);
263 }
264
265 static int snmp6_alloc_dev(struct inet6_dev *idev)
266 {
267         if (snmp_mib_init((void **)idev->stats.ipv6,
268                           sizeof(struct ipstats_mib)) < 0)
269                 goto err_ip;
270         if (snmp_mib_init((void **)idev->stats.icmpv6,
271                           sizeof(struct icmpv6_mib)) < 0)
272                 goto err_icmp;
273         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
274                           sizeof(struct icmpv6msg_mib)) < 0)
275                 goto err_icmpmsg;
276
277         return 0;
278
279 err_icmpmsg:
280         snmp_mib_free((void **)idev->stats.icmpv6);
281 err_icmp:
282         snmp_mib_free((void **)idev->stats.ipv6);
283 err_ip:
284         return -ENOMEM;
285 }
286
287 static void snmp6_free_dev(struct inet6_dev *idev)
288 {
289         snmp_mib_free((void **)idev->stats.icmpv6msg);
290         snmp_mib_free((void **)idev->stats.icmpv6);
291         snmp_mib_free((void **)idev->stats.ipv6);
292 }
293
294 /* Nobody refers to this device, we may destroy it. */
295
296 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
297 {
298         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
299         kfree(idev);
300 }
301
302 void in6_dev_finish_destroy(struct inet6_dev *idev)
303 {
304         struct net_device *dev = idev->dev;
305         BUG_TRAP(idev->addr_list==NULL);
306         BUG_TRAP(idev->mc_list==NULL);
307 #ifdef NET_REFCNT_DEBUG
308         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
309 #endif
310         dev_put(dev);
311         if (!idev->dead) {
312                 printk("Freeing alive inet6 device %p\n", idev);
313                 return;
314         }
315         snmp6_free_dev(idev);
316         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
317 }
318
319 EXPORT_SYMBOL(in6_dev_finish_destroy);
320
321 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
322 {
323         struct inet6_dev *ndev;
324         struct in6_addr maddr;
325
326         ASSERT_RTNL();
327
328         if (dev->mtu < IPV6_MIN_MTU)
329                 return NULL;
330
331         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
332
333         if (ndev == NULL)
334                 return NULL;
335
336         rwlock_init(&ndev->lock);
337         ndev->dev = dev;
338         memcpy(&ndev->cnf, dev->nd_net->ipv6.devconf_dflt, sizeof(ndev->cnf));
339         ndev->cnf.mtu6 = dev->mtu;
340         ndev->cnf.sysctl = NULL;
341         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
342         if (ndev->nd_parms == NULL) {
343                 kfree(ndev);
344                 return NULL;
345         }
346         /* We refer to the device */
347         dev_hold(dev);
348
349         if (snmp6_alloc_dev(ndev) < 0) {
350                 ADBG((KERN_WARNING
351                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
352                         __FUNCTION__, dev->name));
353                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
354                 ndev->dead = 1;
355                 in6_dev_finish_destroy(ndev);
356                 return NULL;
357         }
358
359         if (snmp6_register_dev(ndev) < 0) {
360                 ADBG((KERN_WARNING
361                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
362                         __FUNCTION__, dev->name));
363                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
364                 ndev->dead = 1;
365                 in6_dev_finish_destroy(ndev);
366                 return NULL;
367         }
368
369         /* One reference from device.  We must do this before
370          * we invoke __ipv6_regen_rndid().
371          */
372         in6_dev_hold(ndev);
373
374 #ifdef CONFIG_IPV6_PRIVACY
375         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
376         if ((dev->flags&IFF_LOOPBACK) ||
377             dev->type == ARPHRD_TUNNEL ||
378 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
379             dev->type == ARPHRD_SIT ||
380 #endif
381             dev->type == ARPHRD_NONE) {
382                 printk(KERN_INFO
383                        "%s: Disabled Privacy Extensions\n",
384                        dev->name);
385                 ndev->cnf.use_tempaddr = -1;
386
387                 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
388                         printk(KERN_INFO
389                                "%s: Disabled Multicast RS\n",
390                                dev->name);
391                         ndev->cnf.rtr_solicits = 0;
392                 }
393         } else {
394                 in6_dev_hold(ndev);
395                 ipv6_regen_rndid((unsigned long) ndev);
396         }
397 #endif
398
399         if (netif_running(dev) && addrconf_qdisc_ok(dev))
400                 ndev->if_flags |= IF_READY;
401
402         ipv6_mc_init_dev(ndev);
403         ndev->tstamp = jiffies;
404         addrconf_sysctl_register(ndev);
405         /* protected by rtnl_lock */
406         rcu_assign_pointer(dev->ip6_ptr, ndev);
407
408         /* Join all-node multicast group */
409         ipv6_addr_all_nodes(&maddr);
410         ipv6_dev_mc_inc(dev, &maddr);
411
412         return ndev;
413 }
414
415 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
416 {
417         struct inet6_dev *idev;
418
419         ASSERT_RTNL();
420
421         if ((idev = __in6_dev_get(dev)) == NULL) {
422                 if ((idev = ipv6_add_dev(dev)) == NULL)
423                         return NULL;
424         }
425
426         if (dev->flags&IFF_UP)
427                 ipv6_mc_up(idev);
428         return idev;
429 }
430
431 #ifdef CONFIG_SYSCTL
432 static void dev_forward_change(struct inet6_dev *idev)
433 {
434         struct net_device *dev;
435         struct inet6_ifaddr *ifa;
436         struct in6_addr addr;
437
438         if (!idev)
439                 return;
440         dev = idev->dev;
441         if (dev && (dev->flags & IFF_MULTICAST)) {
442                 ipv6_addr_all_routers(&addr);
443
444                 if (idev->cnf.forwarding)
445                         ipv6_dev_mc_inc(dev, &addr);
446                 else
447                         ipv6_dev_mc_dec(dev, &addr);
448         }
449         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
450                 if (ifa->flags&IFA_F_TENTATIVE)
451                         continue;
452                 if (idev->cnf.forwarding)
453                         addrconf_join_anycast(ifa);
454                 else
455                         addrconf_leave_anycast(ifa);
456         }
457 }
458
459
460 static void addrconf_forward_change(struct net *net, __s32 newf)
461 {
462         struct net_device *dev;
463         struct inet6_dev *idev;
464
465         read_lock(&dev_base_lock);
466         for_each_netdev(net, dev) {
467                 rcu_read_lock();
468                 idev = __in6_dev_get(dev);
469                 if (idev) {
470                         int changed = (!idev->cnf.forwarding) ^ (!newf);
471                         idev->cnf.forwarding = newf;
472                         if (changed)
473                                 dev_forward_change(idev);
474                 }
475                 rcu_read_unlock();
476         }
477         read_unlock(&dev_base_lock);
478 }
479
480 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
481 {
482         struct net *net;
483
484         net = (struct net *)table->extra2;
485         if (p == &net->ipv6.devconf_dflt->forwarding)
486                 return;
487
488         if (p == &net->ipv6.devconf_all->forwarding) {
489                 __s32 newf = net->ipv6.devconf_all->forwarding;
490                 net->ipv6.devconf_dflt->forwarding = newf;
491                 addrconf_forward_change(net, newf);
492         } else if ((!*p) ^ (!old))
493                 dev_forward_change((struct inet6_dev *)table->extra1);
494
495         if (*p)
496                 rt6_purge_dflt_routers();
497 }
498 #endif
499
500 /* Nobody refers to this ifaddr, destroy it */
501
502 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
503 {
504         BUG_TRAP(ifp->if_next==NULL);
505         BUG_TRAP(ifp->lst_next==NULL);
506 #ifdef NET_REFCNT_DEBUG
507         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
508 #endif
509
510         in6_dev_put(ifp->idev);
511
512         if (del_timer(&ifp->timer))
513                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
514
515         if (!ifp->dead) {
516                 printk("Freeing alive inet6 address %p\n", ifp);
517                 return;
518         }
519         dst_release(&ifp->rt->u.dst);
520
521         kfree(ifp);
522 }
523
524 static void
525 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
526 {
527         struct inet6_ifaddr *ifa, **ifap;
528         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
529
530         /*
531          * Each device address list is sorted in order of scope -
532          * global before linklocal.
533          */
534         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
535              ifap = &ifa->if_next) {
536                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
537                         break;
538         }
539
540         ifp->if_next = *ifap;
541         *ifap = ifp;
542 }
543
544 /* On success it returns ifp with increased reference count */
545
546 static struct inet6_ifaddr *
547 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
548               int scope, u32 flags)
549 {
550         struct inet6_ifaddr *ifa = NULL;
551         struct rt6_info *rt;
552         int hash;
553         int err = 0;
554
555         rcu_read_lock_bh();
556         if (idev->dead) {
557                 err = -ENODEV;                  /*XXX*/
558                 goto out2;
559         }
560
561         write_lock(&addrconf_hash_lock);
562
563         /* Ignore adding duplicate addresses on an interface */
564         if (ipv6_chk_same_addr(&init_net, addr, idev->dev)) {
565                 ADBG(("ipv6_add_addr: already assigned\n"));
566                 err = -EEXIST;
567                 goto out;
568         }
569
570         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
571
572         if (ifa == NULL) {
573                 ADBG(("ipv6_add_addr: malloc failed\n"));
574                 err = -ENOBUFS;
575                 goto out;
576         }
577
578         rt = addrconf_dst_alloc(idev, addr, 0);
579         if (IS_ERR(rt)) {
580                 err = PTR_ERR(rt);
581                 goto out;
582         }
583
584         ipv6_addr_copy(&ifa->addr, addr);
585
586         spin_lock_init(&ifa->lock);
587         init_timer(&ifa->timer);
588         ifa->timer.data = (unsigned long) ifa;
589         ifa->scope = scope;
590         ifa->prefix_len = pfxlen;
591         ifa->flags = flags | IFA_F_TENTATIVE;
592         ifa->cstamp = ifa->tstamp = jiffies;
593
594         ifa->rt = rt;
595
596         /*
597          * part one of RFC 4429, section 3.3
598          * We should not configure an address as
599          * optimistic if we do not yet know the link
600          * layer address of our nexhop router
601          */
602
603         if (rt->rt6i_nexthop == NULL)
604                 ifa->flags &= ~IFA_F_OPTIMISTIC;
605
606         ifa->idev = idev;
607         in6_dev_hold(idev);
608         /* For caller */
609         in6_ifa_hold(ifa);
610
611         /* Add to big hash table */
612         hash = ipv6_addr_hash(addr);
613
614         ifa->lst_next = inet6_addr_lst[hash];
615         inet6_addr_lst[hash] = ifa;
616         in6_ifa_hold(ifa);
617         write_unlock(&addrconf_hash_lock);
618
619         write_lock(&idev->lock);
620         /* Add to inet6_dev unicast addr list. */
621         ipv6_link_dev_addr(idev, ifa);
622
623 #ifdef CONFIG_IPV6_PRIVACY
624         if (ifa->flags&IFA_F_TEMPORARY) {
625                 ifa->tmp_next = idev->tempaddr_list;
626                 idev->tempaddr_list = ifa;
627                 in6_ifa_hold(ifa);
628         }
629 #endif
630
631         in6_ifa_hold(ifa);
632         write_unlock(&idev->lock);
633 out2:
634         rcu_read_unlock_bh();
635
636         if (likely(err == 0))
637                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
638         else {
639                 kfree(ifa);
640                 ifa = ERR_PTR(err);
641         }
642
643         return ifa;
644 out:
645         write_unlock(&addrconf_hash_lock);
646         goto out2;
647 }
648
649 /* This function wants to get referenced ifp and releases it before return */
650
651 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
652 {
653         struct inet6_ifaddr *ifa, **ifap;
654         struct inet6_dev *idev = ifp->idev;
655         int hash;
656         int deleted = 0, onlink = 0;
657         unsigned long expires = jiffies;
658
659         hash = ipv6_addr_hash(&ifp->addr);
660
661         ifp->dead = 1;
662
663         write_lock_bh(&addrconf_hash_lock);
664         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
665              ifap = &ifa->lst_next) {
666                 if (ifa == ifp) {
667                         *ifap = ifa->lst_next;
668                         __in6_ifa_put(ifp);
669                         ifa->lst_next = NULL;
670                         break;
671                 }
672         }
673         write_unlock_bh(&addrconf_hash_lock);
674
675         write_lock_bh(&idev->lock);
676 #ifdef CONFIG_IPV6_PRIVACY
677         if (ifp->flags&IFA_F_TEMPORARY) {
678                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
679                      ifap = &ifa->tmp_next) {
680                         if (ifa == ifp) {
681                                 *ifap = ifa->tmp_next;
682                                 if (ifp->ifpub) {
683                                         in6_ifa_put(ifp->ifpub);
684                                         ifp->ifpub = NULL;
685                                 }
686                                 __in6_ifa_put(ifp);
687                                 ifa->tmp_next = NULL;
688                                 break;
689                         }
690                 }
691         }
692 #endif
693
694         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
695                 if (ifa == ifp) {
696                         *ifap = ifa->if_next;
697                         __in6_ifa_put(ifp);
698                         ifa->if_next = NULL;
699                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
700                                 break;
701                         deleted = 1;
702                         continue;
703                 } else if (ifp->flags & IFA_F_PERMANENT) {
704                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
705                                               ifp->prefix_len)) {
706                                 if (ifa->flags & IFA_F_PERMANENT) {
707                                         onlink = 1;
708                                         if (deleted)
709                                                 break;
710                                 } else {
711                                         unsigned long lifetime;
712
713                                         if (!onlink)
714                                                 onlink = -1;
715
716                                         spin_lock(&ifa->lock);
717                                         lifetime = min_t(unsigned long,
718                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
719                                         if (time_before(expires,
720                                                         ifa->tstamp + lifetime * HZ))
721                                                 expires = ifa->tstamp + lifetime * HZ;
722                                         spin_unlock(&ifa->lock);
723                                 }
724                         }
725                 }
726                 ifap = &ifa->if_next;
727         }
728         write_unlock_bh(&idev->lock);
729
730         ipv6_ifa_notify(RTM_DELADDR, ifp);
731
732         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
733
734         addrconf_del_timer(ifp);
735
736         /*
737          * Purge or update corresponding prefix
738          *
739          * 1) we don't purge prefix here if address was not permanent.
740          *    prefix is managed by its own lifetime.
741          * 2) if there're no addresses, delete prefix.
742          * 3) if there're still other permanent address(es),
743          *    corresponding prefix is still permanent.
744          * 4) otherwise, update prefix lifetime to the
745          *    longest valid lifetime among the corresponding
746          *    addresses on the device.
747          *    Note: subsequent RA will update lifetime.
748          *
749          * --yoshfuji
750          */
751         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
752                 struct in6_addr prefix;
753                 struct rt6_info *rt;
754
755                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
756                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
757
758                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
759                         if (onlink == 0) {
760                                 ip6_del_rt(rt);
761                                 rt = NULL;
762                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
763                                 rt->rt6i_expires = expires;
764                                 rt->rt6i_flags |= RTF_EXPIRES;
765                         }
766                 }
767                 dst_release(&rt->u.dst);
768         }
769
770         in6_ifa_put(ifp);
771 }
772
773 #ifdef CONFIG_IPV6_PRIVACY
774 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
775 {
776         struct inet6_dev *idev = ifp->idev;
777         struct in6_addr addr, *tmpaddr;
778         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
779         int tmp_plen;
780         int ret = 0;
781         int max_addresses;
782         u32 addr_flags;
783
784         write_lock(&idev->lock);
785         if (ift) {
786                 spin_lock_bh(&ift->lock);
787                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
788                 spin_unlock_bh(&ift->lock);
789                 tmpaddr = &addr;
790         } else {
791                 tmpaddr = NULL;
792         }
793 retry:
794         in6_dev_hold(idev);
795         if (idev->cnf.use_tempaddr <= 0) {
796                 write_unlock(&idev->lock);
797                 printk(KERN_INFO
798                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
799                 in6_dev_put(idev);
800                 ret = -1;
801                 goto out;
802         }
803         spin_lock_bh(&ifp->lock);
804         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
805                 idev->cnf.use_tempaddr = -1;    /*XXX*/
806                 spin_unlock_bh(&ifp->lock);
807                 write_unlock(&idev->lock);
808                 printk(KERN_WARNING
809                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
810                 in6_dev_put(idev);
811                 ret = -1;
812                 goto out;
813         }
814         in6_ifa_hold(ifp);
815         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
816         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
817                 spin_unlock_bh(&ifp->lock);
818                 write_unlock(&idev->lock);
819                 printk(KERN_WARNING
820                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
821                 in6_ifa_put(ifp);
822                 in6_dev_put(idev);
823                 ret = -1;
824                 goto out;
825         }
826         memcpy(&addr.s6_addr[8], idev->rndid, 8);
827         tmp_valid_lft = min_t(__u32,
828                               ifp->valid_lft,
829                               idev->cnf.temp_valid_lft);
830         tmp_prefered_lft = min_t(__u32,
831                                  ifp->prefered_lft,
832                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
833         tmp_plen = ifp->prefix_len;
834         max_addresses = idev->cnf.max_addresses;
835         tmp_cstamp = ifp->cstamp;
836         tmp_tstamp = ifp->tstamp;
837         spin_unlock_bh(&ifp->lock);
838
839         write_unlock(&idev->lock);
840
841         addr_flags = IFA_F_TEMPORARY;
842         /* set in addrconf_prefix_rcv() */
843         if (ifp->flags & IFA_F_OPTIMISTIC)
844                 addr_flags |= IFA_F_OPTIMISTIC;
845
846         ift = !max_addresses ||
847               ipv6_count_addresses(idev) < max_addresses ?
848                 ipv6_add_addr(idev, &addr, tmp_plen,
849                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
850                               addr_flags) : NULL;
851         if (!ift || IS_ERR(ift)) {
852                 in6_ifa_put(ifp);
853                 in6_dev_put(idev);
854                 printk(KERN_INFO
855                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
856                 tmpaddr = &addr;
857                 write_lock(&idev->lock);
858                 goto retry;
859         }
860
861         spin_lock_bh(&ift->lock);
862         ift->ifpub = ifp;
863         ift->valid_lft = tmp_valid_lft;
864         ift->prefered_lft = tmp_prefered_lft;
865         ift->cstamp = tmp_cstamp;
866         ift->tstamp = tmp_tstamp;
867         spin_unlock_bh(&ift->lock);
868
869         addrconf_dad_start(ift, 0);
870         in6_ifa_put(ift);
871         in6_dev_put(idev);
872 out:
873         return ret;
874 }
875 #endif
876
877 /*
878  *      Choose an appropriate source address (RFC3484)
879  */
880 struct ipv6_saddr_score {
881         int             addr_type;
882         unsigned int    attrs;
883         int             matchlen;
884         int             scope;
885         unsigned int    rule;
886 };
887
888 #define IPV6_SADDR_SCORE_LOCAL          0x0001
889 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
890 #define IPV6_SADDR_SCORE_HOA            0x0008
891 #define IPV6_SADDR_SCORE_OIF            0x0010
892 #define IPV6_SADDR_SCORE_LABEL          0x0020
893 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
894
895 static inline int ipv6_saddr_preferred(int type)
896 {
897         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
898                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
899                 return 1;
900         return 0;
901 }
902
903 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
904                        struct in6_addr *daddr, struct in6_addr *saddr)
905 {
906         struct ipv6_saddr_score hiscore;
907         struct inet6_ifaddr *ifa_result = NULL;
908         int daddr_type = __ipv6_addr_type(daddr);
909         int daddr_scope = __ipv6_addr_src_scope(daddr_type);
910         int daddr_ifindex = daddr_dev ? daddr_dev->ifindex : 0;
911         u32 daddr_label = ipv6_addr_label(daddr, daddr_type, daddr_ifindex);
912         struct net_device *dev;
913
914         memset(&hiscore, 0, sizeof(hiscore));
915
916         read_lock(&dev_base_lock);
917         rcu_read_lock();
918
919         for_each_netdev(&init_net, dev) {
920                 struct inet6_dev *idev;
921                 struct inet6_ifaddr *ifa;
922
923                 /* Rule 0: Candidate Source Address (section 4)
924                  *  - multicast and link-local destination address,
925                  *    the set of candidate source address MUST only
926                  *    include addresses assigned to interfaces
927                  *    belonging to the same link as the outgoing
928                  *    interface.
929                  * (- For site-local destination addresses, the
930                  *    set of candidate source addresses MUST only
931                  *    include addresses assigned to interfaces
932                  *    belonging to the same site as the outgoing
933                  *    interface.)
934                  */
935                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
936                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
937                     daddr_dev && dev != daddr_dev)
938                         continue;
939
940                 idev = __in6_dev_get(dev);
941                 if (!idev)
942                         continue;
943
944                 read_lock_bh(&idev->lock);
945                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
946                         struct ipv6_saddr_score score;
947
948                         score.addr_type = __ipv6_addr_type(&ifa->addr);
949
950                         /* Rule 0:
951                          * - Tentative Address (RFC2462 section 5.4)
952                          *  - A tentative address is not considered
953                          *    "assigned to an interface" in the traditional
954                          *    sense, unless it is also flagged as optimistic.
955                          * - Candidate Source Address (section 4)
956                          *  - In any case, anycast addresses, multicast
957                          *    addresses, and the unspecified address MUST
958                          *    NOT be included in a candidate set.
959                          */
960                         if ((ifa->flags & IFA_F_TENTATIVE) &&
961                             (!(ifa->flags & IFA_F_OPTIMISTIC)))
962                                 continue;
963                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
964                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
965                                 LIMIT_NETDEBUG(KERN_DEBUG
966                                                "ADDRCONF: unspecified / multicast address "
967                                                "assigned as unicast address on %s",
968                                                dev->name);
969                                 continue;
970                         }
971
972                         score.attrs = 0;
973                         score.matchlen = 0;
974                         score.scope = 0;
975                         score.rule = 0;
976
977                         if (ifa_result == NULL) {
978                                 /* record it if the first available entry */
979                                 goto record_it;
980                         }
981
982                         /* Rule 1: Prefer same address */
983                         if (hiscore.rule < 1) {
984                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
985                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
986                                 hiscore.rule++;
987                         }
988                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
989                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
990                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
991                                         score.rule = 1;
992                                         goto record_it;
993                                 }
994                         } else {
995                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
996                                         continue;
997                         }
998
999                         /* Rule 2: Prefer appropriate scope */
1000                         if (hiscore.rule < 2) {
1001                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1002                                 hiscore.rule++;
1003                         }
1004                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1005                         if (hiscore.scope < score.scope) {
1006                                 if (hiscore.scope < daddr_scope) {
1007                                         score.rule = 2;
1008                                         goto record_it;
1009                                 } else
1010                                         continue;
1011                         } else if (score.scope < hiscore.scope) {
1012                                 if (score.scope < daddr_scope)
1013                                         break; /* addresses sorted by scope */
1014                                 else {
1015                                         score.rule = 2;
1016                                         goto record_it;
1017                                 }
1018                         }
1019
1020                         /* Rule 3: Avoid deprecated and optimistic addresses */
1021                         if (hiscore.rule < 3) {
1022                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1023                                    (((ifa_result->flags &
1024                                     (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0)))
1025                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1026                                 hiscore.rule++;
1027                         }
1028                         if (ipv6_saddr_preferred(score.addr_type) ||
1029                            (((ifa->flags &
1030                             (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0))) {
1031                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1032                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1033                                         score.rule = 3;
1034                                         goto record_it;
1035                                 }
1036                         } else {
1037                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1038                                         continue;
1039                         }
1040
1041                         /* Rule 4: Prefer home address */
1042 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
1043                         if (hiscore.rule < 4) {
1044                                 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1045                                         hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1046                                 hiscore.rule++;
1047                         }
1048                         if (ifa->flags & IFA_F_HOMEADDRESS) {
1049                                 score.attrs |= IPV6_SADDR_SCORE_HOA;
1050                                 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1051                                         score.rule = 4;
1052                                         goto record_it;
1053                                 }
1054                         } else {
1055                                 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1056                                         continue;
1057                         }
1058 #else
1059                         if (hiscore.rule < 4)
1060                                 hiscore.rule++;
1061 #endif
1062
1063                         /* Rule 5: Prefer outgoing interface */
1064                         if (hiscore.rule < 5) {
1065                                 if (daddr_dev == NULL ||
1066                                     daddr_dev == ifa_result->idev->dev)
1067                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1068                                 hiscore.rule++;
1069                         }
1070                         if (daddr_dev == NULL ||
1071                             daddr_dev == ifa->idev->dev) {
1072                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1073                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1074                                         score.rule = 5;
1075                                         goto record_it;
1076                                 }
1077                         } else {
1078                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1079                                         continue;
1080                         }
1081
1082                         /* Rule 6: Prefer matching label */
1083                         if (hiscore.rule < 6) {
1084                                 if (ipv6_addr_label(&ifa_result->addr,
1085                                                     hiscore.addr_type,
1086                                                     ifa_result->idev->dev->ifindex) == daddr_label)
1087                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1088                                 hiscore.rule++;
1089                         }
1090                         if (ipv6_addr_label(&ifa->addr,
1091                                             score.addr_type,
1092                                             ifa->idev->dev->ifindex) == daddr_label) {
1093                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1094                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1095                                         score.rule = 6;
1096                                         goto record_it;
1097                                 }
1098                         } else {
1099                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1100                                         continue;
1101                         }
1102
1103 #ifdef CONFIG_IPV6_PRIVACY
1104                         /* Rule 7: Prefer public address
1105                          * Note: prefer temprary address if use_tempaddr >= 2
1106                          */
1107                         if (hiscore.rule < 7) {
1108                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1109                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1110                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1111                                 hiscore.rule++;
1112                         }
1113                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1114                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1115                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1116                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1117                                         score.rule = 7;
1118                                         goto record_it;
1119                                 }
1120                         } else {
1121                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1122                                         continue;
1123                         }
1124 #else
1125                         if (hiscore.rule < 7)
1126                                 hiscore.rule++;
1127 #endif
1128
1129                         /* Skip rule 8 for orchid -> non-orchid address pairs. */
1130                         if (ipv6_addr_orchid(&ifa->addr) && !ipv6_addr_orchid(daddr))
1131                                 continue;
1132
1133                         /* Rule 8: Use longest matching prefix */
1134                         if (hiscore.rule < 8) {
1135                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1136                                 hiscore.rule++;
1137                         }
1138                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1139                         if (score.matchlen > hiscore.matchlen) {
1140                                 score.rule = 8;
1141                                 goto record_it;
1142                         }
1143 #if 0
1144                         else if (score.matchlen < hiscore.matchlen)
1145                                 continue;
1146 #endif
1147
1148                         /* Final Rule: choose first available one */
1149                         continue;
1150 record_it:
1151                         if (ifa_result)
1152                                 in6_ifa_put(ifa_result);
1153                         in6_ifa_hold(ifa);
1154                         ifa_result = ifa;
1155                         hiscore = score;
1156                 }
1157                 read_unlock_bh(&idev->lock);
1158         }
1159         rcu_read_unlock();
1160         read_unlock(&dev_base_lock);
1161
1162         if (!ifa_result)
1163                 return -EADDRNOTAVAIL;
1164
1165         ipv6_addr_copy(saddr, &ifa_result->addr);
1166         in6_ifa_put(ifa_result);
1167         return 0;
1168 }
1169
1170
1171 int ipv6_get_saddr(struct dst_entry *dst,
1172                    struct in6_addr *daddr, struct in6_addr *saddr)
1173 {
1174         return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1175 }
1176
1177 EXPORT_SYMBOL(ipv6_get_saddr);
1178
1179 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1180                     unsigned char banned_flags)
1181 {
1182         struct inet6_dev *idev;
1183         int err = -EADDRNOTAVAIL;
1184
1185         rcu_read_lock();
1186         if ((idev = __in6_dev_get(dev)) != NULL) {
1187                 struct inet6_ifaddr *ifp;
1188
1189                 read_lock_bh(&idev->lock);
1190                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1191                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1192                                 ipv6_addr_copy(addr, &ifp->addr);
1193                                 err = 0;
1194                                 break;
1195                         }
1196                 }
1197                 read_unlock_bh(&idev->lock);
1198         }
1199         rcu_read_unlock();
1200         return err;
1201 }
1202
1203 static int ipv6_count_addresses(struct inet6_dev *idev)
1204 {
1205         int cnt = 0;
1206         struct inet6_ifaddr *ifp;
1207
1208         read_lock_bh(&idev->lock);
1209         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1210                 cnt++;
1211         read_unlock_bh(&idev->lock);
1212         return cnt;
1213 }
1214
1215 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1216                   struct net_device *dev, int strict)
1217 {
1218         struct inet6_ifaddr * ifp;
1219         u8 hash = ipv6_addr_hash(addr);
1220
1221         read_lock_bh(&addrconf_hash_lock);
1222         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1223                 if (ifp->idev->dev->nd_net != net)
1224                         continue;
1225                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1226                     !(ifp->flags&IFA_F_TENTATIVE)) {
1227                         if (dev == NULL || ifp->idev->dev == dev ||
1228                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1229                                 break;
1230                 }
1231         }
1232         read_unlock_bh(&addrconf_hash_lock);
1233         return ifp != NULL;
1234 }
1235 EXPORT_SYMBOL(ipv6_chk_addr);
1236
1237 static
1238 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1239                        struct net_device *dev)
1240 {
1241         struct inet6_ifaddr * ifp;
1242         u8 hash = ipv6_addr_hash(addr);
1243
1244         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1245                 if (ifp->idev->dev->nd_net != net)
1246                         continue;
1247                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1248                         if (dev == NULL || ifp->idev->dev == dev)
1249                                 break;
1250                 }
1251         }
1252         return ifp != NULL;
1253 }
1254
1255 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1256                                      struct net_device *dev, int strict)
1257 {
1258         struct inet6_ifaddr * ifp;
1259         u8 hash = ipv6_addr_hash(addr);
1260
1261         read_lock_bh(&addrconf_hash_lock);
1262         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1263                 if (ifp->idev->dev->nd_net != net)
1264                         continue;
1265                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1266                         if (dev == NULL || ifp->idev->dev == dev ||
1267                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1268                                 in6_ifa_hold(ifp);
1269                                 break;
1270                         }
1271                 }
1272         }
1273         read_unlock_bh(&addrconf_hash_lock);
1274
1275         return ifp;
1276 }
1277
1278 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1279 {
1280         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1281         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1282         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1283         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1284         int sk_ipv6only = ipv6_only_sock(sk);
1285         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1286         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1287         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1288
1289         if (!sk2_rcv_saddr && !sk_ipv6only)
1290                 return 1;
1291
1292         if (addr_type2 == IPV6_ADDR_ANY &&
1293             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1294                 return 1;
1295
1296         if (addr_type == IPV6_ADDR_ANY &&
1297             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1298                 return 1;
1299
1300         if (sk2_rcv_saddr6 &&
1301             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1302                 return 1;
1303
1304         if (addr_type == IPV6_ADDR_MAPPED &&
1305             !sk2_ipv6only &&
1306             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1307                 return 1;
1308
1309         return 0;
1310 }
1311
1312 /* Gets referenced address, destroys ifaddr */
1313
1314 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1315 {
1316         if (ifp->flags&IFA_F_PERMANENT) {
1317                 spin_lock_bh(&ifp->lock);
1318                 addrconf_del_timer(ifp);
1319                 ifp->flags |= IFA_F_TENTATIVE;
1320                 spin_unlock_bh(&ifp->lock);
1321                 in6_ifa_put(ifp);
1322 #ifdef CONFIG_IPV6_PRIVACY
1323         } else if (ifp->flags&IFA_F_TEMPORARY) {
1324                 struct inet6_ifaddr *ifpub;
1325                 spin_lock_bh(&ifp->lock);
1326                 ifpub = ifp->ifpub;
1327                 if (ifpub) {
1328                         in6_ifa_hold(ifpub);
1329                         spin_unlock_bh(&ifp->lock);
1330                         ipv6_create_tempaddr(ifpub, ifp);
1331                         in6_ifa_put(ifpub);
1332                 } else {
1333                         spin_unlock_bh(&ifp->lock);
1334                 }
1335                 ipv6_del_addr(ifp);
1336 #endif
1337         } else
1338                 ipv6_del_addr(ifp);
1339 }
1340
1341 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1342 {
1343         if (net_ratelimit())
1344                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1345         addrconf_dad_stop(ifp);
1346 }
1347
1348 /* Join to solicited addr multicast group. */
1349
1350 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1351 {
1352         struct in6_addr maddr;
1353
1354         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1355                 return;
1356
1357         addrconf_addr_solict_mult(addr, &maddr);
1358         ipv6_dev_mc_inc(dev, &maddr);
1359 }
1360
1361 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1362 {
1363         struct in6_addr maddr;
1364
1365         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1366                 return;
1367
1368         addrconf_addr_solict_mult(addr, &maddr);
1369         __ipv6_dev_mc_dec(idev, &maddr);
1370 }
1371
1372 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1373 {
1374         struct in6_addr addr;
1375         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1376         if (ipv6_addr_any(&addr))
1377                 return;
1378         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1379 }
1380
1381 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1382 {
1383         struct in6_addr addr;
1384         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1385         if (ipv6_addr_any(&addr))
1386                 return;
1387         __ipv6_dev_ac_dec(ifp->idev, &addr);
1388 }
1389
1390 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1391 {
1392         if (dev->addr_len != ETH_ALEN)
1393                 return -1;
1394         memcpy(eui, dev->dev_addr, 3);
1395         memcpy(eui + 5, dev->dev_addr + 3, 3);
1396
1397         /*
1398          * The zSeries OSA network cards can be shared among various
1399          * OS instances, but the OSA cards have only one MAC address.
1400          * This leads to duplicate address conflicts in conjunction
1401          * with IPv6 if more than one instance uses the same card.
1402          *
1403          * The driver for these cards can deliver a unique 16-bit
1404          * identifier for each instance sharing the same card.  It is
1405          * placed instead of 0xFFFE in the interface identifier.  The
1406          * "u" bit of the interface identifier is not inverted in this
1407          * case.  Hence the resulting interface identifier has local
1408          * scope according to RFC2373.
1409          */
1410         if (dev->dev_id) {
1411                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1412                 eui[4] = dev->dev_id & 0xFF;
1413         } else {
1414                 eui[3] = 0xFF;
1415                 eui[4] = 0xFE;
1416                 eui[0] ^= 2;
1417         }
1418         return 0;
1419 }
1420
1421 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1422 {
1423         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1424         if (dev->addr_len != ARCNET_ALEN)
1425                 return -1;
1426         memset(eui, 0, 7);
1427         eui[7] = *(u8*)dev->dev_addr;
1428         return 0;
1429 }
1430
1431 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1432 {
1433         if (dev->addr_len != INFINIBAND_ALEN)
1434                 return -1;
1435         memcpy(eui, dev->dev_addr + 12, 8);
1436         eui[0] |= 2;
1437         return 0;
1438 }
1439
1440 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1441 {
1442         switch (dev->type) {
1443         case ARPHRD_ETHER:
1444         case ARPHRD_FDDI:
1445         case ARPHRD_IEEE802_TR:
1446                 return addrconf_ifid_eui48(eui, dev);
1447         case ARPHRD_ARCNET:
1448                 return addrconf_ifid_arcnet(eui, dev);
1449         case ARPHRD_INFINIBAND:
1450                 return addrconf_ifid_infiniband(eui, dev);
1451         case ARPHRD_SIT:
1452                 if (dev->priv_flags & IFF_ISATAP)
1453                         return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1454         }
1455         return -1;
1456 }
1457
1458 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1459 {
1460         int err = -1;
1461         struct inet6_ifaddr *ifp;
1462
1463         read_lock_bh(&idev->lock);
1464         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1465                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1466                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1467                         err = 0;
1468                         break;
1469                 }
1470         }
1471         read_unlock_bh(&idev->lock);
1472         return err;
1473 }
1474
1475 #ifdef CONFIG_IPV6_PRIVACY
1476 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1477 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1478 {
1479 regen:
1480         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1481         idev->rndid[0] &= ~0x02;
1482
1483         /*
1484          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1485          * check if generated address is not inappropriate
1486          *
1487          *  - Reserved subnet anycast (RFC 2526)
1488          *      11111101 11....11 1xxxxxxx
1489          *  - ISATAP (RFC4214) 6.1
1490          *      00-00-5E-FE-xx-xx-xx-xx
1491          *  - value 0
1492          *  - XXX: already assigned to an address on the device
1493          */
1494         if (idev->rndid[0] == 0xfd &&
1495             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1496             (idev->rndid[7]&0x80))
1497                 goto regen;
1498         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1499                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1500                         goto regen;
1501                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1502                         goto regen;
1503         }
1504
1505         return 0;
1506 }
1507
1508 static void ipv6_regen_rndid(unsigned long data)
1509 {
1510         struct inet6_dev *idev = (struct inet6_dev *) data;
1511         unsigned long expires;
1512
1513         rcu_read_lock_bh();
1514         write_lock_bh(&idev->lock);
1515
1516         if (idev->dead)
1517                 goto out;
1518
1519         if (__ipv6_regen_rndid(idev) < 0)
1520                 goto out;
1521
1522         expires = jiffies +
1523                 idev->cnf.temp_prefered_lft * HZ -
1524                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1525         if (time_before(expires, jiffies)) {
1526                 printk(KERN_WARNING
1527                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1528                         idev->dev->name);
1529                 goto out;
1530         }
1531
1532         if (!mod_timer(&idev->regen_timer, expires))
1533                 in6_dev_hold(idev);
1534
1535 out:
1536         write_unlock_bh(&idev->lock);
1537         rcu_read_unlock_bh();
1538         in6_dev_put(idev);
1539 }
1540
1541 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1542         int ret = 0;
1543
1544         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1545                 ret = __ipv6_regen_rndid(idev);
1546         return ret;
1547 }
1548 #endif
1549
1550 /*
1551  *      Add prefix route.
1552  */
1553
1554 static void
1555 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1556                       unsigned long expires, u32 flags)
1557 {
1558         struct fib6_config cfg = {
1559                 .fc_table = RT6_TABLE_PREFIX,
1560                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1561                 .fc_ifindex = dev->ifindex,
1562                 .fc_expires = expires,
1563                 .fc_dst_len = plen,
1564                 .fc_flags = RTF_UP | flags,
1565                 .fc_nlinfo.nl_net = &init_net,
1566         };
1567
1568         ipv6_addr_copy(&cfg.fc_dst, pfx);
1569
1570         /* Prevent useless cloning on PtP SIT.
1571            This thing is done here expecting that the whole
1572            class of non-broadcast devices need not cloning.
1573          */
1574 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1575         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1576                 cfg.fc_flags |= RTF_NONEXTHOP;
1577 #endif
1578
1579         ip6_route_add(&cfg);
1580 }
1581
1582 /* Create "default" multicast route to the interface */
1583
1584 static void addrconf_add_mroute(struct net_device *dev)
1585 {
1586         struct fib6_config cfg = {
1587                 .fc_table = RT6_TABLE_LOCAL,
1588                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1589                 .fc_ifindex = dev->ifindex,
1590                 .fc_dst_len = 8,
1591                 .fc_flags = RTF_UP,
1592                 .fc_nlinfo.nl_net = &init_net,
1593         };
1594
1595         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1596
1597         ip6_route_add(&cfg);
1598 }
1599
1600 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1601 static void sit_route_add(struct net_device *dev)
1602 {
1603         struct fib6_config cfg = {
1604                 .fc_table = RT6_TABLE_MAIN,
1605                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1606                 .fc_ifindex = dev->ifindex,
1607                 .fc_dst_len = 96,
1608                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1609                 .fc_nlinfo.nl_net = &init_net,
1610         };
1611
1612         /* prefix length - 96 bits "::d.d.d.d" */
1613         ip6_route_add(&cfg);
1614 }
1615 #endif
1616
1617 static void addrconf_add_lroute(struct net_device *dev)
1618 {
1619         struct in6_addr addr;
1620
1621         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1622         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1623 }
1624
1625 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1626 {
1627         struct inet6_dev *idev;
1628
1629         ASSERT_RTNL();
1630
1631         if ((idev = ipv6_find_idev(dev)) == NULL)
1632                 return NULL;
1633
1634         /* Add default multicast route */
1635         addrconf_add_mroute(dev);
1636
1637         /* Add link local route */
1638         addrconf_add_lroute(dev);
1639         return idev;
1640 }
1641
1642 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1643 {
1644         struct prefix_info *pinfo;
1645         __u32 valid_lft;
1646         __u32 prefered_lft;
1647         int addr_type;
1648         unsigned long rt_expires;
1649         struct inet6_dev *in6_dev;
1650
1651         pinfo = (struct prefix_info *) opt;
1652
1653         if (len < sizeof(struct prefix_info)) {
1654                 ADBG(("addrconf: prefix option too short\n"));
1655                 return;
1656         }
1657
1658         /*
1659          *      Validation checks ([ADDRCONF], page 19)
1660          */
1661
1662         addr_type = ipv6_addr_type(&pinfo->prefix);
1663
1664         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1665                 return;
1666
1667         valid_lft = ntohl(pinfo->valid);
1668         prefered_lft = ntohl(pinfo->prefered);
1669
1670         if (prefered_lft > valid_lft) {
1671                 if (net_ratelimit())
1672                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1673                 return;
1674         }
1675
1676         in6_dev = in6_dev_get(dev);
1677
1678         if (in6_dev == NULL) {
1679                 if (net_ratelimit())
1680                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1681                 return;
1682         }
1683
1684         /*
1685          *      Two things going on here:
1686          *      1) Add routes for on-link prefixes
1687          *      2) Configure prefixes with the auto flag set
1688          */
1689
1690         /* Avoid arithmetic overflow. Really, we could
1691            save rt_expires in seconds, likely valid_lft,
1692            but it would require division in fib gc, that it
1693            not good.
1694          */
1695         if (valid_lft >= 0x7FFFFFFF/HZ)
1696                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1697         else
1698                 rt_expires = valid_lft * HZ;
1699
1700         /*
1701          * We convert this (in jiffies) to clock_t later.
1702          * Avoid arithmetic overflow there as well.
1703          * Overflow can happen only if HZ < USER_HZ.
1704          */
1705         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1706                 rt_expires = 0x7FFFFFFF / USER_HZ;
1707
1708         if (pinfo->onlink) {
1709                 struct rt6_info *rt;
1710                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1711
1712                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1713                         if (rt->rt6i_flags&RTF_EXPIRES) {
1714                                 if (valid_lft == 0) {
1715                                         ip6_del_rt(rt);
1716                                         rt = NULL;
1717                                 } else {
1718                                         rt->rt6i_expires = jiffies + rt_expires;
1719                                 }
1720                         }
1721                 } else if (valid_lft) {
1722                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1723                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1724                 }
1725                 if (rt)
1726                         dst_release(&rt->u.dst);
1727         }
1728
1729         /* Try to figure out our local address for this prefix */
1730
1731         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1732                 struct inet6_ifaddr * ifp;
1733                 struct in6_addr addr;
1734                 int create = 0, update_lft = 0;
1735
1736                 if (pinfo->prefix_len == 64) {
1737                         memcpy(&addr, &pinfo->prefix, 8);
1738                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1739                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1740                                 in6_dev_put(in6_dev);
1741                                 return;
1742                         }
1743                         goto ok;
1744                 }
1745                 if (net_ratelimit())
1746                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1747                                pinfo->prefix_len);
1748                 in6_dev_put(in6_dev);
1749                 return;
1750
1751 ok:
1752
1753                 ifp = ipv6_get_ifaddr(&init_net, &addr, dev, 1);
1754
1755                 if (ifp == NULL && valid_lft) {
1756                         int max_addresses = in6_dev->cnf.max_addresses;
1757                         u32 addr_flags = 0;
1758
1759 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1760                         if (in6_dev->cnf.optimistic_dad &&
1761                             !ipv6_devconf.forwarding)
1762                                 addr_flags = IFA_F_OPTIMISTIC;
1763 #endif
1764
1765                         /* Do not allow to create too much of autoconfigured
1766                          * addresses; this would be too easy way to crash kernel.
1767                          */
1768                         if (!max_addresses ||
1769                             ipv6_count_addresses(in6_dev) < max_addresses)
1770                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1771                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1772                                                     addr_flags);
1773
1774                         if (!ifp || IS_ERR(ifp)) {
1775                                 in6_dev_put(in6_dev);
1776                                 return;
1777                         }
1778
1779                         update_lft = create = 1;
1780                         ifp->cstamp = jiffies;
1781                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1782                 }
1783
1784                 if (ifp) {
1785                         int flags;
1786                         unsigned long now;
1787 #ifdef CONFIG_IPV6_PRIVACY
1788                         struct inet6_ifaddr *ift;
1789 #endif
1790                         u32 stored_lft;
1791
1792                         /* update lifetime (RFC2462 5.5.3 e) */
1793                         spin_lock(&ifp->lock);
1794                         now = jiffies;
1795                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1796                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1797                         else
1798                                 stored_lft = 0;
1799                         if (!update_lft && stored_lft) {
1800                                 if (valid_lft > MIN_VALID_LIFETIME ||
1801                                     valid_lft > stored_lft)
1802                                         update_lft = 1;
1803                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1804                                         /* valid_lft <= stored_lft is always true */
1805                                         /* XXX: IPsec */
1806                                         update_lft = 0;
1807                                 } else {
1808                                         valid_lft = MIN_VALID_LIFETIME;
1809                                         if (valid_lft < prefered_lft)
1810                                                 prefered_lft = valid_lft;
1811                                         update_lft = 1;
1812                                 }
1813                         }
1814
1815                         if (update_lft) {
1816                                 ifp->valid_lft = valid_lft;
1817                                 ifp->prefered_lft = prefered_lft;
1818                                 ifp->tstamp = now;
1819                                 flags = ifp->flags;
1820                                 ifp->flags &= ~IFA_F_DEPRECATED;
1821                                 spin_unlock(&ifp->lock);
1822
1823                                 if (!(flags&IFA_F_TENTATIVE))
1824                                         ipv6_ifa_notify(0, ifp);
1825                         } else
1826                                 spin_unlock(&ifp->lock);
1827
1828 #ifdef CONFIG_IPV6_PRIVACY
1829                         read_lock_bh(&in6_dev->lock);
1830                         /* update all temporary addresses in the list */
1831                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1832                                 /*
1833                                  * When adjusting the lifetimes of an existing
1834                                  * temporary address, only lower the lifetimes.
1835                                  * Implementations must not increase the
1836                                  * lifetimes of an existing temporary address
1837                                  * when processing a Prefix Information Option.
1838                                  */
1839                                 spin_lock(&ift->lock);
1840                                 flags = ift->flags;
1841                                 if (ift->valid_lft > valid_lft &&
1842                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1843                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1844                                 if (ift->prefered_lft > prefered_lft &&
1845                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1846                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1847                                 spin_unlock(&ift->lock);
1848                                 if (!(flags&IFA_F_TENTATIVE))
1849                                         ipv6_ifa_notify(0, ift);
1850                         }
1851
1852                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1853                                 /*
1854                                  * When a new public address is created as described in [ADDRCONF],
1855                                  * also create a new temporary address.
1856                                  */
1857                                 read_unlock_bh(&in6_dev->lock);
1858                                 ipv6_create_tempaddr(ifp, NULL);
1859                         } else {
1860                                 read_unlock_bh(&in6_dev->lock);
1861                         }
1862 #endif
1863                         in6_ifa_put(ifp);
1864                         addrconf_verify(0);
1865                 }
1866         }
1867         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1868         in6_dev_put(in6_dev);
1869 }
1870
1871 /*
1872  *      Set destination address.
1873  *      Special case for SIT interfaces where we create a new "virtual"
1874  *      device.
1875  */
1876 int addrconf_set_dstaddr(void __user *arg)
1877 {
1878         struct in6_ifreq ireq;
1879         struct net_device *dev;
1880         int err = -EINVAL;
1881
1882         rtnl_lock();
1883
1884         err = -EFAULT;
1885         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1886                 goto err_exit;
1887
1888         dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1889
1890         err = -ENODEV;
1891         if (dev == NULL)
1892                 goto err_exit;
1893
1894 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1895         if (dev->type == ARPHRD_SIT) {
1896                 struct ifreq ifr;
1897                 mm_segment_t    oldfs;
1898                 struct ip_tunnel_parm p;
1899
1900                 err = -EADDRNOTAVAIL;
1901                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1902                         goto err_exit;
1903
1904                 memset(&p, 0, sizeof(p));
1905                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1906                 p.iph.saddr = 0;
1907                 p.iph.version = 4;
1908                 p.iph.ihl = 5;
1909                 p.iph.protocol = IPPROTO_IPV6;
1910                 p.iph.ttl = 64;
1911                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1912
1913                 oldfs = get_fs(); set_fs(KERNEL_DS);
1914                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1915                 set_fs(oldfs);
1916
1917                 if (err == 0) {
1918                         err = -ENOBUFS;
1919                         if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1920                                 goto err_exit;
1921                         err = dev_open(dev);
1922                 }
1923         }
1924 #endif
1925
1926 err_exit:
1927         rtnl_unlock();
1928         return err;
1929 }
1930
1931 /*
1932  *      Manual configuration of address on an interface
1933  */
1934 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1935                           __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1936 {
1937         struct inet6_ifaddr *ifp;
1938         struct inet6_dev *idev;
1939         struct net_device *dev;
1940         int scope;
1941         u32 flags = RTF_EXPIRES;
1942
1943         ASSERT_RTNL();
1944
1945         /* check the lifetime */
1946         if (!valid_lft || prefered_lft > valid_lft)
1947                 return -EINVAL;
1948
1949         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1950                 return -ENODEV;
1951
1952         if ((idev = addrconf_add_dev(dev)) == NULL)
1953                 return -ENOBUFS;
1954
1955         scope = ipv6_addr_scope(pfx);
1956
1957         if (valid_lft == INFINITY_LIFE_TIME) {
1958                 ifa_flags |= IFA_F_PERMANENT;
1959                 flags = 0;
1960         } else if (valid_lft >= 0x7FFFFFFF/HZ)
1961                 valid_lft = 0x7FFFFFFF/HZ;
1962
1963         if (prefered_lft == 0)
1964                 ifa_flags |= IFA_F_DEPRECATED;
1965         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1966                  (prefered_lft != INFINITY_LIFE_TIME))
1967                 prefered_lft = 0x7FFFFFFF/HZ;
1968
1969         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1970
1971         if (!IS_ERR(ifp)) {
1972                 spin_lock_bh(&ifp->lock);
1973                 ifp->valid_lft = valid_lft;
1974                 ifp->prefered_lft = prefered_lft;
1975                 ifp->tstamp = jiffies;
1976                 spin_unlock_bh(&ifp->lock);
1977
1978                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1979                                       jiffies_to_clock_t(valid_lft * HZ), flags);
1980                 /*
1981                  * Note that section 3.1 of RFC 4429 indicates
1982                  * that the Optimistic flag should not be set for
1983                  * manually configured addresses
1984                  */
1985                 addrconf_dad_start(ifp, 0);
1986                 in6_ifa_put(ifp);
1987                 addrconf_verify(0);
1988                 return 0;
1989         }
1990
1991         return PTR_ERR(ifp);
1992 }
1993
1994 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1995 {
1996         struct inet6_ifaddr *ifp;
1997         struct inet6_dev *idev;
1998         struct net_device *dev;
1999
2000         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
2001                 return -ENODEV;
2002
2003         if ((idev = __in6_dev_get(dev)) == NULL)
2004                 return -ENXIO;
2005
2006         read_lock_bh(&idev->lock);
2007         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2008                 if (ifp->prefix_len == plen &&
2009                     ipv6_addr_equal(pfx, &ifp->addr)) {
2010                         in6_ifa_hold(ifp);
2011                         read_unlock_bh(&idev->lock);
2012
2013                         ipv6_del_addr(ifp);
2014
2015                         /* If the last address is deleted administratively,
2016                            disable IPv6 on this interface.
2017                          */
2018                         if (idev->addr_list == NULL)
2019                                 addrconf_ifdown(idev->dev, 1);
2020                         return 0;
2021                 }
2022         }
2023         read_unlock_bh(&idev->lock);
2024         return -EADDRNOTAVAIL;
2025 }
2026
2027
2028 int addrconf_add_ifaddr(void __user *arg)
2029 {
2030         struct in6_ifreq ireq;
2031         int err;
2032
2033         if (!capable(CAP_NET_ADMIN))
2034                 return -EPERM;
2035
2036         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2037                 return -EFAULT;
2038
2039         rtnl_lock();
2040         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2041                              IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2042         rtnl_unlock();
2043         return err;
2044 }
2045
2046 int addrconf_del_ifaddr(void __user *arg)
2047 {
2048         struct in6_ifreq ireq;
2049         int err;
2050
2051         if (!capable(CAP_NET_ADMIN))
2052                 return -EPERM;
2053
2054         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2055                 return -EFAULT;
2056
2057         rtnl_lock();
2058         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2059         rtnl_unlock();
2060         return err;
2061 }
2062
2063 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2064 static void sit_add_v4_addrs(struct inet6_dev *idev)
2065 {
2066         struct inet6_ifaddr * ifp;
2067         struct in6_addr addr;
2068         struct net_device *dev;
2069         int scope;
2070
2071         ASSERT_RTNL();
2072
2073         memset(&addr, 0, sizeof(struct in6_addr));
2074         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2075
2076         if (idev->dev->flags&IFF_POINTOPOINT) {
2077                 addr.s6_addr32[0] = htonl(0xfe800000);
2078                 scope = IFA_LINK;
2079         } else {
2080                 scope = IPV6_ADDR_COMPATv4;
2081         }
2082
2083         if (addr.s6_addr32[3]) {
2084                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2085                 if (!IS_ERR(ifp)) {
2086                         spin_lock_bh(&ifp->lock);
2087                         ifp->flags &= ~IFA_F_TENTATIVE;
2088                         spin_unlock_bh(&ifp->lock);
2089                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2090                         in6_ifa_put(ifp);
2091                 }
2092                 return;
2093         }
2094
2095         for_each_netdev(&init_net, dev) {
2096                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2097                 if (in_dev && (dev->flags & IFF_UP)) {
2098                         struct in_ifaddr * ifa;
2099
2100                         int flag = scope;
2101
2102                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2103                                 int plen;
2104
2105                                 addr.s6_addr32[3] = ifa->ifa_local;
2106
2107                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2108                                         continue;
2109                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2110                                         if (idev->dev->flags&IFF_POINTOPOINT)
2111                                                 continue;
2112                                         flag |= IFA_HOST;
2113                                 }
2114                                 if (idev->dev->flags&IFF_POINTOPOINT)
2115                                         plen = 64;
2116                                 else
2117                                         plen = 96;
2118
2119                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2120                                                     IFA_F_PERMANENT);
2121                                 if (!IS_ERR(ifp)) {
2122                                         spin_lock_bh(&ifp->lock);
2123                                         ifp->flags &= ~IFA_F_TENTATIVE;
2124                                         spin_unlock_bh(&ifp->lock);
2125                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2126                                         in6_ifa_put(ifp);
2127                                 }
2128                         }
2129                 }
2130         }
2131 }
2132 #endif
2133
2134 static void init_loopback(struct net_device *dev)
2135 {
2136         struct inet6_dev  *idev;
2137         struct inet6_ifaddr * ifp;
2138
2139         /* ::1 */
2140
2141         ASSERT_RTNL();
2142
2143         if ((idev = ipv6_find_idev(dev)) == NULL) {
2144                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2145                 return;
2146         }
2147
2148         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2149         if (!IS_ERR(ifp)) {
2150                 spin_lock_bh(&ifp->lock);
2151                 ifp->flags &= ~IFA_F_TENTATIVE;
2152                 spin_unlock_bh(&ifp->lock);
2153                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2154                 in6_ifa_put(ifp);
2155         }
2156 }
2157
2158 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2159 {
2160         struct inet6_ifaddr * ifp;
2161         u32 addr_flags = IFA_F_PERMANENT;
2162
2163 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2164         if (idev->cnf.optimistic_dad &&
2165             !ipv6_devconf.forwarding)
2166                 addr_flags |= IFA_F_OPTIMISTIC;
2167 #endif
2168
2169
2170         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2171         if (!IS_ERR(ifp)) {
2172                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2173                 addrconf_dad_start(ifp, 0);
2174                 in6_ifa_put(ifp);
2175         }
2176 }
2177
2178 static void addrconf_dev_config(struct net_device *dev)
2179 {
2180         struct in6_addr addr;
2181         struct inet6_dev    * idev;
2182
2183         ASSERT_RTNL();
2184
2185         if ((dev->type != ARPHRD_ETHER) &&
2186             (dev->type != ARPHRD_FDDI) &&
2187             (dev->type != ARPHRD_IEEE802_TR) &&
2188             (dev->type != ARPHRD_ARCNET) &&
2189             (dev->type != ARPHRD_INFINIBAND)) {
2190                 /* Alas, we support only Ethernet autoconfiguration. */
2191                 return;
2192         }
2193
2194         idev = addrconf_add_dev(dev);
2195         if (idev == NULL)
2196                 return;
2197
2198         memset(&addr, 0, sizeof(struct in6_addr));
2199         addr.s6_addr32[0] = htonl(0xFE800000);
2200
2201         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2202                 addrconf_add_linklocal(idev, &addr);
2203 }
2204
2205 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2206 static void addrconf_sit_config(struct net_device *dev)
2207 {
2208         struct inet6_dev *idev;
2209
2210         ASSERT_RTNL();
2211
2212         /*
2213          * Configure the tunnel with one of our IPv4
2214          * addresses... we should configure all of
2215          * our v4 addrs in the tunnel
2216          */
2217
2218         if ((idev = ipv6_find_idev(dev)) == NULL) {
2219                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2220                 return;
2221         }
2222
2223         if (dev->priv_flags & IFF_ISATAP) {
2224                 struct in6_addr addr;
2225
2226                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2227                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2228                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2229                         addrconf_add_linklocal(idev, &addr);
2230                 return;
2231         }
2232
2233         sit_add_v4_addrs(idev);
2234
2235         if (dev->flags&IFF_POINTOPOINT) {
2236                 addrconf_add_mroute(dev);
2237                 addrconf_add_lroute(dev);
2238         } else
2239                 sit_route_add(dev);
2240 }
2241 #endif
2242
2243 static inline int
2244 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2245 {
2246         struct in6_addr lladdr;
2247
2248         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2249                 addrconf_add_linklocal(idev, &lladdr);
2250                 return 0;
2251         }
2252         return -1;
2253 }
2254
2255 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2256 {
2257         struct net_device *link_dev;
2258
2259         /* first try to inherit the link-local address from the link device */
2260         if (idev->dev->iflink &&
2261             (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2262                 if (!ipv6_inherit_linklocal(idev, link_dev))
2263                         return;
2264         }
2265         /* then try to inherit it from any device */
2266         for_each_netdev(&init_net, link_dev) {
2267                 if (!ipv6_inherit_linklocal(idev, link_dev))
2268                         return;
2269         }
2270         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2271 }
2272
2273 /*
2274  * Autoconfigure tunnel with a link-local address so routing protocols,
2275  * DHCPv6, MLD etc. can be run over the virtual link
2276  */
2277
2278 static void addrconf_ip6_tnl_config(struct net_device *dev)
2279 {
2280         struct inet6_dev *idev;
2281
2282         ASSERT_RTNL();
2283
2284         if ((idev = addrconf_add_dev(dev)) == NULL) {
2285                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2286                 return;
2287         }
2288         ip6_tnl_add_linklocal(idev);
2289 }
2290
2291 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2292                            void * data)
2293 {
2294         struct net_device *dev = (struct net_device *) data;
2295         struct inet6_dev *idev = __in6_dev_get(dev);
2296         int run_pending = 0;
2297         int err;
2298
2299         if (dev->nd_net != &init_net)
2300                 return NOTIFY_DONE;
2301
2302         switch(event) {
2303         case NETDEV_REGISTER:
2304                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2305                         idev = ipv6_add_dev(dev);
2306                         if (!idev)
2307                                 return notifier_from_errno(-ENOMEM);
2308                 }
2309                 break;
2310         case NETDEV_UP:
2311         case NETDEV_CHANGE:
2312                 if (dev->flags & IFF_SLAVE)
2313                         break;
2314
2315                 if (event == NETDEV_UP) {
2316                         if (!addrconf_qdisc_ok(dev)) {
2317                                 /* device is not ready yet. */
2318                                 printk(KERN_INFO
2319                                         "ADDRCONF(NETDEV_UP): %s: "
2320                                         "link is not ready\n",
2321                                         dev->name);
2322                                 break;
2323                         }
2324
2325                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2326                                 idev = ipv6_add_dev(dev);
2327
2328                         if (idev)
2329                                 idev->if_flags |= IF_READY;
2330                 } else {
2331                         if (!addrconf_qdisc_ok(dev)) {
2332                                 /* device is still not ready. */
2333                                 break;
2334                         }
2335
2336                         if (idev) {
2337                                 if (idev->if_flags & IF_READY) {
2338                                         /* device is already configured. */
2339                                         break;
2340                                 }
2341                                 idev->if_flags |= IF_READY;
2342                         }
2343
2344                         printk(KERN_INFO
2345                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2346                                         "link becomes ready\n",
2347                                         dev->name);
2348
2349                         run_pending = 1;
2350                 }
2351
2352                 switch(dev->type) {
2353 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2354                 case ARPHRD_SIT:
2355                         addrconf_sit_config(dev);
2356                         break;
2357 #endif
2358                 case ARPHRD_TUNNEL6:
2359                         addrconf_ip6_tnl_config(dev);
2360                         break;
2361                 case ARPHRD_LOOPBACK:
2362                         init_loopback(dev);
2363                         break;
2364
2365                 default:
2366                         addrconf_dev_config(dev);
2367                         break;
2368                 }
2369                 if (idev) {
2370                         if (run_pending)
2371                                 addrconf_dad_run(idev);
2372
2373                         /* If the MTU changed during the interface down, when the
2374                            interface up, the changed MTU must be reflected in the
2375                            idev as well as routers.
2376                          */
2377                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2378                                 rt6_mtu_change(dev, dev->mtu);
2379                                 idev->cnf.mtu6 = dev->mtu;
2380                         }
2381                         idev->tstamp = jiffies;
2382                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2383                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2384                            stop IPv6 on this interface.
2385                          */
2386                         if (dev->mtu < IPV6_MIN_MTU)
2387                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2388                 }
2389                 break;
2390
2391         case NETDEV_CHANGEMTU:
2392                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2393                         rt6_mtu_change(dev, dev->mtu);
2394                         idev->cnf.mtu6 = dev->mtu;
2395                         break;
2396                 }
2397
2398                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2399                         idev = ipv6_add_dev(dev);
2400                         if (idev)
2401                                 break;
2402                 }
2403
2404                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2405
2406         case NETDEV_DOWN:
2407         case NETDEV_UNREGISTER:
2408                 /*
2409                  *      Remove all addresses from this interface.
2410                  */
2411                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2412                 break;
2413
2414         case NETDEV_CHANGENAME:
2415                 if (idev) {
2416                         snmp6_unregister_dev(idev);
2417                         addrconf_sysctl_unregister(idev);
2418                         addrconf_sysctl_register(idev);
2419                         err = snmp6_register_dev(idev);
2420                         if (err)
2421                                 return notifier_from_errno(err);
2422                 }
2423                 break;
2424         }
2425
2426         return NOTIFY_OK;
2427 }
2428
2429 /*
2430  *      addrconf module should be notified of a device going up
2431  */
2432 static struct notifier_block ipv6_dev_notf = {
2433         .notifier_call = addrconf_notify,
2434         .priority = 0
2435 };
2436
2437 static int addrconf_ifdown(struct net_device *dev, int how)
2438 {
2439         struct inet6_dev *idev;
2440         struct inet6_ifaddr *ifa, **bifa;
2441         int i;
2442
2443         ASSERT_RTNL();
2444
2445         if (dev == init_net.loopback_dev && how == 1)
2446                 how = 0;
2447
2448         rt6_ifdown(dev);
2449         neigh_ifdown(&nd_tbl, dev);
2450
2451         idev = __in6_dev_get(dev);
2452         if (idev == NULL)
2453                 return -ENODEV;
2454
2455         /* Step 1: remove reference to ipv6 device from parent device.
2456                    Do not dev_put!
2457          */
2458         if (how == 1) {
2459                 idev->dead = 1;
2460
2461                 /* protected by rtnl_lock */
2462                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2463
2464                 /* Step 1.5: remove snmp6 entry */
2465                 snmp6_unregister_dev(idev);
2466
2467         }
2468
2469         /* Step 2: clear hash table */
2470         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2471                 bifa = &inet6_addr_lst[i];
2472
2473                 write_lock_bh(&addrconf_hash_lock);
2474                 while ((ifa = *bifa) != NULL) {
2475                         if (ifa->idev == idev) {
2476                                 *bifa = ifa->lst_next;
2477                                 ifa->lst_next = NULL;
2478                                 addrconf_del_timer(ifa);
2479                                 in6_ifa_put(ifa);
2480                                 continue;
2481                         }
2482                         bifa = &ifa->lst_next;
2483                 }
2484                 write_unlock_bh(&addrconf_hash_lock);
2485         }
2486
2487         write_lock_bh(&idev->lock);
2488
2489         /* Step 3: clear flags for stateless addrconf */
2490         if (how != 1)
2491                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2492
2493         /* Step 4: clear address list */
2494 #ifdef CONFIG_IPV6_PRIVACY
2495         if (how == 1 && del_timer(&idev->regen_timer))
2496                 in6_dev_put(idev);
2497
2498         /* clear tempaddr list */
2499         while ((ifa = idev->tempaddr_list) != NULL) {
2500                 idev->tempaddr_list = ifa->tmp_next;
2501                 ifa->tmp_next = NULL;
2502                 ifa->dead = 1;
2503                 write_unlock_bh(&idev->lock);
2504                 spin_lock_bh(&ifa->lock);
2505
2506                 if (ifa->ifpub) {
2507                         in6_ifa_put(ifa->ifpub);
2508                         ifa->ifpub = NULL;
2509                 }
2510                 spin_unlock_bh(&ifa->lock);
2511                 in6_ifa_put(ifa);
2512                 write_lock_bh(&idev->lock);
2513         }
2514 #endif
2515         while ((ifa = idev->addr_list) != NULL) {
2516                 idev->addr_list = ifa->if_next;
2517                 ifa->if_next = NULL;
2518                 ifa->dead = 1;
2519                 addrconf_del_timer(ifa);
2520                 write_unlock_bh(&idev->lock);
2521
2522                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2523                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2524                 in6_ifa_put(ifa);
2525
2526                 write_lock_bh(&idev->lock);
2527         }
2528         write_unlock_bh(&idev->lock);
2529
2530         /* Step 5: Discard multicast list */
2531
2532         if (how == 1)
2533                 ipv6_mc_destroy_dev(idev);
2534         else
2535                 ipv6_mc_down(idev);
2536
2537         idev->tstamp = jiffies;
2538
2539         /* Shot the device (if unregistered) */
2540
2541         if (how == 1) {
2542                 addrconf_sysctl_unregister(idev);
2543                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2544                 neigh_ifdown(&nd_tbl, dev);
2545                 in6_dev_put(idev);
2546         }
2547         return 0;
2548 }
2549
2550 static void addrconf_rs_timer(unsigned long data)
2551 {
2552         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2553
2554         if (ifp->idev->cnf.forwarding)
2555                 goto out;
2556
2557         if (ifp->idev->if_flags & IF_RA_RCVD) {
2558                 /*
2559                  *      Announcement received after solicitation
2560                  *      was sent
2561                  */
2562                 goto out;
2563         }
2564
2565         spin_lock(&ifp->lock);
2566         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2567                 struct in6_addr all_routers;
2568
2569                 /* The wait after the last probe can be shorter */
2570                 addrconf_mod_timer(ifp, AC_RS,
2571                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2572                                    ifp->idev->cnf.rtr_solicit_delay :
2573                                    ifp->idev->cnf.rtr_solicit_interval);
2574                 spin_unlock(&ifp->lock);
2575
2576                 ipv6_addr_all_routers(&all_routers);
2577
2578                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2579         } else {
2580                 spin_unlock(&ifp->lock);
2581                 /*
2582                  * Note: we do not support deprecated "all on-link"
2583                  * assumption any longer.
2584                  */
2585                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2586                        ifp->idev->dev->name);
2587         }
2588
2589 out:
2590         in6_ifa_put(ifp);
2591 }
2592
2593 /*
2594  *      Duplicate Address Detection
2595  */
2596 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2597 {
2598         unsigned long rand_num;
2599         struct inet6_dev *idev = ifp->idev;
2600
2601         if (ifp->flags & IFA_F_OPTIMISTIC)
2602                 rand_num = 0;
2603         else
2604                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2605
2606         ifp->probes = idev->cnf.dad_transmits;
2607         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2608 }
2609
2610 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2611 {
2612         struct inet6_dev *idev = ifp->idev;
2613         struct net_device *dev = idev->dev;
2614
2615         addrconf_join_solict(dev, &ifp->addr);
2616
2617         net_srandom(ifp->addr.s6_addr32[3]);
2618
2619         read_lock_bh(&idev->lock);
2620         if (ifp->dead)
2621                 goto out;
2622         spin_lock_bh(&ifp->lock);
2623
2624         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2625             !(ifp->flags&IFA_F_TENTATIVE) ||
2626             ifp->flags & IFA_F_NODAD) {
2627                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2628                 spin_unlock_bh(&ifp->lock);
2629                 read_unlock_bh(&idev->lock);
2630
2631                 addrconf_dad_completed(ifp);
2632                 return;
2633         }
2634
2635         if (!(idev->if_flags & IF_READY)) {
2636                 spin_unlock_bh(&ifp->lock);
2637                 read_unlock_bh(&idev->lock);
2638                 /*
2639                  * If the defice is not ready:
2640                  * - keep it tentative if it is a permanent address.
2641                  * - otherwise, kill it.
2642                  */
2643                 in6_ifa_hold(ifp);
2644                 addrconf_dad_stop(ifp);
2645                 return;
2646         }
2647
2648         /*
2649          * Optimistic nodes can start receiving
2650          * Frames right away
2651          */
2652         if(ifp->flags & IFA_F_OPTIMISTIC)
2653                 ip6_ins_rt(ifp->rt);
2654
2655         addrconf_dad_kick(ifp);
2656         spin_unlock_bh(&ifp->lock);
2657 out:
2658         read_unlock_bh(&idev->lock);
2659 }
2660
2661 static void addrconf_dad_timer(unsigned long data)
2662 {
2663         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2664         struct inet6_dev *idev = ifp->idev;
2665         struct in6_addr unspec;
2666         struct in6_addr mcaddr;
2667
2668         read_lock_bh(&idev->lock);
2669         if (idev->dead) {
2670                 read_unlock_bh(&idev->lock);
2671                 goto out;
2672         }
2673         spin_lock_bh(&ifp->lock);
2674         if (ifp->probes == 0) {
2675                 /*
2676                  * DAD was successful
2677                  */
2678
2679                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2680                 spin_unlock_bh(&ifp->lock);
2681                 read_unlock_bh(&idev->lock);
2682
2683                 addrconf_dad_completed(ifp);
2684
2685                 goto out;
2686         }
2687
2688         ifp->probes--;
2689         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2690         spin_unlock_bh(&ifp->lock);
2691         read_unlock_bh(&idev->lock);
2692
2693         /* send a neighbour solicitation for our addr */
2694         memset(&unspec, 0, sizeof(unspec));
2695         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2696         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2697 out:
2698         in6_ifa_put(ifp);
2699 }
2700
2701 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2702 {
2703         struct net_device *     dev = ifp->idev->dev;
2704
2705         /*
2706          *      Configure the address for reception. Now it is valid.
2707          */
2708
2709         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2710
2711         /* If added prefix is link local and forwarding is off,
2712            start sending router solicitations.
2713          */
2714
2715         if (ifp->idev->cnf.forwarding == 0 &&
2716             ifp->idev->cnf.rtr_solicits > 0 &&
2717             (dev->flags&IFF_LOOPBACK) == 0 &&
2718             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2719                 struct in6_addr all_routers;
2720
2721                 ipv6_addr_all_routers(&all_routers);
2722
2723                 /*
2724                  *      If a host as already performed a random delay
2725                  *      [...] as part of DAD [...] there is no need
2726                  *      to delay again before sending the first RS
2727                  */
2728                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2729
2730                 spin_lock_bh(&ifp->lock);
2731                 ifp->probes = 1;
2732                 ifp->idev->if_flags |= IF_RS_SENT;
2733                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2734                 spin_unlock_bh(&ifp->lock);
2735         }
2736 }
2737
2738 static void addrconf_dad_run(struct inet6_dev *idev) {
2739         struct inet6_ifaddr *ifp;
2740
2741         read_lock_bh(&idev->lock);
2742         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2743                 spin_lock_bh(&ifp->lock);
2744                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2745                         spin_unlock_bh(&ifp->lock);
2746                         continue;
2747                 }
2748                 spin_unlock_bh(&ifp->lock);
2749                 addrconf_dad_kick(ifp);
2750         }
2751         read_unlock_bh(&idev->lock);
2752 }
2753
2754 #ifdef CONFIG_PROC_FS
2755 struct if6_iter_state {
2756         struct seq_net_private p;
2757         int bucket;
2758 };
2759
2760 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2761 {
2762         struct inet6_ifaddr *ifa = NULL;
2763         struct if6_iter_state *state = seq->private;
2764         struct net *net = state->p.net;
2765
2766         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2767                 ifa = inet6_addr_lst[state->bucket];
2768
2769                 while (ifa && ifa->idev->dev->nd_net != net)
2770                         ifa = ifa->lst_next;
2771                 if (ifa)
2772                         break;
2773         }
2774         return ifa;
2775 }
2776
2777 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2778 {
2779         struct if6_iter_state *state = seq->private;
2780         struct net *net = state->p.net;
2781
2782         ifa = ifa->lst_next;
2783 try_again:
2784         if (ifa) {
2785                 if (ifa->idev->dev->nd_net != net) {
2786                         ifa = ifa->lst_next;
2787                         goto try_again;
2788                 }
2789         }
2790
2791         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2792                 ifa = inet6_addr_lst[state->bucket];
2793                 goto try_again;
2794         }
2795
2796         return ifa;
2797 }
2798
2799 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2800 {
2801         struct inet6_ifaddr *ifa = if6_get_first(seq);
2802
2803         if (ifa)
2804                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2805                         --pos;
2806         return pos ? NULL : ifa;
2807 }
2808
2809 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2810         __acquires(addrconf_hash_lock)
2811 {
2812         read_lock_bh(&addrconf_hash_lock);
2813         return if6_get_idx(seq, *pos);
2814 }
2815
2816 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2817 {
2818         struct inet6_ifaddr *ifa;
2819
2820         ifa = if6_get_next(seq, v);
2821         ++*pos;
2822         return ifa;
2823 }
2824
2825 static void if6_seq_stop(struct seq_file *seq, void *v)
2826         __releases(addrconf_hash_lock)
2827 {
2828         read_unlock_bh(&addrconf_hash_lock);
2829 }
2830
2831 static int if6_seq_show(struct seq_file *seq, void *v)
2832 {
2833         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2834         seq_printf(seq,
2835                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2836                    NIP6(ifp->addr),
2837                    ifp->idev->dev->ifindex,
2838                    ifp->prefix_len,
2839                    ifp->scope,
2840                    ifp->flags,
2841                    ifp->idev->dev->name);
2842         return 0;
2843 }
2844
2845 static const struct seq_operations if6_seq_ops = {
2846         .start  = if6_seq_start,
2847         .next   = if6_seq_next,
2848         .show   = if6_seq_show,
2849         .stop   = if6_seq_stop,
2850 };
2851
2852 static int if6_seq_open(struct inode *inode, struct file *file)
2853 {
2854         return seq_open_net(inode, file, &if6_seq_ops,
2855                             sizeof(struct if6_iter_state));
2856 }
2857
2858 static const struct file_operations if6_fops = {
2859         .owner          = THIS_MODULE,
2860         .open           = if6_seq_open,
2861         .read           = seq_read,
2862         .llseek         = seq_lseek,
2863         .release        = seq_release_net,
2864 };
2865
2866 static int if6_proc_net_init(struct net *net)
2867 {
2868         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2869                 return -ENOMEM;
2870         return 0;
2871 }
2872
2873 static void if6_proc_net_exit(struct net *net)
2874 {
2875        proc_net_remove(net, "if_inet6");
2876 }
2877
2878 static struct pernet_operations if6_proc_net_ops = {
2879        .init = if6_proc_net_init,
2880        .exit = if6_proc_net_exit,
2881 };
2882
2883 int __init if6_proc_init(void)
2884 {
2885         return register_pernet_subsys(&if6_proc_net_ops);
2886 }
2887
2888 void if6_proc_exit(void)
2889 {
2890         unregister_pernet_subsys(&if6_proc_net_ops);
2891 }
2892 #endif  /* CONFIG_PROC_FS */
2893
2894 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2895 /* Check if address is a home address configured on any interface. */
2896 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2897 {
2898         int ret = 0;
2899         struct inet6_ifaddr * ifp;
2900         u8 hash = ipv6_addr_hash(addr);
2901         read_lock_bh(&addrconf_hash_lock);
2902         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2903                 if (ifp->idev->dev->nd_net != net)
2904                         continue;
2905                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2906                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2907                         ret = 1;
2908                         break;
2909                 }
2910         }
2911         read_unlock_bh(&addrconf_hash_lock);
2912         return ret;
2913 }
2914 #endif
2915
2916 /*
2917  *      Periodic address status verification
2918  */
2919
2920 static void addrconf_verify(unsigned long foo)
2921 {
2922         struct inet6_ifaddr *ifp;
2923         unsigned long now, next;
2924         int i;
2925
2926         spin_lock_bh(&addrconf_verify_lock);
2927         now = jiffies;
2928         next = now + ADDR_CHECK_FREQUENCY;
2929
2930         del_timer(&addr_chk_timer);
2931
2932         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2933
2934 restart:
2935                 read_lock(&addrconf_hash_lock);
2936                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2937                         unsigned long age;
2938 #ifdef CONFIG_IPV6_PRIVACY
2939                         unsigned long regen_advance;
2940 #endif
2941
2942                         if (ifp->flags & IFA_F_PERMANENT)
2943                                 continue;
2944
2945                         spin_lock(&ifp->lock);
2946                         age = (now - ifp->tstamp) / HZ;
2947
2948 #ifdef CONFIG_IPV6_PRIVACY
2949                         regen_advance = ifp->idev->cnf.regen_max_retry *
2950                                         ifp->idev->cnf.dad_transmits *
2951                                         ifp->idev->nd_parms->retrans_time / HZ;
2952 #endif
2953
2954                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2955                             age >= ifp->valid_lft) {
2956                                 spin_unlock(&ifp->lock);
2957                                 in6_ifa_hold(ifp);
2958                                 read_unlock(&addrconf_hash_lock);
2959                                 ipv6_del_addr(ifp);
2960                                 goto restart;
2961                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2962                                 spin_unlock(&ifp->lock);
2963                                 continue;
2964                         } else if (age >= ifp->prefered_lft) {
2965                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2966                                 int deprecate = 0;
2967
2968                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2969                                         deprecate = 1;
2970                                         ifp->flags |= IFA_F_DEPRECATED;
2971                                 }
2972
2973                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2974                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2975
2976                                 spin_unlock(&ifp->lock);
2977
2978                                 if (deprecate) {
2979                                         in6_ifa_hold(ifp);
2980                                         read_unlock(&addrconf_hash_lock);
2981
2982                                         ipv6_ifa_notify(0, ifp);
2983                                         in6_ifa_put(ifp);
2984                                         goto restart;
2985                                 }
2986 #ifdef CONFIG_IPV6_PRIVACY
2987                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2988                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2989                                 if (age >= ifp->prefered_lft - regen_advance) {
2990                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2991                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2992                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2993                                         if (!ifp->regen_count && ifpub) {
2994                                                 ifp->regen_count++;
2995                                                 in6_ifa_hold(ifp);
2996                                                 in6_ifa_hold(ifpub);
2997                                                 spin_unlock(&ifp->lock);
2998                                                 read_unlock(&addrconf_hash_lock);
2999                                                 spin_lock(&ifpub->lock);
3000                                                 ifpub->regen_count = 0;
3001                                                 spin_unlock(&ifpub->lock);
3002                                                 ipv6_create_tempaddr(ifpub, ifp);
3003                                                 in6_ifa_put(ifpub);
3004                                                 in6_ifa_put(ifp);
3005                                                 goto restart;
3006                                         }
3007                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3008                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3009                                 spin_unlock(&ifp->lock);
3010 #endif
3011                         } else {
3012                                 /* ifp->prefered_lft <= ifp->valid_lft */
3013                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3014                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3015                                 spin_unlock(&ifp->lock);
3016                         }
3017                 }
3018                 read_unlock(&addrconf_hash_lock);
3019         }
3020
3021         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3022         add_timer(&addr_chk_timer);
3023         spin_unlock_bh(&addrconf_verify_lock);
3024 }
3025
3026 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3027 {
3028         struct in6_addr *pfx = NULL;
3029
3030         if (addr)
3031                 pfx = nla_data(addr);
3032
3033         if (local) {
3034                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3035                         pfx = NULL;
3036                 else
3037                         pfx = nla_data(local);
3038         }
3039
3040         return pfx;
3041 }
3042
3043 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3044         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3045         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3046         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3047 };
3048
3049 static int
3050 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3051 {
3052         struct net *net = skb->sk->sk_net;
3053         struct ifaddrmsg *ifm;
3054         struct nlattr *tb[IFA_MAX+1];
3055         struct in6_addr *pfx;
3056         int err;
3057
3058         if (net != &init_net)
3059                 return -EINVAL;
3060
3061         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3062         if (err < 0)
3063                 return err;
3064
3065         ifm = nlmsg_data(nlh);
3066         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3067         if (pfx == NULL)
3068                 return -EINVAL;
3069
3070         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3071 }
3072
3073 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3074                              u32 prefered_lft, u32 valid_lft)
3075 {
3076         u32 flags = RTF_EXPIRES;
3077
3078         if (!valid_lft || (prefered_lft > valid_lft))
3079                 return -EINVAL;
3080
3081         if (valid_lft == INFINITY_LIFE_TIME) {
3082                 ifa_flags |= IFA_F_PERMANENT;
3083                 flags = 0;
3084         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3085                 valid_lft = 0x7FFFFFFF/HZ;
3086
3087         if (prefered_lft == 0)
3088                 ifa_flags |= IFA_F_DEPRECATED;
3089         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3090                  (prefered_lft != INFINITY_LIFE_TIME))
3091                 prefered_lft = 0x7FFFFFFF/HZ;
3092
3093         spin_lock_bh(&ifp->lock);
3094         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3095         ifp->tstamp = jiffies;
3096         ifp->valid_lft = valid_lft;
3097         ifp->prefered_lft = prefered_lft;
3098
3099         spin_unlock_bh(&ifp->lock);
3100         if (!(ifp->flags&IFA_F_TENTATIVE))
3101                 ipv6_ifa_notify(0, ifp);
3102
3103         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3104                               jiffies_to_clock_t(valid_lft * HZ), flags);
3105         addrconf_verify(0);
3106
3107         return 0;
3108 }
3109
3110 static int
3111 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3112 {
3113         struct net *net = skb->sk->sk_net;
3114         struct ifaddrmsg *ifm;
3115         struct nlattr *tb[IFA_MAX+1];
3116         struct in6_addr *pfx;
3117         struct inet6_ifaddr *ifa;
3118         struct net_device *dev;
3119         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3120         u8 ifa_flags;
3121         int err;
3122
3123         if (net != &init_net)
3124                 return -EINVAL;
3125
3126         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3127         if (err < 0)
3128                 return err;
3129
3130         ifm = nlmsg_data(nlh);
3131         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3132         if (pfx == NULL)
3133                 return -EINVAL;
3134
3135         if (tb[IFA_CACHEINFO]) {
3136                 struct ifa_cacheinfo *ci;
3137
3138                 ci = nla_data(tb[IFA_CACHEINFO]);
3139                 valid_lft = ci->ifa_valid;
3140                 preferred_lft = ci->ifa_prefered;
3141         } else {
3142                 preferred_lft = INFINITY_LIFE_TIME;
3143                 valid_lft = INFINITY_LIFE_TIME;
3144         }
3145
3146         dev =  __dev_get_by_index(&init_net, ifm->ifa_index);
3147         if (dev == NULL)
3148                 return -ENODEV;
3149
3150         /* We ignore other flags so far. */
3151         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3152
3153         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3154         if (ifa == NULL) {
3155                 /*
3156                  * It would be best to check for !NLM_F_CREATE here but
3157                  * userspace alreay relies on not having to provide this.
3158                  */
3159                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3160                                       ifa_flags, preferred_lft, valid_lft);
3161         }
3162
3163         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3164             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3165                 err = -EEXIST;
3166         else
3167                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3168
3169         in6_ifa_put(ifa);
3170
3171         return err;
3172 }
3173
3174 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3175                           u8 scope, int ifindex)
3176 {
3177         struct ifaddrmsg *ifm;
3178
3179         ifm = nlmsg_data(nlh);
3180         ifm->ifa_family = AF_INET6;
3181         ifm->ifa_prefixlen = prefixlen;
3182         ifm->ifa_flags = flags;
3183         ifm->ifa_scope = scope;
3184         ifm->ifa_index = ifindex;
3185 }
3186
3187 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3188                          unsigned long tstamp, u32 preferred, u32 valid)
3189 {
3190         struct ifa_cacheinfo ci;
3191
3192         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3193                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3194         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3195                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3196         ci.ifa_prefered = preferred;
3197         ci.ifa_valid = valid;
3198
3199         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3200 }
3201
3202 static inline int rt_scope(int ifa_scope)
3203 {
3204         if (ifa_scope & IFA_HOST)
3205                 return RT_SCOPE_HOST;
3206         else if (ifa_scope & IFA_LINK)
3207                 return RT_SCOPE_LINK;
3208         else if (ifa_scope & IFA_SITE)
3209                 return RT_SCOPE_SITE;
3210         else
3211                 return RT_SCOPE_UNIVERSE;
3212 }
3213
3214 static inline int inet6_ifaddr_msgsize(void)
3215 {
3216         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3217                + nla_total_size(16) /* IFA_ADDRESS */
3218                + nla_total_size(sizeof(struct ifa_cacheinfo));
3219 }
3220
3221 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3222                              u32 pid, u32 seq, int event, unsigned int flags)
3223 {
3224         struct nlmsghdr  *nlh;
3225         u32 preferred, valid;
3226
3227         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3228         if (nlh == NULL)
3229                 return -EMSGSIZE;
3230
3231         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3232                       ifa->idev->dev->ifindex);
3233
3234         if (!(ifa->flags&IFA_F_PERMANENT)) {
3235                 preferred = ifa->prefered_lft;
3236                 valid = ifa->valid_lft;
3237                 if (preferred != INFINITY_LIFE_TIME) {
3238                         long tval = (jiffies - ifa->tstamp)/HZ;
3239                         preferred -= tval;
3240                         if (valid != INFINITY_LIFE_TIME)
3241                                 valid -= tval;
3242                 }
3243         } else {
3244                 preferred = INFINITY_LIFE_TIME;
3245                 valid = INFINITY_LIFE_TIME;
3246         }
3247
3248         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3249             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3250                 nlmsg_cancel(skb, nlh);
3251                 return -EMSGSIZE;
3252         }
3253
3254         return nlmsg_end(skb, nlh);
3255 }
3256
3257 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3258                                 u32 pid, u32 seq, int event, u16 flags)
3259 {
3260         struct nlmsghdr  *nlh;
3261         u8 scope = RT_SCOPE_UNIVERSE;
3262         int ifindex = ifmca->idev->dev->ifindex;
3263
3264         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3265                 scope = RT_SCOPE_SITE;
3266
3267         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3268         if (nlh == NULL)
3269                 return -EMSGSIZE;
3270
3271         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3272         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3273             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3274                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3275                 nlmsg_cancel(skb, nlh);
3276                 return -EMSGSIZE;
3277         }
3278
3279         return nlmsg_end(skb, nlh);
3280 }
3281
3282 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3283                                 u32 pid, u32 seq, int event, unsigned int flags)
3284 {
3285         struct nlmsghdr  *nlh;
3286         u8 scope = RT_SCOPE_UNIVERSE;
3287         int ifindex = ifaca->aca_idev->dev->ifindex;
3288
3289         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3290                 scope = RT_SCOPE_SITE;
3291
3292         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3293         if (nlh == NULL)
3294                 return -EMSGSIZE;
3295
3296         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3297         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3298             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3299                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3300                 nlmsg_cancel(skb, nlh);
3301                 return -EMSGSIZE;
3302         }
3303
3304         return nlmsg_end(skb, nlh);
3305 }
3306
3307 enum addr_type_t
3308 {
3309         UNICAST_ADDR,
3310         MULTICAST_ADDR,
3311         ANYCAST_ADDR,
3312 };
3313
3314 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3315                            enum addr_type_t type)
3316 {
3317         int idx, ip_idx;
3318         int s_idx, s_ip_idx;
3319         int err = 1;
3320         struct net_device *dev;
3321         struct inet6_dev *idev = NULL;
3322         struct inet6_ifaddr *ifa;
3323         struct ifmcaddr6 *ifmca;
3324         struct ifacaddr6 *ifaca;
3325
3326         s_idx = cb->args[0];
3327         s_ip_idx = ip_idx = cb->args[1];
3328
3329         idx = 0;
3330         for_each_netdev(&init_net, dev) {
3331                 if (idx < s_idx)
3332                         goto cont;
3333                 if (idx > s_idx)
3334                         s_ip_idx = 0;
3335                 ip_idx = 0;
3336                 if ((idev = in6_dev_get(dev)) == NULL)
3337                         goto cont;
3338                 read_lock_bh(&idev->lock);
3339                 switch (type) {
3340                 case UNICAST_ADDR:
3341                         /* unicast address incl. temp addr */
3342                         for (ifa = idev->addr_list; ifa;
3343                              ifa = ifa->if_next, ip_idx++) {
3344                                 if (ip_idx < s_ip_idx)
3345                                         continue;
3346                                 err = inet6_fill_ifaddr(skb, ifa,
3347                                                         NETLINK_CB(cb->skb).pid,
3348                                                         cb->nlh->nlmsg_seq,
3349                                                         RTM_NEWADDR,
3350                                                         NLM_F_MULTI);
3351                         }
3352                         break;
3353                 case MULTICAST_ADDR:
3354                         /* multicast address */
3355                         for (ifmca = idev->mc_list; ifmca;
3356                              ifmca = ifmca->next, ip_idx++) {
3357                                 if (ip_idx < s_ip_idx)
3358                                         continue;
3359                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3360                                                           NETLINK_CB(cb->skb).pid,
3361                                                           cb->nlh->nlmsg_seq,
3362                                                           RTM_GETMULTICAST,
3363                                                           NLM_F_MULTI);
3364                         }
3365                         break;
3366                 case ANYCAST_ADDR:
3367                         /* anycast address */
3368                         for (ifaca = idev->ac_list; ifaca;
3369                              ifaca = ifaca->aca_next, ip_idx++) {
3370                                 if (ip_idx < s_ip_idx)
3371                                         continue;
3372                                 err = inet6_fill_ifacaddr(skb, ifaca,
3373                                                           NETLINK_CB(cb->skb).pid,
3374                                                           cb->nlh->nlmsg_seq,
3375                                                           RTM_GETANYCAST,
3376                                                           NLM_F_MULTI);
3377                         }
3378                         break;
3379                 default:
3380                         break;
3381                 }
3382                 read_unlock_bh(&idev->lock);
3383                 in6_dev_put(idev);
3384
3385                 if (err <= 0)
3386                         break;
3387 cont:
3388                 idx++;
3389         }
3390         cb->args[0] = idx;
3391         cb->args[1] = ip_idx;
3392         return skb->len;
3393 }
3394
3395 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3396 {
3397         struct net *net = skb->sk->sk_net;
3398         enum addr_type_t type = UNICAST_ADDR;
3399
3400         if (net != &init_net)
3401                 return 0;
3402
3403         return inet6_dump_addr(skb, cb, type);
3404 }
3405
3406 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3407 {
3408         struct net *net = skb->sk->sk_net;
3409         enum addr_type_t type = MULTICAST_ADDR;
3410
3411         if (net != &init_net)
3412                 return 0;
3413
3414         return inet6_dump_addr(skb, cb, type);
3415 }
3416
3417
3418 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3419 {
3420         struct net *net = skb->sk->sk_net;
3421         enum addr_type_t type = ANYCAST_ADDR;
3422
3423         if (net != &init_net)
3424                 return 0;
3425
3426         return inet6_dump_addr(skb, cb, type);
3427 }
3428
3429 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3430                              void *arg)
3431 {
3432         struct net *net = in_skb->sk->sk_net;
3433         struct ifaddrmsg *ifm;
3434         struct nlattr *tb[IFA_MAX+1];
3435         struct in6_addr *addr = NULL;
3436         struct net_device *dev = NULL;
3437         struct inet6_ifaddr *ifa;
3438         struct sk_buff *skb;
3439         int err;
3440
3441         if (net != &init_net)
3442                 return -EINVAL;
3443
3444         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3445         if (err < 0)
3446                 goto errout;
3447
3448         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3449         if (addr == NULL) {
3450                 err = -EINVAL;
3451                 goto errout;
3452         }
3453
3454         ifm = nlmsg_data(nlh);
3455         if (ifm->ifa_index)
3456                 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3457
3458         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3459                 err = -EADDRNOTAVAIL;
3460                 goto errout;
3461         }
3462
3463         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3464                 err = -ENOBUFS;
3465                 goto errout_ifa;
3466         }
3467
3468         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3469                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3470         if (err < 0) {
3471                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3472                 WARN_ON(err == -EMSGSIZE);
3473                 kfree_skb(skb);
3474                 goto errout_ifa;
3475         }
3476         err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
3477 errout_ifa:
3478         in6_ifa_put(ifa);
3479 errout:
3480         return err;
3481 }
3482
3483 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3484 {
3485         struct sk_buff *skb;
3486         int err = -ENOBUFS;
3487
3488         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3489         if (skb == NULL)
3490                 goto errout;
3491
3492         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3493         if (err < 0) {
3494                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3495                 WARN_ON(err == -EMSGSIZE);
3496                 kfree_skb(skb);
3497                 goto errout;
3498         }
3499         err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3500 errout:
3501         if (err < 0)
3502                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3503 }
3504
3505 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3506                                 __s32 *array, int bytes)
3507 {
3508         BUG_ON(bytes < (DEVCONF_MAX * 4));
3509
3510         memset(array, 0, bytes);
3511         array[DEVCONF_FORWARDING] = cnf->forwarding;
3512         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3513         array[DEVCONF_MTU6] = cnf->mtu6;
3514         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3515         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3516         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3517         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3518         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3519         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3520         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3521         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3522 #ifdef CONFIG_IPV6_PRIVACY
3523         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3524         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3525         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3526         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3527         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3528 #endif
3529         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3530         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3531         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3532 #ifdef CONFIG_IPV6_ROUTER_PREF
3533         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3534         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3535 #ifdef CONFIG_IPV6_ROUTE_INFO
3536         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3537 #endif
3538 #endif
3539         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3540         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3541 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3542         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3543 #endif
3544 }
3545
3546 static inline size_t inet6_if_nlmsg_size(void)
3547 {
3548         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3549                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3550                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3551                + nla_total_size(4) /* IFLA_MTU */
3552                + nla_total_size(4) /* IFLA_LINK */
3553                + nla_total_size( /* IFLA_PROTINFO */
3554                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3555                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3556                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3557                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3558                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3559                  );
3560 }
3561
3562 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3563                                       int bytes)
3564 {
3565         int i;
3566         int pad = bytes - sizeof(u64) * items;
3567         BUG_ON(pad < 0);
3568
3569         /* Use put_unaligned() because stats may not be aligned for u64. */
3570         put_unaligned(items, &stats[0]);
3571         for (i = 1; i < items; i++)
3572                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3573
3574         memset(&stats[items], 0, pad);
3575 }
3576
3577 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3578                              int bytes)
3579 {
3580         switch(attrtype) {
3581         case IFLA_INET6_STATS:
3582                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3583                 break;
3584         case IFLA_INET6_ICMP6STATS:
3585                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3586                 break;
3587         }
3588 }
3589
3590 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3591                              u32 pid, u32 seq, int event, unsigned int flags)
3592 {
3593         struct net_device *dev = idev->dev;
3594         struct nlattr *nla;
3595         struct ifinfomsg *hdr;
3596         struct nlmsghdr *nlh;
3597         void *protoinfo;
3598         struct ifla_cacheinfo ci;
3599
3600         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3601         if (nlh == NULL)
3602                 return -EMSGSIZE;
3603
3604         hdr = nlmsg_data(nlh);
3605         hdr->ifi_family = AF_INET6;
3606         hdr->__ifi_pad = 0;
3607         hdr->ifi_type = dev->type;
3608         hdr->ifi_index = dev->ifindex;
3609         hdr->ifi_flags = dev_get_flags(dev);
3610         hdr->ifi_change = 0;
3611
3612         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3613
3614         if (dev->addr_len)
3615                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3616
3617         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3618         if (dev->ifindex != dev->iflink)
3619                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3620
3621         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3622         if (protoinfo == NULL)
3623                 goto nla_put_failure;
3624
3625         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3626
3627         ci.max_reasm_len = IPV6_MAXPLEN;
3628         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3629                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3630         ci.reachable_time = idev->nd_parms->reachable_time;
3631         ci.retrans_time = idev->nd_parms->retrans_time;
3632         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3633
3634         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3635         if (nla == NULL)
3636                 goto nla_put_failure;
3637         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3638
3639         /* XXX - MC not implemented */
3640
3641         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3642         if (nla == NULL)
3643                 goto nla_put_failure;
3644         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3645
3646         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3647         if (nla == NULL)
3648                 goto nla_put_failure;
3649         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3650
3651         nla_nest_end(skb, protoinfo);
3652         return nlmsg_end(skb, nlh);
3653
3654 nla_put_failure:
3655         nlmsg_cancel(skb, nlh);
3656         return -EMSGSIZE;
3657 }
3658
3659 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3660 {
3661         struct net *net = skb->sk->sk_net;
3662         int idx, err;
3663         int s_idx = cb->args[0];
3664         struct net_device *dev;
3665         struct inet6_dev *idev;
3666
3667         if (net != &init_net)
3668                 return 0;
3669
3670         read_lock(&dev_base_lock);
3671         idx = 0;
3672         for_each_netdev(&init_net, dev) {
3673                 if (idx < s_idx)
3674                         goto cont;
3675                 if ((idev = in6_dev_get(dev)) == NULL)
3676                         goto cont;
3677                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3678                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3679                 in6_dev_put(idev);
3680                 if (err <= 0)
3681                         break;
3682 cont:
3683                 idx++;
3684         }
3685         read_unlock(&dev_base_lock);
3686         cb->args[0] = idx;
3687
3688         return skb->len;
3689 }
3690
3691 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3692 {
3693         struct sk_buff *skb;
3694         int err = -ENOBUFS;
3695
3696         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3697         if (skb == NULL)
3698                 goto errout;
3699
3700         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3701         if (err < 0) {
3702                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3703                 WARN_ON(err == -EMSGSIZE);
3704                 kfree_skb(skb);
3705                 goto errout;
3706         }
3707         err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3708 errout:
3709         if (err < 0)
3710                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3711 }
3712
3713 static inline size_t inet6_prefix_nlmsg_size(void)
3714 {
3715         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3716                + nla_total_size(sizeof(struct in6_addr))
3717                + nla_total_size(sizeof(struct prefix_cacheinfo));
3718 }
3719
3720 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3721                              struct prefix_info *pinfo, u32 pid, u32 seq,
3722                              int event, unsigned int flags)
3723 {
3724         struct prefixmsg *pmsg;
3725         struct nlmsghdr *nlh;
3726         struct prefix_cacheinfo ci;
3727
3728         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3729         if (nlh == NULL)
3730                 return -EMSGSIZE;
3731
3732         pmsg = nlmsg_data(nlh);
3733         pmsg->prefix_family = AF_INET6;
3734         pmsg->prefix_pad1 = 0;
3735         pmsg->prefix_pad2 = 0;
3736         pmsg->prefix_ifindex = idev->dev->ifindex;
3737         pmsg->prefix_len = pinfo->prefix_len;
3738         pmsg->prefix_type = pinfo->type;
3739         pmsg->prefix_pad3 = 0;
3740         pmsg->prefix_flags = 0;
3741         if (pinfo->onlink)
3742                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3743         if (pinfo->autoconf)
3744                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3745
3746         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3747
3748         ci.preferred_time = ntohl(pinfo->prefered);
3749         ci.valid_time = ntohl(pinfo->valid);
3750         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3751
3752         return nlmsg_end(skb, nlh);
3753
3754 nla_put_failure:
3755         nlmsg_cancel(skb, nlh);
3756         return -EMSGSIZE;
3757 }
3758
3759 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3760                          struct prefix_info *pinfo)
3761 {
3762         struct sk_buff *skb;
3763         int err = -ENOBUFS;
3764
3765         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3766         if (skb == NULL)
3767                 goto errout;
3768
3769         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3770         if (err < 0) {
3771                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3772                 WARN_ON(err == -EMSGSIZE);
3773                 kfree_skb(skb);
3774                 goto errout;
3775         }
3776         err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3777 errout:
3778         if (err < 0)
3779                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_PREFIX, err);
3780 }
3781
3782 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3783 {
3784         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3785
3786         switch (event) {
3787         case RTM_NEWADDR:
3788                 /*
3789                  * If the address was optimistic
3790                  * we inserted the route at the start of
3791                  * our DAD process, so we don't need
3792                  * to do it again
3793                  */
3794                 if (!(ifp->rt->rt6i_node))
3795                         ip6_ins_rt(ifp->rt);
3796                 if (ifp->idev->cnf.forwarding)
3797                         addrconf_join_anycast(ifp);
3798                 break;
3799         case RTM_DELADDR:
3800                 if (ifp->idev->cnf.forwarding)
3801                         addrconf_leave_anycast(ifp);
3802                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3803                 dst_hold(&ifp->rt->u.dst);
3804                 if (ip6_del_rt(ifp->rt))
3805                         dst_free(&ifp->rt->u.dst);
3806                 break;
3807         }
3808 }
3809
3810 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3811 {
3812         rcu_read_lock_bh();
3813         if (likely(ifp->idev->dead == 0))
3814                 __ipv6_ifa_notify(event, ifp);
3815         rcu_read_unlock_bh();
3816 }
3817
3818 #ifdef CONFIG_SYSCTL
3819
3820 static
3821 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3822                            void __user *buffer, size_t *lenp, loff_t *ppos)
3823 {
3824         int *valp = ctl->data;
3825         int val = *valp;
3826         int ret;
3827
3828         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3829
3830         if (write)
3831                 addrconf_fixup_forwarding(ctl, valp, val);
3832         return ret;
3833 }
3834
3835 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3836                                             int __user *name, int nlen,
3837                                             void __user *oldval,
3838                                             size_t __user *oldlenp,
3839                                             void __user *newval, size_t newlen)
3840 {
3841         int *valp = table->data;
3842         int val = *valp;
3843         int new;
3844
3845         if (!newval || !newlen)
3846                 return 0;
3847         if (newlen != sizeof(int))
3848                 return -EINVAL;
3849         if (get_user(new, (int __user *)newval))
3850                 return -EFAULT;
3851         if (new == *valp)
3852                 return 0;
3853         if (oldval && oldlenp) {
3854                 size_t len;
3855                 if (get_user(len, oldlenp))
3856                         return -EFAULT;
3857                 if (len) {
3858                         if (len > table->maxlen)
3859                                 len = table->maxlen;
3860                         if (copy_to_user(oldval, valp, len))
3861                                 return -EFAULT;
3862                         if (put_user(len, oldlenp))
3863                                 return -EFAULT;
3864                 }
3865         }
3866
3867         *valp = new;
3868         addrconf_fixup_forwarding(table, valp, val);
3869         return 1;
3870 }
3871
3872 static struct addrconf_sysctl_table
3873 {
3874         struct ctl_table_header *sysctl_header;
3875         ctl_table addrconf_vars[__NET_IPV6_MAX];
3876         char *dev_name;
3877 } addrconf_sysctl __read_mostly = {
3878         .sysctl_header = NULL,
3879         .addrconf_vars = {
3880                 {
3881                         .ctl_name       =       NET_IPV6_FORWARDING,
3882                         .procname       =       "forwarding",
3883                         .data           =       &ipv6_devconf.forwarding,
3884                         .maxlen         =       sizeof(int),
3885                         .mode           =       0644,
3886                         .proc_handler   =       &addrconf_sysctl_forward,
3887                         .strategy       =       &addrconf_sysctl_forward_strategy,
3888                 },
3889                 {
3890                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3891                         .procname       =       "hop_limit",
3892                         .data           =       &ipv6_devconf.hop_limit,
3893                         .maxlen         =       sizeof(int),
3894                         .mode           =       0644,
3895                         .proc_handler   =       proc_dointvec,
3896                 },
3897                 {
3898                         .ctl_name       =       NET_IPV6_MTU,
3899                         .procname       =       "mtu",
3900                         .data           =       &ipv6_devconf.mtu6,
3901                         .maxlen         =       sizeof(int),
3902                         .mode           =       0644,
3903                         .proc_handler   =       &proc_dointvec,
3904                 },
3905                 {
3906                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3907                         .procname       =       "accept_ra",
3908                         .data           =       &ipv6_devconf.accept_ra,
3909                         .maxlen         =       sizeof(int),
3910                         .mode           =       0644,
3911                         .proc_handler   =       &proc_dointvec,
3912                 },
3913                 {
3914                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3915                         .procname       =       "accept_redirects",
3916                         .data           =       &ipv6_devconf.accept_redirects,
3917                         .maxlen         =       sizeof(int),
3918                         .mode           =       0644,
3919                         .proc_handler   =       &proc_dointvec,
3920                 },
3921                 {
3922                         .ctl_name       =       NET_IPV6_AUTOCONF,
3923                         .procname       =       "autoconf",
3924                         .data           =       &ipv6_devconf.autoconf,
3925                         .maxlen         =       sizeof(int),
3926                         .mode           =       0644,
3927                         .proc_handler   =       &proc_dointvec,
3928                 },
3929                 {
3930                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3931                         .procname       =       "dad_transmits",
3932                         .data           =       &ipv6_devconf.dad_transmits,
3933                         .maxlen         =       sizeof(int),
3934                         .mode           =       0644,
3935                         .proc_handler   =       &proc_dointvec,
3936                 },
3937                 {
3938                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3939                         .procname       =       "router_solicitations",
3940                         .data           =       &ipv6_devconf.rtr_solicits,
3941                         .maxlen         =       sizeof(int),
3942                         .mode           =       0644,
3943                         .proc_handler   =       &proc_dointvec,
3944                 },
3945                 {
3946                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3947                         .procname       =       "router_solicitation_interval",
3948                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3949                         .maxlen         =       sizeof(int),
3950                         .mode           =       0644,
3951                         .proc_handler   =       &proc_dointvec_jiffies,
3952                         .strategy       =       &sysctl_jiffies,
3953                 },
3954                 {
3955                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3956                         .procname       =       "router_solicitation_delay",
3957                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3958                         .maxlen         =       sizeof(int),
3959                         .mode           =       0644,
3960                         .proc_handler   =       &proc_dointvec_jiffies,
3961                         .strategy       =       &sysctl_jiffies,
3962                 },
3963                 {
3964                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3965                         .procname       =       "force_mld_version",
3966                         .data           =       &ipv6_devconf.force_mld_version,
3967                         .maxlen         =       sizeof(int),
3968                         .mode           =       0644,
3969                         .proc_handler   =       &proc_dointvec,
3970                 },
3971 #ifdef CONFIG_IPV6_PRIVACY
3972                 {
3973                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3974                         .procname       =       "use_tempaddr",
3975                         .data           =       &ipv6_devconf.use_tempaddr,
3976                         .maxlen         =       sizeof(int),
3977                         .mode           =       0644,
3978                         .proc_handler   =       &proc_dointvec,
3979                 },
3980                 {
3981                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3982                         .procname       =       "temp_valid_lft",
3983                         .data           =       &ipv6_devconf.temp_valid_lft,
3984                         .maxlen         =       sizeof(int),
3985                         .mode           =       0644,
3986                         .proc_handler   =       &proc_dointvec,
3987                 },
3988                 {
3989                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3990                         .procname       =       "temp_prefered_lft",
3991                         .data           =       &ipv6_devconf.temp_prefered_lft,
3992                         .maxlen         =       sizeof(int),
3993                         .mode           =       0644,
3994                         .proc_handler   =       &proc_dointvec,
3995                 },
3996                 {
3997                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3998                         .procname       =       "regen_max_retry",
3999                         .data           =       &ipv6_devconf.regen_max_retry,
4000                         .maxlen         =       sizeof(int),
4001                         .mode           =       0644,
4002                         .proc_handler   =       &proc_dointvec,
4003                 },
4004                 {
4005                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4006                         .procname       =       "max_desync_factor",
4007                         .data           =       &ipv6_devconf.max_desync_factor,
4008                         .maxlen         =       sizeof(int),
4009                         .mode           =       0644,
4010                         .proc_handler   =       &proc_dointvec,
4011                 },
4012 #endif
4013                 {
4014                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4015                         .procname       =       "max_addresses",
4016                         .data           =       &ipv6_devconf.max_addresses,
4017                         .maxlen         =       sizeof(int),
4018                         .mode           =       0644,
4019                         .proc_handler   =       &proc_dointvec,
4020                 },
4021                 {
4022                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4023                         .procname       =       "accept_ra_defrtr",
4024                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4025                         .maxlen         =       sizeof(int),
4026                         .mode           =       0644,
4027                         .proc_handler   =       &proc_dointvec,
4028                 },
4029                 {
4030                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4031                         .procname       =       "accept_ra_pinfo",
4032                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4033                         .maxlen         =       sizeof(int),
4034                         .mode           =       0644,
4035                         .proc_handler   =       &proc_dointvec,
4036                 },
4037 #ifdef CONFIG_IPV6_ROUTER_PREF
4038                 {
4039                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4040                         .procname       =       "accept_ra_rtr_pref",
4041                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4042                         .maxlen         =       sizeof(int),
4043                         .mode           =       0644,
4044                         .proc_handler   =       &proc_dointvec,
4045                 },
4046                 {
4047                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4048                         .procname       =       "router_probe_interval",
4049                         .data           =       &ipv6_devconf.rtr_probe_interval,
4050                         .maxlen         =       sizeof(int),
4051                         .mode           =       0644,
4052                         .proc_handler   =       &proc_dointvec_jiffies,
4053                         .strategy       =       &sysctl_jiffies,
4054                 },
4055 #ifdef CONFIG_IPV6_ROUTE_INFO
4056                 {
4057                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4058                         .procname       =       "accept_ra_rt_info_max_plen",
4059                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4060                         .maxlen         =       sizeof(int),
4061                         .mode           =       0644,
4062                         .proc_handler   =       &proc_dointvec,
4063                 },
4064 #endif
4065 #endif
4066                 {
4067                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4068                         .procname       =       "proxy_ndp",
4069                         .data           =       &ipv6_devconf.proxy_ndp,
4070                         .maxlen         =       sizeof(int),
4071                         .mode           =       0644,
4072                         .proc_handler   =       &proc_dointvec,
4073                 },
4074                 {
4075                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4076                         .procname       =       "accept_source_route",
4077                         .data           =       &ipv6_devconf.accept_source_route,
4078                         .maxlen         =       sizeof(int),
4079                         .mode           =       0644,
4080                         .proc_handler   =       &proc_dointvec,
4081                 },
4082 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4083                 {
4084                         .ctl_name       =       CTL_UNNUMBERED,
4085                         .procname       =       "optimistic_dad",
4086                         .data           =       &ipv6_devconf.optimistic_dad,
4087                         .maxlen         =       sizeof(int),
4088                         .mode           =       0644,
4089                         .proc_handler   =       &proc_dointvec,
4090
4091                 },
4092 #endif
4093                 {
4094                         .ctl_name       =       0,      /* sentinel */
4095                 }
4096         },
4097 };
4098
4099 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4100                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4101 {
4102         int i;
4103         struct addrconf_sysctl_table *t;
4104
4105 #define ADDRCONF_CTL_PATH_DEV   3
4106
4107         struct ctl_path addrconf_ctl_path[] = {
4108                 { .procname = "net", .ctl_name = CTL_NET, },
4109                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4110                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4111                 { /* to be set */ },
4112                 { },
4113         };
4114
4115
4116         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4117         if (t == NULL)
4118                 goto out;
4119
4120         for (i=0; t->addrconf_vars[i].data; i++) {
4121                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4122                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4123                 t->addrconf_vars[i].extra2 = net;
4124         }
4125
4126         /*
4127          * Make a copy of dev_name, because '.procname' is regarded as const
4128          * by sysctl and we wouldn't want anyone to change it under our feet
4129          * (see SIOCSIFNAME).
4130          */
4131         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4132         if (!t->dev_name)
4133                 goto free;
4134
4135         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4136         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4137
4138         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4139                         t->addrconf_vars);
4140         if (t->sysctl_header == NULL)
4141                 goto free_procname;
4142
4143         p->sysctl = t;
4144         return 0;
4145
4146 free_procname:
4147         kfree(t->dev_name);
4148 free:
4149         kfree(t);
4150 out:
4151         return -ENOBUFS;
4152 }
4153
4154 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4155 {
4156         struct addrconf_sysctl_table *t;
4157
4158         if (p->sysctl == NULL)
4159                 return;
4160
4161         t = p->sysctl;
4162         p->sysctl = NULL;
4163         unregister_sysctl_table(t->sysctl_header);
4164         kfree(t->dev_name);
4165         kfree(t);
4166 }
4167
4168 static void addrconf_sysctl_register(struct inet6_dev *idev)
4169 {
4170         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4171                               NET_IPV6_NEIGH, "ipv6",
4172                               &ndisc_ifinfo_sysctl_change,
4173                               NULL);
4174         __addrconf_sysctl_register(idev->dev->nd_net, idev->dev->name,
4175                         idev->dev->ifindex, idev, &idev->cnf);
4176 }
4177
4178 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4179 {
4180         __addrconf_sysctl_unregister(&idev->cnf);
4181         neigh_sysctl_unregister(idev->nd_parms);
4182 }
4183
4184
4185 #endif
4186
4187 static int addrconf_init_net(struct net *net)
4188 {
4189         int err;
4190         struct ipv6_devconf *all, *dflt;
4191
4192         err = -ENOMEM;
4193         all = &ipv6_devconf;
4194         dflt = &ipv6_devconf_dflt;
4195
4196         if (net != &init_net) {
4197                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4198                 if (all == NULL)
4199                         goto err_alloc_all;
4200
4201                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4202                 if (dflt == NULL)
4203                         goto err_alloc_dflt;
4204         }
4205
4206         net->ipv6.devconf_all = all;
4207         net->ipv6.devconf_dflt = dflt;
4208
4209 #ifdef CONFIG_SYSCTL
4210         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4211                         NULL, all);
4212         if (err < 0)
4213                 goto err_reg_all;
4214
4215         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4216                         NULL, dflt);
4217         if (err < 0)
4218                 goto err_reg_dflt;
4219 #endif
4220         return 0;
4221
4222 #ifdef CONFIG_SYSCTL
4223 err_reg_dflt:
4224         __addrconf_sysctl_unregister(all);
4225 err_reg_all:
4226         kfree(dflt);
4227 #endif
4228 err_alloc_dflt:
4229         kfree(all);
4230 err_alloc_all:
4231         return err;
4232 }
4233
4234 static void addrconf_exit_net(struct net *net)
4235 {
4236 #ifdef CONFIG_SYSCTL
4237         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4238         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4239 #endif
4240         if (net != &init_net) {
4241                 kfree(net->ipv6.devconf_dflt);
4242                 kfree(net->ipv6.devconf_all);
4243         }
4244 }
4245
4246 static struct pernet_operations addrconf_ops = {
4247         .init = addrconf_init_net,
4248         .exit = addrconf_exit_net,
4249 };
4250
4251 /*
4252  *      Device notifier
4253  */
4254
4255 int register_inet6addr_notifier(struct notifier_block *nb)
4256 {
4257         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4258 }
4259
4260 EXPORT_SYMBOL(register_inet6addr_notifier);
4261
4262 int unregister_inet6addr_notifier(struct notifier_block *nb)
4263 {
4264         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4265 }
4266
4267 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4268
4269 /*
4270  *      Init / cleanup code
4271  */
4272
4273 int __init addrconf_init(void)
4274 {
4275         int err;
4276
4277         if ((err = ipv6_addr_label_init()) < 0) {
4278                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4279                         err);
4280                 return err;
4281         }
4282
4283         register_pernet_subsys(&addrconf_ops);
4284
4285         /* The addrconf netdev notifier requires that loopback_dev
4286          * has it's ipv6 private information allocated and setup
4287          * before it can bring up and give link-local addresses
4288          * to other devices which are up.
4289          *
4290          * Unfortunately, loopback_dev is not necessarily the first
4291          * entry in the global dev_base list of net devices.  In fact,
4292          * it is likely to be the very last entry on that list.
4293          * So this causes the notifier registry below to try and
4294          * give link-local addresses to all devices besides loopback_dev
4295          * first, then loopback_dev, which cases all the non-loopback_dev
4296          * devices to fail to get a link-local address.
4297          *
4298          * So, as a temporary fix, allocate the ipv6 structure for
4299          * loopback_dev first by hand.
4300          * Longer term, all of the dependencies ipv6 has upon the loopback
4301          * device and it being up should be removed.
4302          */
4303         rtnl_lock();
4304         if (!ipv6_add_dev(init_net.loopback_dev))
4305                 err = -ENOMEM;
4306         rtnl_unlock();
4307         if (err)
4308                 goto errlo;
4309
4310         ip6_null_entry.u.dst.dev = init_net.loopback_dev;
4311         ip6_null_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4312 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4313         ip6_prohibit_entry.u.dst.dev = init_net.loopback_dev;
4314         ip6_prohibit_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4315         ip6_blk_hole_entry.u.dst.dev = init_net.loopback_dev;
4316         ip6_blk_hole_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4317 #endif
4318
4319         register_netdevice_notifier(&ipv6_dev_notf);
4320
4321         addrconf_verify(0);
4322
4323         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4324         if (err < 0)
4325                 goto errout;
4326
4327         /* Only the first call to __rtnl_register can fail */
4328         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4329         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4330         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4331         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4332         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4333
4334         ipv6_addr_label_rtnl_register();
4335
4336         return 0;
4337 errout:
4338         unregister_netdevice_notifier(&ipv6_dev_notf);
4339 errlo:
4340         unregister_pernet_subsys(&addrconf_ops);
4341
4342         return err;
4343 }
4344
4345 void addrconf_cleanup(void)
4346 {
4347         struct net_device *dev;
4348         struct inet6_ifaddr *ifa;
4349         int i;
4350
4351         unregister_netdevice_notifier(&ipv6_dev_notf);
4352
4353         unregister_pernet_subsys(&addrconf_ops);
4354
4355         rtnl_lock();
4356
4357         /*
4358          *      clean dev list.
4359          */
4360
4361         for_each_netdev(&init_net, dev) {
4362                 if (__in6_dev_get(dev) == NULL)
4363                         continue;
4364                 addrconf_ifdown(dev, 1);
4365         }
4366         addrconf_ifdown(init_net.loopback_dev, 2);
4367
4368         /*
4369          *      Check hash table.
4370          */
4371
4372         write_lock_bh(&addrconf_hash_lock);
4373         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4374                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4375                         struct inet6_ifaddr *bifa;
4376
4377                         bifa = ifa;
4378                         ifa = ifa->lst_next;
4379                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4380                         /* Do not free it; something is wrong.
4381                            Now we can investigate it with debugger.
4382                          */
4383                 }
4384         }
4385         write_unlock_bh(&addrconf_hash_lock);
4386
4387         del_timer(&addr_chk_timer);
4388
4389         rtnl_unlock();
4390 }