41cc31ee297e958fadcfcbea4cb6794741484015
[pandora-kernel.git] / 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                         /* Rule 8: Use longest matching prefix */
1129                         if (hiscore.rule < 8) {
1130                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1131                                 hiscore.rule++;
1132                         }
1133                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1134                         if (score.matchlen > hiscore.matchlen) {
1135                                 score.rule = 8;
1136                                 goto record_it;
1137                         }
1138 #if 0
1139                         else if (score.matchlen < hiscore.matchlen)
1140                                 continue;
1141 #endif
1142
1143                         /* Final Rule: choose first available one */
1144                         continue;
1145 record_it:
1146                         if (ifa_result)
1147                                 in6_ifa_put(ifa_result);
1148                         in6_ifa_hold(ifa);
1149                         ifa_result = ifa;
1150                         hiscore = score;
1151                 }
1152                 read_unlock_bh(&idev->lock);
1153         }
1154         rcu_read_unlock();
1155         read_unlock(&dev_base_lock);
1156
1157         if (!ifa_result)
1158                 return -EADDRNOTAVAIL;
1159
1160         ipv6_addr_copy(saddr, &ifa_result->addr);
1161         in6_ifa_put(ifa_result);
1162         return 0;
1163 }
1164
1165
1166 int ipv6_get_saddr(struct dst_entry *dst,
1167                    struct in6_addr *daddr, struct in6_addr *saddr)
1168 {
1169         return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1170 }
1171
1172 EXPORT_SYMBOL(ipv6_get_saddr);
1173
1174 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1175                     unsigned char banned_flags)
1176 {
1177         struct inet6_dev *idev;
1178         int err = -EADDRNOTAVAIL;
1179
1180         rcu_read_lock();
1181         if ((idev = __in6_dev_get(dev)) != NULL) {
1182                 struct inet6_ifaddr *ifp;
1183
1184                 read_lock_bh(&idev->lock);
1185                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1186                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1187                                 ipv6_addr_copy(addr, &ifp->addr);
1188                                 err = 0;
1189                                 break;
1190                         }
1191                 }
1192                 read_unlock_bh(&idev->lock);
1193         }
1194         rcu_read_unlock();
1195         return err;
1196 }
1197
1198 static int ipv6_count_addresses(struct inet6_dev *idev)
1199 {
1200         int cnt = 0;
1201         struct inet6_ifaddr *ifp;
1202
1203         read_lock_bh(&idev->lock);
1204         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1205                 cnt++;
1206         read_unlock_bh(&idev->lock);
1207         return cnt;
1208 }
1209
1210 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1211                   struct net_device *dev, int strict)
1212 {
1213         struct inet6_ifaddr * ifp;
1214         u8 hash = ipv6_addr_hash(addr);
1215
1216         read_lock_bh(&addrconf_hash_lock);
1217         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1218                 if (ifp->idev->dev->nd_net != net)
1219                         continue;
1220                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1221                     !(ifp->flags&IFA_F_TENTATIVE)) {
1222                         if (dev == NULL || ifp->idev->dev == dev ||
1223                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1224                                 break;
1225                 }
1226         }
1227         read_unlock_bh(&addrconf_hash_lock);
1228         return ifp != NULL;
1229 }
1230 EXPORT_SYMBOL(ipv6_chk_addr);
1231
1232 static
1233 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1234                        struct net_device *dev)
1235 {
1236         struct inet6_ifaddr * ifp;
1237         u8 hash = ipv6_addr_hash(addr);
1238
1239         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1240                 if (ifp->idev->dev->nd_net != net)
1241                         continue;
1242                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1243                         if (dev == NULL || ifp->idev->dev == dev)
1244                                 break;
1245                 }
1246         }
1247         return ifp != NULL;
1248 }
1249
1250 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1251 {
1252         struct inet6_ifaddr * ifp;
1253         u8 hash = ipv6_addr_hash(addr);
1254
1255         read_lock_bh(&addrconf_hash_lock);
1256         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1257                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1258                         if (dev == NULL || ifp->idev->dev == dev ||
1259                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1260                                 in6_ifa_hold(ifp);
1261                                 break;
1262                         }
1263                 }
1264         }
1265         read_unlock_bh(&addrconf_hash_lock);
1266
1267         return ifp;
1268 }
1269
1270 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1271 {
1272         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1273         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1274         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1275         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1276         int sk_ipv6only = ipv6_only_sock(sk);
1277         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1278         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1279         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1280
1281         if (!sk2_rcv_saddr && !sk_ipv6only)
1282                 return 1;
1283
1284         if (addr_type2 == IPV6_ADDR_ANY &&
1285             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1286                 return 1;
1287
1288         if (addr_type == IPV6_ADDR_ANY &&
1289             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1290                 return 1;
1291
1292         if (sk2_rcv_saddr6 &&
1293             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1294                 return 1;
1295
1296         if (addr_type == IPV6_ADDR_MAPPED &&
1297             !sk2_ipv6only &&
1298             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1299                 return 1;
1300
1301         return 0;
1302 }
1303
1304 /* Gets referenced address, destroys ifaddr */
1305
1306 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1307 {
1308         if (ifp->flags&IFA_F_PERMANENT) {
1309                 spin_lock_bh(&ifp->lock);
1310                 addrconf_del_timer(ifp);
1311                 ifp->flags |= IFA_F_TENTATIVE;
1312                 spin_unlock_bh(&ifp->lock);
1313                 in6_ifa_put(ifp);
1314 #ifdef CONFIG_IPV6_PRIVACY
1315         } else if (ifp->flags&IFA_F_TEMPORARY) {
1316                 struct inet6_ifaddr *ifpub;
1317                 spin_lock_bh(&ifp->lock);
1318                 ifpub = ifp->ifpub;
1319                 if (ifpub) {
1320                         in6_ifa_hold(ifpub);
1321                         spin_unlock_bh(&ifp->lock);
1322                         ipv6_create_tempaddr(ifpub, ifp);
1323                         in6_ifa_put(ifpub);
1324                 } else {
1325                         spin_unlock_bh(&ifp->lock);
1326                 }
1327                 ipv6_del_addr(ifp);
1328 #endif
1329         } else
1330                 ipv6_del_addr(ifp);
1331 }
1332
1333 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1334 {
1335         if (net_ratelimit())
1336                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1337         addrconf_dad_stop(ifp);
1338 }
1339
1340 /* Join to solicited addr multicast group. */
1341
1342 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1343 {
1344         struct in6_addr maddr;
1345
1346         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1347                 return;
1348
1349         addrconf_addr_solict_mult(addr, &maddr);
1350         ipv6_dev_mc_inc(dev, &maddr);
1351 }
1352
1353 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1354 {
1355         struct in6_addr maddr;
1356
1357         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1358                 return;
1359
1360         addrconf_addr_solict_mult(addr, &maddr);
1361         __ipv6_dev_mc_dec(idev, &maddr);
1362 }
1363
1364 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1365 {
1366         struct in6_addr addr;
1367         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1368         if (ipv6_addr_any(&addr))
1369                 return;
1370         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1371 }
1372
1373 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1374 {
1375         struct in6_addr addr;
1376         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1377         if (ipv6_addr_any(&addr))
1378                 return;
1379         __ipv6_dev_ac_dec(ifp->idev, &addr);
1380 }
1381
1382 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1383 {
1384         if (dev->addr_len != ETH_ALEN)
1385                 return -1;
1386         memcpy(eui, dev->dev_addr, 3);
1387         memcpy(eui + 5, dev->dev_addr + 3, 3);
1388
1389         /*
1390          * The zSeries OSA network cards can be shared among various
1391          * OS instances, but the OSA cards have only one MAC address.
1392          * This leads to duplicate address conflicts in conjunction
1393          * with IPv6 if more than one instance uses the same card.
1394          *
1395          * The driver for these cards can deliver a unique 16-bit
1396          * identifier for each instance sharing the same card.  It is
1397          * placed instead of 0xFFFE in the interface identifier.  The
1398          * "u" bit of the interface identifier is not inverted in this
1399          * case.  Hence the resulting interface identifier has local
1400          * scope according to RFC2373.
1401          */
1402         if (dev->dev_id) {
1403                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1404                 eui[4] = dev->dev_id & 0xFF;
1405         } else {
1406                 eui[3] = 0xFF;
1407                 eui[4] = 0xFE;
1408                 eui[0] ^= 2;
1409         }
1410         return 0;
1411 }
1412
1413 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1414 {
1415         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1416         if (dev->addr_len != ARCNET_ALEN)
1417                 return -1;
1418         memset(eui, 0, 7);
1419         eui[7] = *(u8*)dev->dev_addr;
1420         return 0;
1421 }
1422
1423 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1424 {
1425         if (dev->addr_len != INFINIBAND_ALEN)
1426                 return -1;
1427         memcpy(eui, dev->dev_addr + 12, 8);
1428         eui[0] |= 2;
1429         return 0;
1430 }
1431
1432 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1433 {
1434         switch (dev->type) {
1435         case ARPHRD_ETHER:
1436         case ARPHRD_FDDI:
1437         case ARPHRD_IEEE802_TR:
1438                 return addrconf_ifid_eui48(eui, dev);
1439         case ARPHRD_ARCNET:
1440                 return addrconf_ifid_arcnet(eui, dev);
1441         case ARPHRD_INFINIBAND:
1442                 return addrconf_ifid_infiniband(eui, dev);
1443         case ARPHRD_SIT:
1444                 if (dev->priv_flags & IFF_ISATAP)
1445                         return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1446         }
1447         return -1;
1448 }
1449
1450 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1451 {
1452         int err = -1;
1453         struct inet6_ifaddr *ifp;
1454
1455         read_lock_bh(&idev->lock);
1456         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1457                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1458                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1459                         err = 0;
1460                         break;
1461                 }
1462         }
1463         read_unlock_bh(&idev->lock);
1464         return err;
1465 }
1466
1467 #ifdef CONFIG_IPV6_PRIVACY
1468 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1469 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1470 {
1471 regen:
1472         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1473         idev->rndid[0] &= ~0x02;
1474
1475         /*
1476          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1477          * check if generated address is not inappropriate
1478          *
1479          *  - Reserved subnet anycast (RFC 2526)
1480          *      11111101 11....11 1xxxxxxx
1481          *  - ISATAP (RFC4214) 6.1
1482          *      00-00-5E-FE-xx-xx-xx-xx
1483          *  - value 0
1484          *  - XXX: already assigned to an address on the device
1485          */
1486         if (idev->rndid[0] == 0xfd &&
1487             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1488             (idev->rndid[7]&0x80))
1489                 goto regen;
1490         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1491                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1492                         goto regen;
1493                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1494                         goto regen;
1495         }
1496
1497         return 0;
1498 }
1499
1500 static void ipv6_regen_rndid(unsigned long data)
1501 {
1502         struct inet6_dev *idev = (struct inet6_dev *) data;
1503         unsigned long expires;
1504
1505         rcu_read_lock_bh();
1506         write_lock_bh(&idev->lock);
1507
1508         if (idev->dead)
1509                 goto out;
1510
1511         if (__ipv6_regen_rndid(idev) < 0)
1512                 goto out;
1513
1514         expires = jiffies +
1515                 idev->cnf.temp_prefered_lft * HZ -
1516                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1517         if (time_before(expires, jiffies)) {
1518                 printk(KERN_WARNING
1519                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1520                         idev->dev->name);
1521                 goto out;
1522         }
1523
1524         if (!mod_timer(&idev->regen_timer, expires))
1525                 in6_dev_hold(idev);
1526
1527 out:
1528         write_unlock_bh(&idev->lock);
1529         rcu_read_unlock_bh();
1530         in6_dev_put(idev);
1531 }
1532
1533 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1534         int ret = 0;
1535
1536         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1537                 ret = __ipv6_regen_rndid(idev);
1538         return ret;
1539 }
1540 #endif
1541
1542 /*
1543  *      Add prefix route.
1544  */
1545
1546 static void
1547 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1548                       unsigned long expires, u32 flags)
1549 {
1550         struct fib6_config cfg = {
1551                 .fc_table = RT6_TABLE_PREFIX,
1552                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1553                 .fc_ifindex = dev->ifindex,
1554                 .fc_expires = expires,
1555                 .fc_dst_len = plen,
1556                 .fc_flags = RTF_UP | flags,
1557         };
1558
1559         ipv6_addr_copy(&cfg.fc_dst, pfx);
1560
1561         /* Prevent useless cloning on PtP SIT.
1562            This thing is done here expecting that the whole
1563            class of non-broadcast devices need not cloning.
1564          */
1565 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1566         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1567                 cfg.fc_flags |= RTF_NONEXTHOP;
1568 #endif
1569
1570         ip6_route_add(&cfg);
1571 }
1572
1573 /* Create "default" multicast route to the interface */
1574
1575 static void addrconf_add_mroute(struct net_device *dev)
1576 {
1577         struct fib6_config cfg = {
1578                 .fc_table = RT6_TABLE_LOCAL,
1579                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1580                 .fc_ifindex = dev->ifindex,
1581                 .fc_dst_len = 8,
1582                 .fc_flags = RTF_UP,
1583         };
1584
1585         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1586
1587         ip6_route_add(&cfg);
1588 }
1589
1590 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1591 static void sit_route_add(struct net_device *dev)
1592 {
1593         struct fib6_config cfg = {
1594                 .fc_table = RT6_TABLE_MAIN,
1595                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1596                 .fc_ifindex = dev->ifindex,
1597                 .fc_dst_len = 96,
1598                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1599         };
1600
1601         /* prefix length - 96 bits "::d.d.d.d" */
1602         ip6_route_add(&cfg);
1603 }
1604 #endif
1605
1606 static void addrconf_add_lroute(struct net_device *dev)
1607 {
1608         struct in6_addr addr;
1609
1610         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1611         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1612 }
1613
1614 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1615 {
1616         struct inet6_dev *idev;
1617
1618         ASSERT_RTNL();
1619
1620         if ((idev = ipv6_find_idev(dev)) == NULL)
1621                 return NULL;
1622
1623         /* Add default multicast route */
1624         addrconf_add_mroute(dev);
1625
1626         /* Add link local route */
1627         addrconf_add_lroute(dev);
1628         return idev;
1629 }
1630
1631 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1632 {
1633         struct prefix_info *pinfo;
1634         __u32 valid_lft;
1635         __u32 prefered_lft;
1636         int addr_type;
1637         unsigned long rt_expires;
1638         struct inet6_dev *in6_dev;
1639
1640         pinfo = (struct prefix_info *) opt;
1641
1642         if (len < sizeof(struct prefix_info)) {
1643                 ADBG(("addrconf: prefix option too short\n"));
1644                 return;
1645         }
1646
1647         /*
1648          *      Validation checks ([ADDRCONF], page 19)
1649          */
1650
1651         addr_type = ipv6_addr_type(&pinfo->prefix);
1652
1653         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1654                 return;
1655
1656         valid_lft = ntohl(pinfo->valid);
1657         prefered_lft = ntohl(pinfo->prefered);
1658
1659         if (prefered_lft > valid_lft) {
1660                 if (net_ratelimit())
1661                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1662                 return;
1663         }
1664
1665         in6_dev = in6_dev_get(dev);
1666
1667         if (in6_dev == NULL) {
1668                 if (net_ratelimit())
1669                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1670                 return;
1671         }
1672
1673         /*
1674          *      Two things going on here:
1675          *      1) Add routes for on-link prefixes
1676          *      2) Configure prefixes with the auto flag set
1677          */
1678
1679         /* Avoid arithmetic overflow. Really, we could
1680            save rt_expires in seconds, likely valid_lft,
1681            but it would require division in fib gc, that it
1682            not good.
1683          */
1684         if (valid_lft >= 0x7FFFFFFF/HZ)
1685                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1686         else
1687                 rt_expires = valid_lft * HZ;
1688
1689         /*
1690          * We convert this (in jiffies) to clock_t later.
1691          * Avoid arithmetic overflow there as well.
1692          * Overflow can happen only if HZ < USER_HZ.
1693          */
1694         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1695                 rt_expires = 0x7FFFFFFF / USER_HZ;
1696
1697         if (pinfo->onlink) {
1698                 struct rt6_info *rt;
1699                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1700
1701                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1702                         if (rt->rt6i_flags&RTF_EXPIRES) {
1703                                 if (valid_lft == 0) {
1704                                         ip6_del_rt(rt);
1705                                         rt = NULL;
1706                                 } else {
1707                                         rt->rt6i_expires = jiffies + rt_expires;
1708                                 }
1709                         }
1710                 } else if (valid_lft) {
1711                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1712                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1713                 }
1714                 if (rt)
1715                         dst_release(&rt->u.dst);
1716         }
1717
1718         /* Try to figure out our local address for this prefix */
1719
1720         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1721                 struct inet6_ifaddr * ifp;
1722                 struct in6_addr addr;
1723                 int create = 0, update_lft = 0;
1724
1725                 if (pinfo->prefix_len == 64) {
1726                         memcpy(&addr, &pinfo->prefix, 8);
1727                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1728                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1729                                 in6_dev_put(in6_dev);
1730                                 return;
1731                         }
1732                         goto ok;
1733                 }
1734                 if (net_ratelimit())
1735                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1736                                pinfo->prefix_len);
1737                 in6_dev_put(in6_dev);
1738                 return;
1739
1740 ok:
1741
1742                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1743
1744                 if (ifp == NULL && valid_lft) {
1745                         int max_addresses = in6_dev->cnf.max_addresses;
1746                         u32 addr_flags = 0;
1747
1748 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1749                         if (in6_dev->cnf.optimistic_dad &&
1750                             !ipv6_devconf.forwarding)
1751                                 addr_flags = IFA_F_OPTIMISTIC;
1752 #endif
1753
1754                         /* Do not allow to create too much of autoconfigured
1755                          * addresses; this would be too easy way to crash kernel.
1756                          */
1757                         if (!max_addresses ||
1758                             ipv6_count_addresses(in6_dev) < max_addresses)
1759                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1760                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1761                                                     addr_flags);
1762
1763                         if (!ifp || IS_ERR(ifp)) {
1764                                 in6_dev_put(in6_dev);
1765                                 return;
1766                         }
1767
1768                         update_lft = create = 1;
1769                         ifp->cstamp = jiffies;
1770                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1771                 }
1772
1773                 if (ifp) {
1774                         int flags;
1775                         unsigned long now;
1776 #ifdef CONFIG_IPV6_PRIVACY
1777                         struct inet6_ifaddr *ift;
1778 #endif
1779                         u32 stored_lft;
1780
1781                         /* update lifetime (RFC2462 5.5.3 e) */
1782                         spin_lock(&ifp->lock);
1783                         now = jiffies;
1784                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1785                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1786                         else
1787                                 stored_lft = 0;
1788                         if (!update_lft && stored_lft) {
1789                                 if (valid_lft > MIN_VALID_LIFETIME ||
1790                                     valid_lft > stored_lft)
1791                                         update_lft = 1;
1792                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1793                                         /* valid_lft <= stored_lft is always true */
1794                                         /* XXX: IPsec */
1795                                         update_lft = 0;
1796                                 } else {
1797                                         valid_lft = MIN_VALID_LIFETIME;
1798                                         if (valid_lft < prefered_lft)
1799                                                 prefered_lft = valid_lft;
1800                                         update_lft = 1;
1801                                 }
1802                         }
1803
1804                         if (update_lft) {
1805                                 ifp->valid_lft = valid_lft;
1806                                 ifp->prefered_lft = prefered_lft;
1807                                 ifp->tstamp = now;
1808                                 flags = ifp->flags;
1809                                 ifp->flags &= ~IFA_F_DEPRECATED;
1810                                 spin_unlock(&ifp->lock);
1811
1812                                 if (!(flags&IFA_F_TENTATIVE))
1813                                         ipv6_ifa_notify(0, ifp);
1814                         } else
1815                                 spin_unlock(&ifp->lock);
1816
1817 #ifdef CONFIG_IPV6_PRIVACY
1818                         read_lock_bh(&in6_dev->lock);
1819                         /* update all temporary addresses in the list */
1820                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1821                                 /*
1822                                  * When adjusting the lifetimes of an existing
1823                                  * temporary address, only lower the lifetimes.
1824                                  * Implementations must not increase the
1825                                  * lifetimes of an existing temporary address
1826                                  * when processing a Prefix Information Option.
1827                                  */
1828                                 spin_lock(&ift->lock);
1829                                 flags = ift->flags;
1830                                 if (ift->valid_lft > valid_lft &&
1831                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1832                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1833                                 if (ift->prefered_lft > prefered_lft &&
1834                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1835                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1836                                 spin_unlock(&ift->lock);
1837                                 if (!(flags&IFA_F_TENTATIVE))
1838                                         ipv6_ifa_notify(0, ift);
1839                         }
1840
1841                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1842                                 /*
1843                                  * When a new public address is created as described in [ADDRCONF],
1844                                  * also create a new temporary address.
1845                                  */
1846                                 read_unlock_bh(&in6_dev->lock);
1847                                 ipv6_create_tempaddr(ifp, NULL);
1848                         } else {
1849                                 read_unlock_bh(&in6_dev->lock);
1850                         }
1851 #endif
1852                         in6_ifa_put(ifp);
1853                         addrconf_verify(0);
1854                 }
1855         }
1856         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1857         in6_dev_put(in6_dev);
1858 }
1859
1860 /*
1861  *      Set destination address.
1862  *      Special case for SIT interfaces where we create a new "virtual"
1863  *      device.
1864  */
1865 int addrconf_set_dstaddr(void __user *arg)
1866 {
1867         struct in6_ifreq ireq;
1868         struct net_device *dev;
1869         int err = -EINVAL;
1870
1871         rtnl_lock();
1872
1873         err = -EFAULT;
1874         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1875                 goto err_exit;
1876
1877         dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1878
1879         err = -ENODEV;
1880         if (dev == NULL)
1881                 goto err_exit;
1882
1883 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1884         if (dev->type == ARPHRD_SIT) {
1885                 struct ifreq ifr;
1886                 mm_segment_t    oldfs;
1887                 struct ip_tunnel_parm p;
1888
1889                 err = -EADDRNOTAVAIL;
1890                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1891                         goto err_exit;
1892
1893                 memset(&p, 0, sizeof(p));
1894                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1895                 p.iph.saddr = 0;
1896                 p.iph.version = 4;
1897                 p.iph.ihl = 5;
1898                 p.iph.protocol = IPPROTO_IPV6;
1899                 p.iph.ttl = 64;
1900                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1901
1902                 oldfs = get_fs(); set_fs(KERNEL_DS);
1903                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1904                 set_fs(oldfs);
1905
1906                 if (err == 0) {
1907                         err = -ENOBUFS;
1908                         if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1909                                 goto err_exit;
1910                         err = dev_open(dev);
1911                 }
1912         }
1913 #endif
1914
1915 err_exit:
1916         rtnl_unlock();
1917         return err;
1918 }
1919
1920 /*
1921  *      Manual configuration of address on an interface
1922  */
1923 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1924                           __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1925 {
1926         struct inet6_ifaddr *ifp;
1927         struct inet6_dev *idev;
1928         struct net_device *dev;
1929         int scope;
1930         u32 flags = RTF_EXPIRES;
1931
1932         ASSERT_RTNL();
1933
1934         /* check the lifetime */
1935         if (!valid_lft || prefered_lft > valid_lft)
1936                 return -EINVAL;
1937
1938         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1939                 return -ENODEV;
1940
1941         if ((idev = addrconf_add_dev(dev)) == NULL)
1942                 return -ENOBUFS;
1943
1944         scope = ipv6_addr_scope(pfx);
1945
1946         if (valid_lft == INFINITY_LIFE_TIME) {
1947                 ifa_flags |= IFA_F_PERMANENT;
1948                 flags = 0;
1949         } else if (valid_lft >= 0x7FFFFFFF/HZ)
1950                 valid_lft = 0x7FFFFFFF/HZ;
1951
1952         if (prefered_lft == 0)
1953                 ifa_flags |= IFA_F_DEPRECATED;
1954         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1955                  (prefered_lft != INFINITY_LIFE_TIME))
1956                 prefered_lft = 0x7FFFFFFF/HZ;
1957
1958         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1959
1960         if (!IS_ERR(ifp)) {
1961                 spin_lock_bh(&ifp->lock);
1962                 ifp->valid_lft = valid_lft;
1963                 ifp->prefered_lft = prefered_lft;
1964                 ifp->tstamp = jiffies;
1965                 spin_unlock_bh(&ifp->lock);
1966
1967                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1968                                       jiffies_to_clock_t(valid_lft * HZ), flags);
1969                 /*
1970                  * Note that section 3.1 of RFC 4429 indicates
1971                  * that the Optimistic flag should not be set for
1972                  * manually configured addresses
1973                  */
1974                 addrconf_dad_start(ifp, 0);
1975                 in6_ifa_put(ifp);
1976                 addrconf_verify(0);
1977                 return 0;
1978         }
1979
1980         return PTR_ERR(ifp);
1981 }
1982
1983 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1984 {
1985         struct inet6_ifaddr *ifp;
1986         struct inet6_dev *idev;
1987         struct net_device *dev;
1988
1989         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1990                 return -ENODEV;
1991
1992         if ((idev = __in6_dev_get(dev)) == NULL)
1993                 return -ENXIO;
1994
1995         read_lock_bh(&idev->lock);
1996         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1997                 if (ifp->prefix_len == plen &&
1998                     ipv6_addr_equal(pfx, &ifp->addr)) {
1999                         in6_ifa_hold(ifp);
2000                         read_unlock_bh(&idev->lock);
2001
2002                         ipv6_del_addr(ifp);
2003
2004                         /* If the last address is deleted administratively,
2005                            disable IPv6 on this interface.
2006                          */
2007                         if (idev->addr_list == NULL)
2008                                 addrconf_ifdown(idev->dev, 1);
2009                         return 0;
2010                 }
2011         }
2012         read_unlock_bh(&idev->lock);
2013         return -EADDRNOTAVAIL;
2014 }
2015
2016
2017 int addrconf_add_ifaddr(void __user *arg)
2018 {
2019         struct in6_ifreq ireq;
2020         int err;
2021
2022         if (!capable(CAP_NET_ADMIN))
2023                 return -EPERM;
2024
2025         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2026                 return -EFAULT;
2027
2028         rtnl_lock();
2029         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2030                              IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2031         rtnl_unlock();
2032         return err;
2033 }
2034
2035 int addrconf_del_ifaddr(void __user *arg)
2036 {
2037         struct in6_ifreq ireq;
2038         int err;
2039
2040         if (!capable(CAP_NET_ADMIN))
2041                 return -EPERM;
2042
2043         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2044                 return -EFAULT;
2045
2046         rtnl_lock();
2047         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2048         rtnl_unlock();
2049         return err;
2050 }
2051
2052 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2053 static void sit_add_v4_addrs(struct inet6_dev *idev)
2054 {
2055         struct inet6_ifaddr * ifp;
2056         struct in6_addr addr;
2057         struct net_device *dev;
2058         int scope;
2059
2060         ASSERT_RTNL();
2061
2062         memset(&addr, 0, sizeof(struct in6_addr));
2063         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2064
2065         if (idev->dev->flags&IFF_POINTOPOINT) {
2066                 addr.s6_addr32[0] = htonl(0xfe800000);
2067                 scope = IFA_LINK;
2068         } else {
2069                 scope = IPV6_ADDR_COMPATv4;
2070         }
2071
2072         if (addr.s6_addr32[3]) {
2073                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2074                 if (!IS_ERR(ifp)) {
2075                         spin_lock_bh(&ifp->lock);
2076                         ifp->flags &= ~IFA_F_TENTATIVE;
2077                         spin_unlock_bh(&ifp->lock);
2078                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2079                         in6_ifa_put(ifp);
2080                 }
2081                 return;
2082         }
2083
2084         for_each_netdev(&init_net, dev) {
2085                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2086                 if (in_dev && (dev->flags & IFF_UP)) {
2087                         struct in_ifaddr * ifa;
2088
2089                         int flag = scope;
2090
2091                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2092                                 int plen;
2093
2094                                 addr.s6_addr32[3] = ifa->ifa_local;
2095
2096                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2097                                         continue;
2098                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2099                                         if (idev->dev->flags&IFF_POINTOPOINT)
2100                                                 continue;
2101                                         flag |= IFA_HOST;
2102                                 }
2103                                 if (idev->dev->flags&IFF_POINTOPOINT)
2104                                         plen = 64;
2105                                 else
2106                                         plen = 96;
2107
2108                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2109                                                     IFA_F_PERMANENT);
2110                                 if (!IS_ERR(ifp)) {
2111                                         spin_lock_bh(&ifp->lock);
2112                                         ifp->flags &= ~IFA_F_TENTATIVE;
2113                                         spin_unlock_bh(&ifp->lock);
2114                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2115                                         in6_ifa_put(ifp);
2116                                 }
2117                         }
2118                 }
2119         }
2120 }
2121 #endif
2122
2123 static void init_loopback(struct net_device *dev)
2124 {
2125         struct inet6_dev  *idev;
2126         struct inet6_ifaddr * ifp;
2127
2128         /* ::1 */
2129
2130         ASSERT_RTNL();
2131
2132         if ((idev = ipv6_find_idev(dev)) == NULL) {
2133                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2134                 return;
2135         }
2136
2137         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2138         if (!IS_ERR(ifp)) {
2139                 spin_lock_bh(&ifp->lock);
2140                 ifp->flags &= ~IFA_F_TENTATIVE;
2141                 spin_unlock_bh(&ifp->lock);
2142                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2143                 in6_ifa_put(ifp);
2144         }
2145 }
2146
2147 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2148 {
2149         struct inet6_ifaddr * ifp;
2150         u32 addr_flags = IFA_F_PERMANENT;
2151
2152 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2153         if (idev->cnf.optimistic_dad &&
2154             !ipv6_devconf.forwarding)
2155                 addr_flags |= IFA_F_OPTIMISTIC;
2156 #endif
2157
2158
2159         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2160         if (!IS_ERR(ifp)) {
2161                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2162                 addrconf_dad_start(ifp, 0);
2163                 in6_ifa_put(ifp);
2164         }
2165 }
2166
2167 static void addrconf_dev_config(struct net_device *dev)
2168 {
2169         struct in6_addr addr;
2170         struct inet6_dev    * idev;
2171
2172         ASSERT_RTNL();
2173
2174         if ((dev->type != ARPHRD_ETHER) &&
2175             (dev->type != ARPHRD_FDDI) &&
2176             (dev->type != ARPHRD_IEEE802_TR) &&
2177             (dev->type != ARPHRD_ARCNET) &&
2178             (dev->type != ARPHRD_INFINIBAND)) {
2179                 /* Alas, we support only Ethernet autoconfiguration. */
2180                 return;
2181         }
2182
2183         idev = addrconf_add_dev(dev);
2184         if (idev == NULL)
2185                 return;
2186
2187         memset(&addr, 0, sizeof(struct in6_addr));
2188         addr.s6_addr32[0] = htonl(0xFE800000);
2189
2190         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2191                 addrconf_add_linklocal(idev, &addr);
2192 }
2193
2194 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2195 static void addrconf_sit_config(struct net_device *dev)
2196 {
2197         struct inet6_dev *idev;
2198
2199         ASSERT_RTNL();
2200
2201         /*
2202          * Configure the tunnel with one of our IPv4
2203          * addresses... we should configure all of
2204          * our v4 addrs in the tunnel
2205          */
2206
2207         if ((idev = ipv6_find_idev(dev)) == NULL) {
2208                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2209                 return;
2210         }
2211
2212         if (dev->priv_flags & IFF_ISATAP) {
2213                 struct in6_addr addr;
2214
2215                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2216                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2217                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2218                         addrconf_add_linklocal(idev, &addr);
2219                 return;
2220         }
2221
2222         sit_add_v4_addrs(idev);
2223
2224         if (dev->flags&IFF_POINTOPOINT) {
2225                 addrconf_add_mroute(dev);
2226                 addrconf_add_lroute(dev);
2227         } else
2228                 sit_route_add(dev);
2229 }
2230 #endif
2231
2232 static inline int
2233 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2234 {
2235         struct in6_addr lladdr;
2236
2237         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2238                 addrconf_add_linklocal(idev, &lladdr);
2239                 return 0;
2240         }
2241         return -1;
2242 }
2243
2244 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2245 {
2246         struct net_device *link_dev;
2247
2248         /* first try to inherit the link-local address from the link device */
2249         if (idev->dev->iflink &&
2250             (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2251                 if (!ipv6_inherit_linklocal(idev, link_dev))
2252                         return;
2253         }
2254         /* then try to inherit it from any device */
2255         for_each_netdev(&init_net, link_dev) {
2256                 if (!ipv6_inherit_linklocal(idev, link_dev))
2257                         return;
2258         }
2259         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2260 }
2261
2262 /*
2263  * Autoconfigure tunnel with a link-local address so routing protocols,
2264  * DHCPv6, MLD etc. can be run over the virtual link
2265  */
2266
2267 static void addrconf_ip6_tnl_config(struct net_device *dev)
2268 {
2269         struct inet6_dev *idev;
2270
2271         ASSERT_RTNL();
2272
2273         if ((idev = addrconf_add_dev(dev)) == NULL) {
2274                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2275                 return;
2276         }
2277         ip6_tnl_add_linklocal(idev);
2278 }
2279
2280 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2281                            void * data)
2282 {
2283         struct net_device *dev = (struct net_device *) data;
2284         struct inet6_dev *idev = __in6_dev_get(dev);
2285         int run_pending = 0;
2286         int err;
2287
2288         if (dev->nd_net != &init_net)
2289                 return NOTIFY_DONE;
2290
2291         switch(event) {
2292         case NETDEV_REGISTER:
2293                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2294                         idev = ipv6_add_dev(dev);
2295                         if (!idev)
2296                                 return notifier_from_errno(-ENOMEM);
2297                 }
2298                 break;
2299         case NETDEV_UP:
2300         case NETDEV_CHANGE:
2301                 if (dev->flags & IFF_SLAVE)
2302                         break;
2303
2304                 if (event == NETDEV_UP) {
2305                         if (!addrconf_qdisc_ok(dev)) {
2306                                 /* device is not ready yet. */
2307                                 printk(KERN_INFO
2308                                         "ADDRCONF(NETDEV_UP): %s: "
2309                                         "link is not ready\n",
2310                                         dev->name);
2311                                 break;
2312                         }
2313
2314                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2315                                 idev = ipv6_add_dev(dev);
2316
2317                         if (idev)
2318                                 idev->if_flags |= IF_READY;
2319                 } else {
2320                         if (!addrconf_qdisc_ok(dev)) {
2321                                 /* device is still not ready. */
2322                                 break;
2323                         }
2324
2325                         if (idev) {
2326                                 if (idev->if_flags & IF_READY) {
2327                                         /* device is already configured. */
2328                                         break;
2329                                 }
2330                                 idev->if_flags |= IF_READY;
2331                         }
2332
2333                         printk(KERN_INFO
2334                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2335                                         "link becomes ready\n",
2336                                         dev->name);
2337
2338                         run_pending = 1;
2339                 }
2340
2341                 switch(dev->type) {
2342 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2343                 case ARPHRD_SIT:
2344                         addrconf_sit_config(dev);
2345                         break;
2346 #endif
2347                 case ARPHRD_TUNNEL6:
2348                         addrconf_ip6_tnl_config(dev);
2349                         break;
2350                 case ARPHRD_LOOPBACK:
2351                         init_loopback(dev);
2352                         break;
2353
2354                 default:
2355                         addrconf_dev_config(dev);
2356                         break;
2357                 }
2358                 if (idev) {
2359                         if (run_pending)
2360                                 addrconf_dad_run(idev);
2361
2362                         /* If the MTU changed during the interface down, when the
2363                            interface up, the changed MTU must be reflected in the
2364                            idev as well as routers.
2365                          */
2366                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2367                                 rt6_mtu_change(dev, dev->mtu);
2368                                 idev->cnf.mtu6 = dev->mtu;
2369                         }
2370                         idev->tstamp = jiffies;
2371                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2372                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2373                            stop IPv6 on this interface.
2374                          */
2375                         if (dev->mtu < IPV6_MIN_MTU)
2376                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2377                 }
2378                 break;
2379
2380         case NETDEV_CHANGEMTU:
2381                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2382                         rt6_mtu_change(dev, dev->mtu);
2383                         idev->cnf.mtu6 = dev->mtu;
2384                         break;
2385                 }
2386
2387                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2388                         idev = ipv6_add_dev(dev);
2389                         if (idev)
2390                                 break;
2391                 }
2392
2393                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2394
2395         case NETDEV_DOWN:
2396         case NETDEV_UNREGISTER:
2397                 /*
2398                  *      Remove all addresses from this interface.
2399                  */
2400                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2401                 break;
2402
2403         case NETDEV_CHANGENAME:
2404                 if (idev) {
2405                         snmp6_unregister_dev(idev);
2406                         addrconf_sysctl_unregister(idev);
2407                         addrconf_sysctl_register(idev);
2408                         err = snmp6_register_dev(idev);
2409                         if (err)
2410                                 return notifier_from_errno(err);
2411                 }
2412                 break;
2413         }
2414
2415         return NOTIFY_OK;
2416 }
2417
2418 /*
2419  *      addrconf module should be notified of a device going up
2420  */
2421 static struct notifier_block ipv6_dev_notf = {
2422         .notifier_call = addrconf_notify,
2423         .priority = 0
2424 };
2425
2426 static int addrconf_ifdown(struct net_device *dev, int how)
2427 {
2428         struct inet6_dev *idev;
2429         struct inet6_ifaddr *ifa, **bifa;
2430         int i;
2431
2432         ASSERT_RTNL();
2433
2434         if (dev == init_net.loopback_dev && how == 1)
2435                 how = 0;
2436
2437         rt6_ifdown(dev);
2438         neigh_ifdown(&nd_tbl, dev);
2439
2440         idev = __in6_dev_get(dev);
2441         if (idev == NULL)
2442                 return -ENODEV;
2443
2444         /* Step 1: remove reference to ipv6 device from parent device.
2445                    Do not dev_put!
2446          */
2447         if (how == 1) {
2448                 idev->dead = 1;
2449
2450                 /* protected by rtnl_lock */
2451                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2452
2453                 /* Step 1.5: remove snmp6 entry */
2454                 snmp6_unregister_dev(idev);
2455
2456         }
2457
2458         /* Step 2: clear hash table */
2459         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2460                 bifa = &inet6_addr_lst[i];
2461
2462                 write_lock_bh(&addrconf_hash_lock);
2463                 while ((ifa = *bifa) != NULL) {
2464                         if (ifa->idev == idev) {
2465                                 *bifa = ifa->lst_next;
2466                                 ifa->lst_next = NULL;
2467                                 addrconf_del_timer(ifa);
2468                                 in6_ifa_put(ifa);
2469                                 continue;
2470                         }
2471                         bifa = &ifa->lst_next;
2472                 }
2473                 write_unlock_bh(&addrconf_hash_lock);
2474         }
2475
2476         write_lock_bh(&idev->lock);
2477
2478         /* Step 3: clear flags for stateless addrconf */
2479         if (how != 1)
2480                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2481
2482         /* Step 4: clear address list */
2483 #ifdef CONFIG_IPV6_PRIVACY
2484         if (how == 1 && del_timer(&idev->regen_timer))
2485                 in6_dev_put(idev);
2486
2487         /* clear tempaddr list */
2488         while ((ifa = idev->tempaddr_list) != NULL) {
2489                 idev->tempaddr_list = ifa->tmp_next;
2490                 ifa->tmp_next = NULL;
2491                 ifa->dead = 1;
2492                 write_unlock_bh(&idev->lock);
2493                 spin_lock_bh(&ifa->lock);
2494
2495                 if (ifa->ifpub) {
2496                         in6_ifa_put(ifa->ifpub);
2497                         ifa->ifpub = NULL;
2498                 }
2499                 spin_unlock_bh(&ifa->lock);
2500                 in6_ifa_put(ifa);
2501                 write_lock_bh(&idev->lock);
2502         }
2503 #endif
2504         while ((ifa = idev->addr_list) != NULL) {
2505                 idev->addr_list = ifa->if_next;
2506                 ifa->if_next = NULL;
2507                 ifa->dead = 1;
2508                 addrconf_del_timer(ifa);
2509                 write_unlock_bh(&idev->lock);
2510
2511                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2512                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2513                 in6_ifa_put(ifa);
2514
2515                 write_lock_bh(&idev->lock);
2516         }
2517         write_unlock_bh(&idev->lock);
2518
2519         /* Step 5: Discard multicast list */
2520
2521         if (how == 1)
2522                 ipv6_mc_destroy_dev(idev);
2523         else
2524                 ipv6_mc_down(idev);
2525
2526         idev->tstamp = jiffies;
2527
2528         /* Shot the device (if unregistered) */
2529
2530         if (how == 1) {
2531                 addrconf_sysctl_unregister(idev);
2532                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2533                 neigh_ifdown(&nd_tbl, dev);
2534                 in6_dev_put(idev);
2535         }
2536         return 0;
2537 }
2538
2539 static void addrconf_rs_timer(unsigned long data)
2540 {
2541         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2542
2543         if (ifp->idev->cnf.forwarding)
2544                 goto out;
2545
2546         if (ifp->idev->if_flags & IF_RA_RCVD) {
2547                 /*
2548                  *      Announcement received after solicitation
2549                  *      was sent
2550                  */
2551                 goto out;
2552         }
2553
2554         spin_lock(&ifp->lock);
2555         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2556                 struct in6_addr all_routers;
2557
2558                 /* The wait after the last probe can be shorter */
2559                 addrconf_mod_timer(ifp, AC_RS,
2560                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2561                                    ifp->idev->cnf.rtr_solicit_delay :
2562                                    ifp->idev->cnf.rtr_solicit_interval);
2563                 spin_unlock(&ifp->lock);
2564
2565                 ipv6_addr_all_routers(&all_routers);
2566
2567                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2568         } else {
2569                 spin_unlock(&ifp->lock);
2570                 /*
2571                  * Note: we do not support deprecated "all on-link"
2572                  * assumption any longer.
2573                  */
2574                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2575                        ifp->idev->dev->name);
2576         }
2577
2578 out:
2579         in6_ifa_put(ifp);
2580 }
2581
2582 /*
2583  *      Duplicate Address Detection
2584  */
2585 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2586 {
2587         unsigned long rand_num;
2588         struct inet6_dev *idev = ifp->idev;
2589
2590         if (ifp->flags & IFA_F_OPTIMISTIC)
2591                 rand_num = 0;
2592         else
2593                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2594
2595         ifp->probes = idev->cnf.dad_transmits;
2596         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2597 }
2598
2599 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2600 {
2601         struct inet6_dev *idev = ifp->idev;
2602         struct net_device *dev = idev->dev;
2603
2604         addrconf_join_solict(dev, &ifp->addr);
2605
2606         net_srandom(ifp->addr.s6_addr32[3]);
2607
2608         read_lock_bh(&idev->lock);
2609         if (ifp->dead)
2610                 goto out;
2611         spin_lock_bh(&ifp->lock);
2612
2613         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2614             !(ifp->flags&IFA_F_TENTATIVE) ||
2615             ifp->flags & IFA_F_NODAD) {
2616                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2617                 spin_unlock_bh(&ifp->lock);
2618                 read_unlock_bh(&idev->lock);
2619
2620                 addrconf_dad_completed(ifp);
2621                 return;
2622         }
2623
2624         if (!(idev->if_flags & IF_READY)) {
2625                 spin_unlock_bh(&ifp->lock);
2626                 read_unlock_bh(&idev->lock);
2627                 /*
2628                  * If the defice is not ready:
2629                  * - keep it tentative if it is a permanent address.
2630                  * - otherwise, kill it.
2631                  */
2632                 in6_ifa_hold(ifp);
2633                 addrconf_dad_stop(ifp);
2634                 return;
2635         }
2636
2637         /*
2638          * Optimistic nodes can start receiving
2639          * Frames right away
2640          */
2641         if(ifp->flags & IFA_F_OPTIMISTIC)
2642                 ip6_ins_rt(ifp->rt);
2643
2644         addrconf_dad_kick(ifp);
2645         spin_unlock_bh(&ifp->lock);
2646 out:
2647         read_unlock_bh(&idev->lock);
2648 }
2649
2650 static void addrconf_dad_timer(unsigned long data)
2651 {
2652         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2653         struct inet6_dev *idev = ifp->idev;
2654         struct in6_addr unspec;
2655         struct in6_addr mcaddr;
2656
2657         read_lock_bh(&idev->lock);
2658         if (idev->dead) {
2659                 read_unlock_bh(&idev->lock);
2660                 goto out;
2661         }
2662         spin_lock_bh(&ifp->lock);
2663         if (ifp->probes == 0) {
2664                 /*
2665                  * DAD was successful
2666                  */
2667
2668                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2669                 spin_unlock_bh(&ifp->lock);
2670                 read_unlock_bh(&idev->lock);
2671
2672                 addrconf_dad_completed(ifp);
2673
2674                 goto out;
2675         }
2676
2677         ifp->probes--;
2678         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2679         spin_unlock_bh(&ifp->lock);
2680         read_unlock_bh(&idev->lock);
2681
2682         /* send a neighbour solicitation for our addr */
2683         memset(&unspec, 0, sizeof(unspec));
2684         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2685         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2686 out:
2687         in6_ifa_put(ifp);
2688 }
2689
2690 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2691 {
2692         struct net_device *     dev = ifp->idev->dev;
2693
2694         /*
2695          *      Configure the address for reception. Now it is valid.
2696          */
2697
2698         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2699
2700         /* If added prefix is link local and forwarding is off,
2701            start sending router solicitations.
2702          */
2703
2704         if (ifp->idev->cnf.forwarding == 0 &&
2705             ifp->idev->cnf.rtr_solicits > 0 &&
2706             (dev->flags&IFF_LOOPBACK) == 0 &&
2707             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2708                 struct in6_addr all_routers;
2709
2710                 ipv6_addr_all_routers(&all_routers);
2711
2712                 /*
2713                  *      If a host as already performed a random delay
2714                  *      [...] as part of DAD [...] there is no need
2715                  *      to delay again before sending the first RS
2716                  */
2717                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2718
2719                 spin_lock_bh(&ifp->lock);
2720                 ifp->probes = 1;
2721                 ifp->idev->if_flags |= IF_RS_SENT;
2722                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2723                 spin_unlock_bh(&ifp->lock);
2724         }
2725 }
2726
2727 static void addrconf_dad_run(struct inet6_dev *idev) {
2728         struct inet6_ifaddr *ifp;
2729
2730         read_lock_bh(&idev->lock);
2731         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2732                 spin_lock_bh(&ifp->lock);
2733                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2734                         spin_unlock_bh(&ifp->lock);
2735                         continue;
2736                 }
2737                 spin_unlock_bh(&ifp->lock);
2738                 addrconf_dad_kick(ifp);
2739         }
2740         read_unlock_bh(&idev->lock);
2741 }
2742
2743 #ifdef CONFIG_PROC_FS
2744 struct if6_iter_state {
2745         struct seq_net_private p;
2746         int bucket;
2747 };
2748
2749 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2750 {
2751         struct inet6_ifaddr *ifa = NULL;
2752         struct if6_iter_state *state = seq->private;
2753         struct net *net = state->p.net;
2754
2755         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2756                 ifa = inet6_addr_lst[state->bucket];
2757
2758                 while (ifa && ifa->idev->dev->nd_net != net)
2759                         ifa = ifa->lst_next;
2760                 if (ifa)
2761                         break;
2762         }
2763         return ifa;
2764 }
2765
2766 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2767 {
2768         struct if6_iter_state *state = seq->private;
2769         struct net *net = state->p.net;
2770
2771         ifa = ifa->lst_next;
2772 try_again:
2773         if (ifa) {
2774                 if (ifa->idev->dev->nd_net != net) {
2775                         ifa = ifa->lst_next;
2776                         goto try_again;
2777                 }
2778         }
2779
2780         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2781                 ifa = inet6_addr_lst[state->bucket];
2782                 goto try_again;
2783         }
2784
2785         return ifa;
2786 }
2787
2788 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2789 {
2790         struct inet6_ifaddr *ifa = if6_get_first(seq);
2791
2792         if (ifa)
2793                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2794                         --pos;
2795         return pos ? NULL : ifa;
2796 }
2797
2798 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2799 {
2800         read_lock_bh(&addrconf_hash_lock);
2801         return if6_get_idx(seq, *pos);
2802 }
2803
2804 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2805 {
2806         struct inet6_ifaddr *ifa;
2807
2808         ifa = if6_get_next(seq, v);
2809         ++*pos;
2810         return ifa;
2811 }
2812
2813 static void if6_seq_stop(struct seq_file *seq, void *v)
2814 {
2815         read_unlock_bh(&addrconf_hash_lock);
2816 }
2817
2818 static int if6_seq_show(struct seq_file *seq, void *v)
2819 {
2820         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2821         seq_printf(seq,
2822                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2823                    NIP6(ifp->addr),
2824                    ifp->idev->dev->ifindex,
2825                    ifp->prefix_len,
2826                    ifp->scope,
2827                    ifp->flags,
2828                    ifp->idev->dev->name);
2829         return 0;
2830 }
2831
2832 static const struct seq_operations if6_seq_ops = {
2833         .start  = if6_seq_start,
2834         .next   = if6_seq_next,
2835         .show   = if6_seq_show,
2836         .stop   = if6_seq_stop,
2837 };
2838
2839 static int if6_seq_open(struct inode *inode, struct file *file)
2840 {
2841         return seq_open_net(inode, file, &if6_seq_ops,
2842                             sizeof(struct if6_iter_state));
2843 }
2844
2845 static const struct file_operations if6_fops = {
2846         .owner          = THIS_MODULE,
2847         .open           = if6_seq_open,
2848         .read           = seq_read,
2849         .llseek         = seq_lseek,
2850         .release        = seq_release_net,
2851 };
2852
2853 static int if6_proc_net_init(struct net *net)
2854 {
2855         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2856                 return -ENOMEM;
2857         return 0;
2858 }
2859
2860 static void if6_proc_net_exit(struct net *net)
2861 {
2862        proc_net_remove(net, "if_inet6");
2863 }
2864
2865 static struct pernet_operations if6_proc_net_ops = {
2866        .init = if6_proc_net_init,
2867        .exit = if6_proc_net_exit,
2868 };
2869
2870 int __init if6_proc_init(void)
2871 {
2872         return register_pernet_subsys(&if6_proc_net_ops);
2873 }
2874
2875 void if6_proc_exit(void)
2876 {
2877         unregister_pernet_subsys(&if6_proc_net_ops);
2878 }
2879 #endif  /* CONFIG_PROC_FS */
2880
2881 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2882 /* Check if address is a home address configured on any interface. */
2883 int ipv6_chk_home_addr(struct in6_addr *addr)
2884 {
2885         int ret = 0;
2886         struct inet6_ifaddr * ifp;
2887         u8 hash = ipv6_addr_hash(addr);
2888         read_lock_bh(&addrconf_hash_lock);
2889         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2890                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2891                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2892                         ret = 1;
2893                         break;
2894                 }
2895         }
2896         read_unlock_bh(&addrconf_hash_lock);
2897         return ret;
2898 }
2899 #endif
2900
2901 /*
2902  *      Periodic address status verification
2903  */
2904
2905 static void addrconf_verify(unsigned long foo)
2906 {
2907         struct inet6_ifaddr *ifp;
2908         unsigned long now, next;
2909         int i;
2910
2911         spin_lock_bh(&addrconf_verify_lock);
2912         now = jiffies;
2913         next = now + ADDR_CHECK_FREQUENCY;
2914
2915         del_timer(&addr_chk_timer);
2916
2917         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2918
2919 restart:
2920                 read_lock(&addrconf_hash_lock);
2921                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2922                         unsigned long age;
2923 #ifdef CONFIG_IPV6_PRIVACY
2924                         unsigned long regen_advance;
2925 #endif
2926
2927                         if (ifp->flags & IFA_F_PERMANENT)
2928                                 continue;
2929
2930                         spin_lock(&ifp->lock);
2931                         age = (now - ifp->tstamp) / HZ;
2932
2933 #ifdef CONFIG_IPV6_PRIVACY
2934                         regen_advance = ifp->idev->cnf.regen_max_retry *
2935                                         ifp->idev->cnf.dad_transmits *
2936                                         ifp->idev->nd_parms->retrans_time / HZ;
2937 #endif
2938
2939                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2940                             age >= ifp->valid_lft) {
2941                                 spin_unlock(&ifp->lock);
2942                                 in6_ifa_hold(ifp);
2943                                 read_unlock(&addrconf_hash_lock);
2944                                 ipv6_del_addr(ifp);
2945                                 goto restart;
2946                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2947                                 spin_unlock(&ifp->lock);
2948                                 continue;
2949                         } else if (age >= ifp->prefered_lft) {
2950                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2951                                 int deprecate = 0;
2952
2953                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2954                                         deprecate = 1;
2955                                         ifp->flags |= IFA_F_DEPRECATED;
2956                                 }
2957
2958                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2959                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2960
2961                                 spin_unlock(&ifp->lock);
2962
2963                                 if (deprecate) {
2964                                         in6_ifa_hold(ifp);
2965                                         read_unlock(&addrconf_hash_lock);
2966
2967                                         ipv6_ifa_notify(0, ifp);
2968                                         in6_ifa_put(ifp);
2969                                         goto restart;
2970                                 }
2971 #ifdef CONFIG_IPV6_PRIVACY
2972                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2973                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2974                                 if (age >= ifp->prefered_lft - regen_advance) {
2975                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2976                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2977                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2978                                         if (!ifp->regen_count && ifpub) {
2979                                                 ifp->regen_count++;
2980                                                 in6_ifa_hold(ifp);
2981                                                 in6_ifa_hold(ifpub);
2982                                                 spin_unlock(&ifp->lock);
2983                                                 read_unlock(&addrconf_hash_lock);
2984                                                 spin_lock(&ifpub->lock);
2985                                                 ifpub->regen_count = 0;
2986                                                 spin_unlock(&ifpub->lock);
2987                                                 ipv6_create_tempaddr(ifpub, ifp);
2988                                                 in6_ifa_put(ifpub);
2989                                                 in6_ifa_put(ifp);
2990                                                 goto restart;
2991                                         }
2992                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2993                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2994                                 spin_unlock(&ifp->lock);
2995 #endif
2996                         } else {
2997                                 /* ifp->prefered_lft <= ifp->valid_lft */
2998                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2999                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3000                                 spin_unlock(&ifp->lock);
3001                         }
3002                 }
3003                 read_unlock(&addrconf_hash_lock);
3004         }
3005
3006         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3007         add_timer(&addr_chk_timer);
3008         spin_unlock_bh(&addrconf_verify_lock);
3009 }
3010
3011 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3012 {
3013         struct in6_addr *pfx = NULL;
3014
3015         if (addr)
3016                 pfx = nla_data(addr);
3017
3018         if (local) {
3019                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3020                         pfx = NULL;
3021                 else
3022                         pfx = nla_data(local);
3023         }
3024
3025         return pfx;
3026 }
3027
3028 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3029         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3030         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3031         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3032 };
3033
3034 static int
3035 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3036 {
3037         struct net *net = skb->sk->sk_net;
3038         struct ifaddrmsg *ifm;
3039         struct nlattr *tb[IFA_MAX+1];
3040         struct in6_addr *pfx;
3041         int err;
3042
3043         if (net != &init_net)
3044                 return -EINVAL;
3045
3046         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3047         if (err < 0)
3048                 return err;
3049
3050         ifm = nlmsg_data(nlh);
3051         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3052         if (pfx == NULL)
3053                 return -EINVAL;
3054
3055         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3056 }
3057
3058 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3059                              u32 prefered_lft, u32 valid_lft)
3060 {
3061         u32 flags = RTF_EXPIRES;
3062
3063         if (!valid_lft || (prefered_lft > valid_lft))
3064                 return -EINVAL;
3065
3066         if (valid_lft == INFINITY_LIFE_TIME) {
3067                 ifa_flags |= IFA_F_PERMANENT;
3068                 flags = 0;
3069         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3070                 valid_lft = 0x7FFFFFFF/HZ;
3071
3072         if (prefered_lft == 0)
3073                 ifa_flags |= IFA_F_DEPRECATED;
3074         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3075                  (prefered_lft != INFINITY_LIFE_TIME))
3076                 prefered_lft = 0x7FFFFFFF/HZ;
3077
3078         spin_lock_bh(&ifp->lock);
3079         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3080         ifp->tstamp = jiffies;
3081         ifp->valid_lft = valid_lft;
3082         ifp->prefered_lft = prefered_lft;
3083
3084         spin_unlock_bh(&ifp->lock);
3085         if (!(ifp->flags&IFA_F_TENTATIVE))
3086                 ipv6_ifa_notify(0, ifp);
3087
3088         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3089                               jiffies_to_clock_t(valid_lft * HZ), flags);
3090         addrconf_verify(0);
3091
3092         return 0;
3093 }
3094
3095 static int
3096 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3097 {
3098         struct net *net = skb->sk->sk_net;
3099         struct ifaddrmsg *ifm;
3100         struct nlattr *tb[IFA_MAX+1];
3101         struct in6_addr *pfx;
3102         struct inet6_ifaddr *ifa;
3103         struct net_device *dev;
3104         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3105         u8 ifa_flags;
3106         int err;
3107
3108         if (net != &init_net)
3109                 return -EINVAL;
3110
3111         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3112         if (err < 0)
3113                 return err;
3114
3115         ifm = nlmsg_data(nlh);
3116         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3117         if (pfx == NULL)
3118                 return -EINVAL;
3119
3120         if (tb[IFA_CACHEINFO]) {
3121                 struct ifa_cacheinfo *ci;
3122
3123                 ci = nla_data(tb[IFA_CACHEINFO]);
3124                 valid_lft = ci->ifa_valid;
3125                 preferred_lft = ci->ifa_prefered;
3126         } else {
3127                 preferred_lft = INFINITY_LIFE_TIME;
3128                 valid_lft = INFINITY_LIFE_TIME;
3129         }
3130
3131         dev =  __dev_get_by_index(&init_net, ifm->ifa_index);
3132         if (dev == NULL)
3133                 return -ENODEV;
3134
3135         /* We ignore other flags so far. */
3136         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3137
3138         ifa = ipv6_get_ifaddr(pfx, dev, 1);
3139         if (ifa == NULL) {
3140                 /*
3141                  * It would be best to check for !NLM_F_CREATE here but
3142                  * userspace alreay relies on not having to provide this.
3143                  */
3144                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3145                                       ifa_flags, preferred_lft, valid_lft);
3146         }
3147
3148         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3149             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3150                 err = -EEXIST;
3151         else
3152                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3153
3154         in6_ifa_put(ifa);
3155
3156         return err;
3157 }
3158
3159 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3160                           u8 scope, int ifindex)
3161 {
3162         struct ifaddrmsg *ifm;
3163
3164         ifm = nlmsg_data(nlh);
3165         ifm->ifa_family = AF_INET6;
3166         ifm->ifa_prefixlen = prefixlen;
3167         ifm->ifa_flags = flags;
3168         ifm->ifa_scope = scope;
3169         ifm->ifa_index = ifindex;
3170 }
3171
3172 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3173                          unsigned long tstamp, u32 preferred, u32 valid)
3174 {
3175         struct ifa_cacheinfo ci;
3176
3177         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3178                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3179         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3180                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3181         ci.ifa_prefered = preferred;
3182         ci.ifa_valid = valid;
3183
3184         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3185 }
3186
3187 static inline int rt_scope(int ifa_scope)
3188 {
3189         if (ifa_scope & IFA_HOST)
3190                 return RT_SCOPE_HOST;
3191         else if (ifa_scope & IFA_LINK)
3192                 return RT_SCOPE_LINK;
3193         else if (ifa_scope & IFA_SITE)
3194                 return RT_SCOPE_SITE;
3195         else
3196                 return RT_SCOPE_UNIVERSE;
3197 }
3198
3199 static inline int inet6_ifaddr_msgsize(void)
3200 {
3201         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3202                + nla_total_size(16) /* IFA_ADDRESS */
3203                + nla_total_size(sizeof(struct ifa_cacheinfo));
3204 }
3205
3206 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3207                              u32 pid, u32 seq, int event, unsigned int flags)
3208 {
3209         struct nlmsghdr  *nlh;
3210         u32 preferred, valid;
3211
3212         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3213         if (nlh == NULL)
3214                 return -EMSGSIZE;
3215
3216         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3217                       ifa->idev->dev->ifindex);
3218
3219         if (!(ifa->flags&IFA_F_PERMANENT)) {
3220                 preferred = ifa->prefered_lft;
3221                 valid = ifa->valid_lft;
3222                 if (preferred != INFINITY_LIFE_TIME) {
3223                         long tval = (jiffies - ifa->tstamp)/HZ;
3224                         preferred -= tval;
3225                         if (valid != INFINITY_LIFE_TIME)
3226                                 valid -= tval;
3227                 }
3228         } else {
3229                 preferred = INFINITY_LIFE_TIME;
3230                 valid = INFINITY_LIFE_TIME;
3231         }
3232
3233         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3234             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3235                 nlmsg_cancel(skb, nlh);
3236                 return -EMSGSIZE;
3237         }
3238
3239         return nlmsg_end(skb, nlh);
3240 }
3241
3242 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3243                                 u32 pid, u32 seq, int event, u16 flags)
3244 {
3245         struct nlmsghdr  *nlh;
3246         u8 scope = RT_SCOPE_UNIVERSE;
3247         int ifindex = ifmca->idev->dev->ifindex;
3248
3249         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3250                 scope = RT_SCOPE_SITE;
3251
3252         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3253         if (nlh == NULL)
3254                 return -EMSGSIZE;
3255
3256         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3257         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3258             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3259                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3260                 nlmsg_cancel(skb, nlh);
3261                 return -EMSGSIZE;
3262         }
3263
3264         return nlmsg_end(skb, nlh);
3265 }
3266
3267 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3268                                 u32 pid, u32 seq, int event, unsigned int flags)
3269 {
3270         struct nlmsghdr  *nlh;
3271         u8 scope = RT_SCOPE_UNIVERSE;
3272         int ifindex = ifaca->aca_idev->dev->ifindex;
3273
3274         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3275                 scope = RT_SCOPE_SITE;
3276
3277         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3278         if (nlh == NULL)
3279                 return -EMSGSIZE;
3280
3281         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3282         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3283             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3284                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3285                 nlmsg_cancel(skb, nlh);
3286                 return -EMSGSIZE;
3287         }
3288
3289         return nlmsg_end(skb, nlh);
3290 }
3291
3292 enum addr_type_t
3293 {
3294         UNICAST_ADDR,
3295         MULTICAST_ADDR,
3296         ANYCAST_ADDR,
3297 };
3298
3299 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3300                            enum addr_type_t type)
3301 {
3302         int idx, ip_idx;
3303         int s_idx, s_ip_idx;
3304         int err = 1;
3305         struct net_device *dev;
3306         struct inet6_dev *idev = NULL;
3307         struct inet6_ifaddr *ifa;
3308         struct ifmcaddr6 *ifmca;
3309         struct ifacaddr6 *ifaca;
3310
3311         s_idx = cb->args[0];
3312         s_ip_idx = ip_idx = cb->args[1];
3313
3314         idx = 0;
3315         for_each_netdev(&init_net, dev) {
3316                 if (idx < s_idx)
3317                         goto cont;
3318                 if (idx > s_idx)
3319                         s_ip_idx = 0;
3320                 ip_idx = 0;
3321                 if ((idev = in6_dev_get(dev)) == NULL)
3322                         goto cont;
3323                 read_lock_bh(&idev->lock);
3324                 switch (type) {
3325                 case UNICAST_ADDR:
3326                         /* unicast address incl. temp addr */
3327                         for (ifa = idev->addr_list; ifa;
3328                              ifa = ifa->if_next, ip_idx++) {
3329                                 if (ip_idx < s_ip_idx)
3330                                         continue;
3331                                 if ((err = inet6_fill_ifaddr(skb, ifa,
3332                                     NETLINK_CB(cb->skb).pid,
3333                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3334                                     NLM_F_MULTI)) <= 0)
3335                                         goto done;
3336                         }
3337                         break;
3338                 case MULTICAST_ADDR:
3339                         /* multicast address */
3340                         for (ifmca = idev->mc_list; ifmca;
3341                              ifmca = ifmca->next, ip_idx++) {
3342                                 if (ip_idx < s_ip_idx)
3343                                         continue;
3344                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3345                                     NETLINK_CB(cb->skb).pid,
3346                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3347                                     NLM_F_MULTI)) <= 0)
3348                                         goto done;
3349                         }
3350                         break;
3351                 case ANYCAST_ADDR:
3352                         /* anycast address */
3353                         for (ifaca = idev->ac_list; ifaca;
3354                              ifaca = ifaca->aca_next, ip_idx++) {
3355                                 if (ip_idx < s_ip_idx)
3356                                         continue;
3357                                 if ((err = inet6_fill_ifacaddr(skb, ifaca,
3358                                     NETLINK_CB(cb->skb).pid,
3359                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3360                                     NLM_F_MULTI)) <= 0)
3361                                         goto done;
3362                         }
3363                         break;
3364                 default:
3365                         break;
3366                 }
3367                 read_unlock_bh(&idev->lock);
3368                 in6_dev_put(idev);
3369 cont:
3370                 idx++;
3371         }
3372 done:
3373         if (err <= 0) {
3374                 read_unlock_bh(&idev->lock);
3375                 in6_dev_put(idev);
3376         }
3377         cb->args[0] = idx;
3378         cb->args[1] = ip_idx;
3379         return skb->len;
3380 }
3381
3382 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3383 {
3384         struct net *net = skb->sk->sk_net;
3385         enum addr_type_t type = UNICAST_ADDR;
3386
3387         if (net != &init_net)
3388                 return 0;
3389
3390         return inet6_dump_addr(skb, cb, type);
3391 }
3392
3393 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3394 {
3395         struct net *net = skb->sk->sk_net;
3396         enum addr_type_t type = MULTICAST_ADDR;
3397
3398         if (net != &init_net)
3399                 return 0;
3400
3401         return inet6_dump_addr(skb, cb, type);
3402 }
3403
3404
3405 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3406 {
3407         struct net *net = skb->sk->sk_net;
3408         enum addr_type_t type = ANYCAST_ADDR;
3409
3410         if (net != &init_net)
3411                 return 0;
3412
3413         return inet6_dump_addr(skb, cb, type);
3414 }
3415
3416 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3417                              void *arg)
3418 {
3419         struct net *net = in_skb->sk->sk_net;
3420         struct ifaddrmsg *ifm;
3421         struct nlattr *tb[IFA_MAX+1];
3422         struct in6_addr *addr = NULL;
3423         struct net_device *dev = NULL;
3424         struct inet6_ifaddr *ifa;
3425         struct sk_buff *skb;
3426         int err;
3427
3428         if (net != &init_net)
3429                 return -EINVAL;
3430
3431         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3432         if (err < 0)
3433                 goto errout;
3434
3435         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3436         if (addr == NULL) {
3437                 err = -EINVAL;
3438                 goto errout;
3439         }
3440
3441         ifm = nlmsg_data(nlh);
3442         if (ifm->ifa_index)
3443                 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3444
3445         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3446                 err = -EADDRNOTAVAIL;
3447                 goto errout;
3448         }
3449
3450         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3451                 err = -ENOBUFS;
3452                 goto errout_ifa;
3453         }
3454
3455         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3456                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3457         if (err < 0) {
3458                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3459                 WARN_ON(err == -EMSGSIZE);
3460                 kfree_skb(skb);
3461                 goto errout_ifa;
3462         }
3463         err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
3464 errout_ifa:
3465         in6_ifa_put(ifa);
3466 errout:
3467         return err;
3468 }
3469
3470 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3471 {
3472         struct sk_buff *skb;
3473         int err = -ENOBUFS;
3474
3475         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3476         if (skb == NULL)
3477                 goto errout;
3478
3479         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3480         if (err < 0) {
3481                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3482                 WARN_ON(err == -EMSGSIZE);
3483                 kfree_skb(skb);
3484                 goto errout;
3485         }
3486         err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3487 errout:
3488         if (err < 0)
3489                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3490 }
3491
3492 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3493                                 __s32 *array, int bytes)
3494 {
3495         BUG_ON(bytes < (DEVCONF_MAX * 4));
3496
3497         memset(array, 0, bytes);
3498         array[DEVCONF_FORWARDING] = cnf->forwarding;
3499         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3500         array[DEVCONF_MTU6] = cnf->mtu6;
3501         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3502         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3503         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3504         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3505         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3506         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3507         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3508         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3509 #ifdef CONFIG_IPV6_PRIVACY
3510         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3511         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3512         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3513         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3514         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3515 #endif
3516         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3517         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3518         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3519 #ifdef CONFIG_IPV6_ROUTER_PREF
3520         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3521         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3522 #ifdef CONFIG_IPV6_ROUTE_INFO
3523         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3524 #endif
3525 #endif
3526         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3527         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3528 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3529         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3530 #endif
3531 }
3532
3533 static inline size_t inet6_if_nlmsg_size(void)
3534 {
3535         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3536                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3537                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3538                + nla_total_size(4) /* IFLA_MTU */
3539                + nla_total_size(4) /* IFLA_LINK */
3540                + nla_total_size( /* IFLA_PROTINFO */
3541                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3542                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3543                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3544                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3545                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3546                  );
3547 }
3548
3549 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3550                                       int bytes)
3551 {
3552         int i;
3553         int pad = bytes - sizeof(u64) * items;
3554         BUG_ON(pad < 0);
3555
3556         /* Use put_unaligned() because stats may not be aligned for u64. */
3557         put_unaligned(items, &stats[0]);
3558         for (i = 1; i < items; i++)
3559                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3560
3561         memset(&stats[items], 0, pad);
3562 }
3563
3564 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3565                              int bytes)
3566 {
3567         switch(attrtype) {
3568         case IFLA_INET6_STATS:
3569                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3570                 break;
3571         case IFLA_INET6_ICMP6STATS:
3572                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3573                 break;
3574         }
3575 }
3576
3577 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3578                              u32 pid, u32 seq, int event, unsigned int flags)
3579 {
3580         struct net_device *dev = idev->dev;
3581         struct nlattr *nla;
3582         struct ifinfomsg *hdr;
3583         struct nlmsghdr *nlh;
3584         void *protoinfo;
3585         struct ifla_cacheinfo ci;
3586
3587         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3588         if (nlh == NULL)
3589                 return -EMSGSIZE;
3590
3591         hdr = nlmsg_data(nlh);
3592         hdr->ifi_family = AF_INET6;
3593         hdr->__ifi_pad = 0;
3594         hdr->ifi_type = dev->type;
3595         hdr->ifi_index = dev->ifindex;
3596         hdr->ifi_flags = dev_get_flags(dev);
3597         hdr->ifi_change = 0;
3598
3599         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3600
3601         if (dev->addr_len)
3602                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3603
3604         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3605         if (dev->ifindex != dev->iflink)
3606                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3607
3608         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3609         if (protoinfo == NULL)
3610                 goto nla_put_failure;
3611
3612         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3613
3614         ci.max_reasm_len = IPV6_MAXPLEN;
3615         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3616                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3617         ci.reachable_time = idev->nd_parms->reachable_time;
3618         ci.retrans_time = idev->nd_parms->retrans_time;
3619         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3620
3621         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3622         if (nla == NULL)
3623                 goto nla_put_failure;
3624         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3625
3626         /* XXX - MC not implemented */
3627
3628         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3629         if (nla == NULL)
3630                 goto nla_put_failure;
3631         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3632
3633         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3634         if (nla == NULL)
3635                 goto nla_put_failure;
3636         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3637
3638         nla_nest_end(skb, protoinfo);
3639         return nlmsg_end(skb, nlh);
3640
3641 nla_put_failure:
3642         nlmsg_cancel(skb, nlh);
3643         return -EMSGSIZE;
3644 }
3645
3646 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3647 {
3648         struct net *net = skb->sk->sk_net;
3649         int idx, err;
3650         int s_idx = cb->args[0];
3651         struct net_device *dev;
3652         struct inet6_dev *idev;
3653
3654         if (net != &init_net)
3655                 return 0;
3656
3657         read_lock(&dev_base_lock);
3658         idx = 0;
3659         for_each_netdev(&init_net, dev) {
3660                 if (idx < s_idx)
3661                         goto cont;
3662                 if ((idev = in6_dev_get(dev)) == NULL)
3663                         goto cont;
3664                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3665                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3666                 in6_dev_put(idev);
3667                 if (err <= 0)
3668                         break;
3669 cont:
3670                 idx++;
3671         }
3672         read_unlock(&dev_base_lock);
3673         cb->args[0] = idx;
3674
3675         return skb->len;
3676 }
3677
3678 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3679 {
3680         struct sk_buff *skb;
3681         int err = -ENOBUFS;
3682
3683         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3684         if (skb == NULL)
3685                 goto errout;
3686
3687         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3688         if (err < 0) {
3689                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3690                 WARN_ON(err == -EMSGSIZE);
3691                 kfree_skb(skb);
3692                 goto errout;
3693         }
3694         err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3695 errout:
3696         if (err < 0)
3697                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3698 }
3699
3700 static inline size_t inet6_prefix_nlmsg_size(void)
3701 {
3702         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3703                + nla_total_size(sizeof(struct in6_addr))
3704                + nla_total_size(sizeof(struct prefix_cacheinfo));
3705 }
3706
3707 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3708                              struct prefix_info *pinfo, u32 pid, u32 seq,
3709                              int event, unsigned int flags)
3710 {
3711         struct prefixmsg *pmsg;
3712         struct nlmsghdr *nlh;
3713         struct prefix_cacheinfo ci;
3714
3715         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3716         if (nlh == NULL)
3717                 return -EMSGSIZE;
3718
3719         pmsg = nlmsg_data(nlh);
3720         pmsg->prefix_family = AF_INET6;
3721         pmsg->prefix_pad1 = 0;
3722         pmsg->prefix_pad2 = 0;
3723         pmsg->prefix_ifindex = idev->dev->ifindex;
3724         pmsg->prefix_len = pinfo->prefix_len;
3725         pmsg->prefix_type = pinfo->type;
3726         pmsg->prefix_pad3 = 0;
3727         pmsg->prefix_flags = 0;
3728         if (pinfo->onlink)
3729                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3730         if (pinfo->autoconf)
3731                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3732
3733         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3734
3735         ci.preferred_time = ntohl(pinfo->prefered);
3736         ci.valid_time = ntohl(pinfo->valid);
3737         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3738
3739         return nlmsg_end(skb, nlh);
3740
3741 nla_put_failure:
3742         nlmsg_cancel(skb, nlh);
3743         return -EMSGSIZE;
3744 }
3745
3746 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3747                          struct prefix_info *pinfo)
3748 {
3749         struct sk_buff *skb;
3750         int err = -ENOBUFS;
3751
3752         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3753         if (skb == NULL)
3754                 goto errout;
3755
3756         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3757         if (err < 0) {
3758                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3759                 WARN_ON(err == -EMSGSIZE);
3760                 kfree_skb(skb);
3761                 goto errout;
3762         }
3763         err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3764 errout:
3765         if (err < 0)
3766                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_PREFIX, err);
3767 }
3768
3769 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3770 {
3771         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3772
3773         switch (event) {
3774         case RTM_NEWADDR:
3775                 /*
3776                  * If the address was optimistic
3777                  * we inserted the route at the start of
3778                  * our DAD process, so we don't need
3779                  * to do it again
3780                  */
3781                 if (!(ifp->rt->rt6i_node))
3782                         ip6_ins_rt(ifp->rt);
3783                 if (ifp->idev->cnf.forwarding)
3784                         addrconf_join_anycast(ifp);
3785                 break;
3786         case RTM_DELADDR:
3787                 if (ifp->idev->cnf.forwarding)
3788                         addrconf_leave_anycast(ifp);
3789                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3790                 dst_hold(&ifp->rt->u.dst);
3791                 if (ip6_del_rt(ifp->rt))
3792                         dst_free(&ifp->rt->u.dst);
3793                 break;
3794         }
3795 }
3796
3797 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3798 {
3799         rcu_read_lock_bh();
3800         if (likely(ifp->idev->dead == 0))
3801                 __ipv6_ifa_notify(event, ifp);
3802         rcu_read_unlock_bh();
3803 }
3804
3805 #ifdef CONFIG_SYSCTL
3806
3807 static
3808 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3809                            void __user *buffer, size_t *lenp, loff_t *ppos)
3810 {
3811         int *valp = ctl->data;
3812         int val = *valp;
3813         int ret;
3814
3815         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3816
3817         if (write)
3818                 addrconf_fixup_forwarding(ctl, valp, val);
3819         return ret;
3820 }
3821
3822 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3823                                             int __user *name, int nlen,
3824                                             void __user *oldval,
3825                                             size_t __user *oldlenp,
3826                                             void __user *newval, size_t newlen)
3827 {
3828         int *valp = table->data;
3829         int val = *valp;
3830         int new;
3831
3832         if (!newval || !newlen)
3833                 return 0;
3834         if (newlen != sizeof(int))
3835                 return -EINVAL;
3836         if (get_user(new, (int __user *)newval))
3837                 return -EFAULT;
3838         if (new == *valp)
3839                 return 0;
3840         if (oldval && oldlenp) {
3841                 size_t len;
3842                 if (get_user(len, oldlenp))
3843                         return -EFAULT;
3844                 if (len) {
3845                         if (len > table->maxlen)
3846                                 len = table->maxlen;
3847                         if (copy_to_user(oldval, valp, len))
3848                                 return -EFAULT;
3849                         if (put_user(len, oldlenp))
3850                                 return -EFAULT;
3851                 }
3852         }
3853
3854         *valp = new;
3855         addrconf_fixup_forwarding(table, valp, val);
3856         return 1;
3857 }
3858
3859 static struct addrconf_sysctl_table
3860 {
3861         struct ctl_table_header *sysctl_header;
3862         ctl_table addrconf_vars[__NET_IPV6_MAX];
3863         char *dev_name;
3864 } addrconf_sysctl __read_mostly = {
3865         .sysctl_header = NULL,
3866         .addrconf_vars = {
3867                 {
3868                         .ctl_name       =       NET_IPV6_FORWARDING,
3869                         .procname       =       "forwarding",
3870                         .data           =       &ipv6_devconf.forwarding,
3871                         .maxlen         =       sizeof(int),
3872                         .mode           =       0644,
3873                         .proc_handler   =       &addrconf_sysctl_forward,
3874                         .strategy       =       &addrconf_sysctl_forward_strategy,
3875                 },
3876                 {
3877                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3878                         .procname       =       "hop_limit",
3879                         .data           =       &ipv6_devconf.hop_limit,
3880                         .maxlen         =       sizeof(int),
3881                         .mode           =       0644,
3882                         .proc_handler   =       proc_dointvec,
3883                 },
3884                 {
3885                         .ctl_name       =       NET_IPV6_MTU,
3886                         .procname       =       "mtu",
3887                         .data           =       &ipv6_devconf.mtu6,
3888                         .maxlen         =       sizeof(int),
3889                         .mode           =       0644,
3890                         .proc_handler   =       &proc_dointvec,
3891                 },
3892                 {
3893                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3894                         .procname       =       "accept_ra",
3895                         .data           =       &ipv6_devconf.accept_ra,
3896                         .maxlen         =       sizeof(int),
3897                         .mode           =       0644,
3898                         .proc_handler   =       &proc_dointvec,
3899                 },
3900                 {
3901                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3902                         .procname       =       "accept_redirects",
3903                         .data           =       &ipv6_devconf.accept_redirects,
3904                         .maxlen         =       sizeof(int),
3905                         .mode           =       0644,
3906                         .proc_handler   =       &proc_dointvec,
3907                 },
3908                 {
3909                         .ctl_name       =       NET_IPV6_AUTOCONF,
3910                         .procname       =       "autoconf",
3911                         .data           =       &ipv6_devconf.autoconf,
3912                         .maxlen         =       sizeof(int),
3913                         .mode           =       0644,
3914                         .proc_handler   =       &proc_dointvec,
3915                 },
3916                 {
3917                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3918                         .procname       =       "dad_transmits",
3919                         .data           =       &ipv6_devconf.dad_transmits,
3920                         .maxlen         =       sizeof(int),
3921                         .mode           =       0644,
3922                         .proc_handler   =       &proc_dointvec,
3923                 },
3924                 {
3925                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3926                         .procname       =       "router_solicitations",
3927                         .data           =       &ipv6_devconf.rtr_solicits,
3928                         .maxlen         =       sizeof(int),
3929                         .mode           =       0644,
3930                         .proc_handler   =       &proc_dointvec,
3931                 },
3932                 {
3933                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3934                         .procname       =       "router_solicitation_interval",
3935                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3936                         .maxlen         =       sizeof(int),
3937                         .mode           =       0644,
3938                         .proc_handler   =       &proc_dointvec_jiffies,
3939                         .strategy       =       &sysctl_jiffies,
3940                 },
3941                 {
3942                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3943                         .procname       =       "router_solicitation_delay",
3944                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3945                         .maxlen         =       sizeof(int),
3946                         .mode           =       0644,
3947                         .proc_handler   =       &proc_dointvec_jiffies,
3948                         .strategy       =       &sysctl_jiffies,
3949                 },
3950                 {
3951                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3952                         .procname       =       "force_mld_version",
3953                         .data           =       &ipv6_devconf.force_mld_version,
3954                         .maxlen         =       sizeof(int),
3955                         .mode           =       0644,
3956                         .proc_handler   =       &proc_dointvec,
3957                 },
3958 #ifdef CONFIG_IPV6_PRIVACY
3959                 {
3960                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3961                         .procname       =       "use_tempaddr",
3962                         .data           =       &ipv6_devconf.use_tempaddr,
3963                         .maxlen         =       sizeof(int),
3964                         .mode           =       0644,
3965                         .proc_handler   =       &proc_dointvec,
3966                 },
3967                 {
3968                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3969                         .procname       =       "temp_valid_lft",
3970                         .data           =       &ipv6_devconf.temp_valid_lft,
3971                         .maxlen         =       sizeof(int),
3972                         .mode           =       0644,
3973                         .proc_handler   =       &proc_dointvec,
3974                 },
3975                 {
3976                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3977                         .procname       =       "temp_prefered_lft",
3978                         .data           =       &ipv6_devconf.temp_prefered_lft,
3979                         .maxlen         =       sizeof(int),
3980                         .mode           =       0644,
3981                         .proc_handler   =       &proc_dointvec,
3982                 },
3983                 {
3984                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3985                         .procname       =       "regen_max_retry",
3986                         .data           =       &ipv6_devconf.regen_max_retry,
3987                         .maxlen         =       sizeof(int),
3988                         .mode           =       0644,
3989                         .proc_handler   =       &proc_dointvec,
3990                 },
3991                 {
3992                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3993                         .procname       =       "max_desync_factor",
3994                         .data           =       &ipv6_devconf.max_desync_factor,
3995                         .maxlen         =       sizeof(int),
3996                         .mode           =       0644,
3997                         .proc_handler   =       &proc_dointvec,
3998                 },
3999 #endif
4000                 {
4001                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4002                         .procname       =       "max_addresses",
4003                         .data           =       &ipv6_devconf.max_addresses,
4004                         .maxlen         =       sizeof(int),
4005                         .mode           =       0644,
4006                         .proc_handler   =       &proc_dointvec,
4007                 },
4008                 {
4009                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4010                         .procname       =       "accept_ra_defrtr",
4011                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4012                         .maxlen         =       sizeof(int),
4013                         .mode           =       0644,
4014                         .proc_handler   =       &proc_dointvec,
4015                 },
4016                 {
4017                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4018                         .procname       =       "accept_ra_pinfo",
4019                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4020                         .maxlen         =       sizeof(int),
4021                         .mode           =       0644,
4022                         .proc_handler   =       &proc_dointvec,
4023                 },
4024 #ifdef CONFIG_IPV6_ROUTER_PREF
4025                 {
4026                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4027                         .procname       =       "accept_ra_rtr_pref",
4028                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4029                         .maxlen         =       sizeof(int),
4030                         .mode           =       0644,
4031                         .proc_handler   =       &proc_dointvec,
4032                 },
4033                 {
4034                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4035                         .procname       =       "router_probe_interval",
4036                         .data           =       &ipv6_devconf.rtr_probe_interval,
4037                         .maxlen         =       sizeof(int),
4038                         .mode           =       0644,
4039                         .proc_handler   =       &proc_dointvec_jiffies,
4040                         .strategy       =       &sysctl_jiffies,
4041                 },
4042 #ifdef CONFIG_IPV6_ROUTE_INFO
4043                 {
4044                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4045                         .procname       =       "accept_ra_rt_info_max_plen",
4046                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4047                         .maxlen         =       sizeof(int),
4048                         .mode           =       0644,
4049                         .proc_handler   =       &proc_dointvec,
4050                 },
4051 #endif
4052 #endif
4053                 {
4054                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4055                         .procname       =       "proxy_ndp",
4056                         .data           =       &ipv6_devconf.proxy_ndp,
4057                         .maxlen         =       sizeof(int),
4058                         .mode           =       0644,
4059                         .proc_handler   =       &proc_dointvec,
4060                 },
4061                 {
4062                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4063                         .procname       =       "accept_source_route",
4064                         .data           =       &ipv6_devconf.accept_source_route,
4065                         .maxlen         =       sizeof(int),
4066                         .mode           =       0644,
4067                         .proc_handler   =       &proc_dointvec,
4068                 },
4069 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4070                 {
4071                         .ctl_name       =       CTL_UNNUMBERED,
4072                         .procname       =       "optimistic_dad",
4073                         .data           =       &ipv6_devconf.optimistic_dad,
4074                         .maxlen         =       sizeof(int),
4075                         .mode           =       0644,
4076                         .proc_handler   =       &proc_dointvec,
4077
4078                 },
4079 #endif
4080                 {
4081                         .ctl_name       =       0,      /* sentinel */
4082                 }
4083         },
4084 };
4085
4086 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4087                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4088 {
4089         int i;
4090         struct addrconf_sysctl_table *t;
4091
4092 #define ADDRCONF_CTL_PATH_DEV   3
4093
4094         struct ctl_path addrconf_ctl_path[] = {
4095                 { .procname = "net", .ctl_name = CTL_NET, },
4096                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4097                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4098                 { /* to be set */ },
4099                 { },
4100         };
4101
4102
4103         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4104         if (t == NULL)
4105                 goto out;
4106
4107         for (i=0; t->addrconf_vars[i].data; i++) {
4108                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4109                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4110                 t->addrconf_vars[i].extra2 = net;
4111         }
4112
4113         /*
4114          * Make a copy of dev_name, because '.procname' is regarded as const
4115          * by sysctl and we wouldn't want anyone to change it under our feet
4116          * (see SIOCSIFNAME).
4117          */
4118         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4119         if (!t->dev_name)
4120                 goto free;
4121
4122         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4123         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4124
4125         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4126                         t->addrconf_vars);
4127         if (t->sysctl_header == NULL)
4128                 goto free_procname;
4129
4130         p->sysctl = t;
4131         return 0;
4132
4133 free_procname:
4134         kfree(t->dev_name);
4135 free:
4136         kfree(t);
4137 out:
4138         return -ENOBUFS;
4139 }
4140
4141 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4142 {
4143         struct addrconf_sysctl_table *t;
4144
4145         if (p->sysctl == NULL)
4146                 return;
4147
4148         t = p->sysctl;
4149         p->sysctl = NULL;
4150         unregister_sysctl_table(t->sysctl_header);
4151         kfree(t->dev_name);
4152         kfree(t);
4153 }
4154
4155 static void addrconf_sysctl_register(struct inet6_dev *idev)
4156 {
4157         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4158                               NET_IPV6_NEIGH, "ipv6",
4159                               &ndisc_ifinfo_sysctl_change,
4160                               NULL);
4161         __addrconf_sysctl_register(idev->dev->nd_net, idev->dev->name,
4162                         idev->dev->ifindex, idev, &idev->cnf);
4163 }
4164
4165 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4166 {
4167         __addrconf_sysctl_unregister(&idev->cnf);
4168         neigh_sysctl_unregister(idev->nd_parms);
4169 }
4170
4171
4172 #endif
4173
4174 static int addrconf_init_net(struct net *net)
4175 {
4176         int err;
4177         struct ipv6_devconf *all, *dflt;
4178
4179         err = -ENOMEM;
4180         all = &ipv6_devconf;
4181         dflt = &ipv6_devconf_dflt;
4182
4183         if (net != &init_net) {
4184                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4185                 if (all == NULL)
4186                         goto err_alloc_all;
4187
4188                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4189                 if (dflt == NULL)
4190                         goto err_alloc_dflt;
4191         }
4192
4193         net->ipv6.devconf_all = all;
4194         net->ipv6.devconf_dflt = dflt;
4195
4196 #ifdef CONFIG_SYSCTL
4197         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4198                         NULL, all);
4199         if (err < 0)
4200                 goto err_reg_all;
4201
4202         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4203                         NULL, dflt);
4204         if (err < 0)
4205                 goto err_reg_dflt;
4206 #endif
4207         return 0;
4208
4209 #ifdef CONFIG_SYSCTL
4210 err_reg_dflt:
4211         __addrconf_sysctl_unregister(all);
4212 err_reg_all:
4213         kfree(dflt);
4214 #endif
4215 err_alloc_dflt:
4216         kfree(all);
4217 err_alloc_all:
4218         return err;
4219 }
4220
4221 static void addrconf_exit_net(struct net *net)
4222 {
4223 #ifdef CONFIG_SYSCTL
4224         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4225         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4226 #endif
4227         if (net != &init_net) {
4228                 kfree(net->ipv6.devconf_dflt);
4229                 kfree(net->ipv6.devconf_all);
4230         }
4231 }
4232
4233 static struct pernet_operations addrconf_ops = {
4234         .init = addrconf_init_net,
4235         .exit = addrconf_exit_net,
4236 };
4237
4238 /*
4239  *      Device notifier
4240  */
4241
4242 int register_inet6addr_notifier(struct notifier_block *nb)
4243 {
4244         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4245 }
4246
4247 EXPORT_SYMBOL(register_inet6addr_notifier);
4248
4249 int unregister_inet6addr_notifier(struct notifier_block *nb)
4250 {
4251         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4252 }
4253
4254 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4255
4256 /*
4257  *      Init / cleanup code
4258  */
4259
4260 int __init addrconf_init(void)
4261 {
4262         int err;
4263
4264         if ((err = ipv6_addr_label_init()) < 0) {
4265                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4266                         err);
4267                 return err;
4268         }
4269
4270         register_pernet_subsys(&addrconf_ops);
4271
4272         /* The addrconf netdev notifier requires that loopback_dev
4273          * has it's ipv6 private information allocated and setup
4274          * before it can bring up and give link-local addresses
4275          * to other devices which are up.
4276          *
4277          * Unfortunately, loopback_dev is not necessarily the first
4278          * entry in the global dev_base list of net devices.  In fact,
4279          * it is likely to be the very last entry on that list.
4280          * So this causes the notifier registry below to try and
4281          * give link-local addresses to all devices besides loopback_dev
4282          * first, then loopback_dev, which cases all the non-loopback_dev
4283          * devices to fail to get a link-local address.
4284          *
4285          * So, as a temporary fix, allocate the ipv6 structure for
4286          * loopback_dev first by hand.
4287          * Longer term, all of the dependencies ipv6 has upon the loopback
4288          * device and it being up should be removed.
4289          */
4290         rtnl_lock();
4291         if (!ipv6_add_dev(init_net.loopback_dev))
4292                 err = -ENOMEM;
4293         rtnl_unlock();
4294         if (err)
4295                 goto errlo;
4296
4297         ip6_null_entry.u.dst.dev = init_net.loopback_dev;
4298         ip6_null_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4299 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4300         ip6_prohibit_entry.u.dst.dev = init_net.loopback_dev;
4301         ip6_prohibit_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4302         ip6_blk_hole_entry.u.dst.dev = init_net.loopback_dev;
4303         ip6_blk_hole_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4304 #endif
4305
4306         register_netdevice_notifier(&ipv6_dev_notf);
4307
4308         addrconf_verify(0);
4309
4310         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4311         if (err < 0)
4312                 goto errout;
4313
4314         /* Only the first call to __rtnl_register can fail */
4315         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4316         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4317         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4318         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4319         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4320
4321         ipv6_addr_label_rtnl_register();
4322
4323         return 0;
4324 errout:
4325         unregister_netdevice_notifier(&ipv6_dev_notf);
4326 errlo:
4327         unregister_pernet_subsys(&addrconf_ops);
4328
4329         return err;
4330 }
4331
4332 void addrconf_cleanup(void)
4333 {
4334         struct net_device *dev;
4335         struct inet6_ifaddr *ifa;
4336         int i;
4337
4338         unregister_netdevice_notifier(&ipv6_dev_notf);
4339
4340         unregister_pernet_subsys(&addrconf_ops);
4341
4342         rtnl_lock();
4343
4344         /*
4345          *      clean dev list.
4346          */
4347
4348         for_each_netdev(&init_net, dev) {
4349                 if (__in6_dev_get(dev) == NULL)
4350                         continue;
4351                 addrconf_ifdown(dev, 1);
4352         }
4353         addrconf_ifdown(init_net.loopback_dev, 2);
4354
4355         /*
4356          *      Check hash table.
4357          */
4358
4359         write_lock_bh(&addrconf_hash_lock);
4360         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4361                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4362                         struct inet6_ifaddr *bifa;
4363
4364                         bifa = ifa;
4365                         ifa = ifa->lst_next;
4366                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4367                         /* Do not free it; something is wrong.
4368                            Now we can investigate it with debugger.
4369                          */
4370                 }
4371         }
4372         write_unlock_bh(&addrconf_hash_lock);
4373
4374         del_timer(&addr_chk_timer);
4375
4376         rtnl_unlock();
4377 }