Merge branch 'net-2.6.26-isatap-20080403' of git://git.linux-ipv6.org/gitroot/yoshfuj...
[safe/jmp/linux-2.6] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16
17 /*
18  *      Changes:
19  *
20  *      Janos Farkas                    :       delete timer on ifdown
21  *      <chexum@bankinf.banki.hu>
22  *      Andi Kleen                      :       kill double kfree on module
23  *                                              unload.
24  *      Maciej W. Rozycki               :       FDDI support
25  *      sekiya@USAGI                    :       Don't send too many RS
26  *                                              packets.
27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
28  *                                              packets.
29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
30  *                                              address validation timer.
31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
32  *                                              support.
33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
34  *                                              address on a same interface.
35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
37  *                                              seq_file.
38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
39  *                                              selection; consider scope,
40  *                                              status etc.
41  */
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99
100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
108 {
109 }
110
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 {
113 }
114 #endif
115
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
120
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
123
124 static int ipv6_count_addresses(struct inet6_dev *idev);
125
126 /*
127  *      Configured unicast address hash table
128  */
129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
131
132 static void addrconf_verify(unsigned long);
133
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
136
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139
140 static int addrconf_ifdown(struct net_device *dev, int how);
141
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
149
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151                                 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153                               struct net_device *dev);
154
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156
157 struct ipv6_devconf ipv6_devconf __read_mostly = {
158         .forwarding             = 0,
159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
160         .mtu6                   = IPV6_MIN_MTU,
161         .accept_ra              = 1,
162         .accept_redirects       = 1,
163         .autoconf               = 1,
164         .force_mld_version      = 0,
165         .dad_transmits          = 1,
166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170         .use_tempaddr           = 0,
171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
173         .regen_max_retry        = REGEN_MAX_RETRY,
174         .max_desync_factor      = MAX_DESYNC_FACTOR,
175 #endif
176         .max_addresses          = IPV6_MAX_ADDRESSES,
177         .accept_ra_defrtr       = 1,
178         .accept_ra_pinfo        = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180         .accept_ra_rtr_pref     = 1,
181         .rtr_probe_interval     = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183         .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186         .proxy_ndp              = 0,
187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
188 };
189
190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
191         .forwarding             = 0,
192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
193         .mtu6                   = IPV6_MIN_MTU,
194         .accept_ra              = 1,
195         .accept_redirects       = 1,
196         .autoconf               = 1,
197         .dad_transmits          = 1,
198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
201 #ifdef CONFIG_IPV6_PRIVACY
202         .use_tempaddr           = 0,
203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
205         .regen_max_retry        = REGEN_MAX_RETRY,
206         .max_desync_factor      = MAX_DESYNC_FACTOR,
207 #endif
208         .max_addresses          = IPV6_MAX_ADDRESSES,
209         .accept_ra_defrtr       = 1,
210         .accept_ra_pinfo        = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212         .accept_ra_rtr_pref     = 1,
213         .rtr_probe_interval     = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215         .accept_ra_rt_info_max_plen = 0,
216 #endif
217 #endif
218         .proxy_ndp              = 0,
219         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
220 };
221
222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
225
226 /* Check if a valid qdisc is available */
227 static inline int addrconf_qdisc_ok(struct net_device *dev)
228 {
229         return (dev->qdisc != &noop_qdisc);
230 }
231
232 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
233 {
234         if (del_timer(&ifp->timer))
235                 __in6_ifa_put(ifp);
236 }
237
238 enum addrconf_timer_t
239 {
240         AC_NONE,
241         AC_DAD,
242         AC_RS,
243 };
244
245 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
246                                enum addrconf_timer_t what,
247                                unsigned long when)
248 {
249         if (!del_timer(&ifp->timer))
250                 in6_ifa_hold(ifp);
251
252         switch (what) {
253         case AC_DAD:
254                 ifp->timer.function = addrconf_dad_timer;
255                 break;
256         case AC_RS:
257                 ifp->timer.function = addrconf_rs_timer;
258                 break;
259         default:;
260         }
261         ifp->timer.expires = jiffies + when;
262         add_timer(&ifp->timer);
263 }
264
265 static int snmp6_alloc_dev(struct inet6_dev *idev)
266 {
267         if (snmp_mib_init((void **)idev->stats.ipv6,
268                           sizeof(struct ipstats_mib)) < 0)
269                 goto err_ip;
270         if (snmp_mib_init((void **)idev->stats.icmpv6,
271                           sizeof(struct icmpv6_mib)) < 0)
272                 goto err_icmp;
273         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
274                           sizeof(struct icmpv6msg_mib)) < 0)
275                 goto err_icmpmsg;
276
277         return 0;
278
279 err_icmpmsg:
280         snmp_mib_free((void **)idev->stats.icmpv6);
281 err_icmp:
282         snmp_mib_free((void **)idev->stats.ipv6);
283 err_ip:
284         return -ENOMEM;
285 }
286
287 static void snmp6_free_dev(struct inet6_dev *idev)
288 {
289         snmp_mib_free((void **)idev->stats.icmpv6msg);
290         snmp_mib_free((void **)idev->stats.icmpv6);
291         snmp_mib_free((void **)idev->stats.ipv6);
292 }
293
294 /* Nobody refers to this device, we may destroy it. */
295
296 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
297 {
298         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
299         kfree(idev);
300 }
301
302 void in6_dev_finish_destroy(struct inet6_dev *idev)
303 {
304         struct net_device *dev = idev->dev;
305         BUG_TRAP(idev->addr_list==NULL);
306         BUG_TRAP(idev->mc_list==NULL);
307 #ifdef NET_REFCNT_DEBUG
308         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
309 #endif
310         dev_put(dev);
311         if (!idev->dead) {
312                 printk("Freeing alive inet6 device %p\n", idev);
313                 return;
314         }
315         snmp6_free_dev(idev);
316         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
317 }
318
319 EXPORT_SYMBOL(in6_dev_finish_destroy);
320
321 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
322 {
323         struct inet6_dev *ndev;
324         struct in6_addr maddr;
325
326         ASSERT_RTNL();
327
328         if (dev->mtu < IPV6_MIN_MTU)
329                 return NULL;
330
331         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
332
333         if (ndev == NULL)
334                 return NULL;
335
336         rwlock_init(&ndev->lock);
337         ndev->dev = dev;
338         memcpy(&ndev->cnf, dev_net(dev)->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                         __func__, 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                         __func__, 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 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(net);
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(dev_net(idev->dev), 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                 struct net *net = dev_net(ifp->idev->dev);
755                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
756                 rt = rt6_lookup(net, &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         unsigned long regen_advance;
780         int tmp_plen;
781         int ret = 0;
782         int max_addresses;
783         u32 addr_flags;
784
785         write_lock(&idev->lock);
786         if (ift) {
787                 spin_lock_bh(&ift->lock);
788                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
789                 spin_unlock_bh(&ift->lock);
790                 tmpaddr = &addr;
791         } else {
792                 tmpaddr = NULL;
793         }
794 retry:
795         in6_dev_hold(idev);
796         if (idev->cnf.use_tempaddr <= 0) {
797                 write_unlock(&idev->lock);
798                 printk(KERN_INFO
799                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
800                 in6_dev_put(idev);
801                 ret = -1;
802                 goto out;
803         }
804         spin_lock_bh(&ifp->lock);
805         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
806                 idev->cnf.use_tempaddr = -1;    /*XXX*/
807                 spin_unlock_bh(&ifp->lock);
808                 write_unlock(&idev->lock);
809                 printk(KERN_WARNING
810                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
811                 in6_dev_put(idev);
812                 ret = -1;
813                 goto out;
814         }
815         in6_ifa_hold(ifp);
816         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
817         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
818                 spin_unlock_bh(&ifp->lock);
819                 write_unlock(&idev->lock);
820                 printk(KERN_WARNING
821                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
822                 in6_ifa_put(ifp);
823                 in6_dev_put(idev);
824                 ret = -1;
825                 goto out;
826         }
827         memcpy(&addr.s6_addr[8], idev->rndid, 8);
828         tmp_valid_lft = min_t(__u32,
829                               ifp->valid_lft,
830                               idev->cnf.temp_valid_lft);
831         tmp_prefered_lft = min_t(__u32,
832                                  ifp->prefered_lft,
833                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
834         tmp_plen = ifp->prefix_len;
835         max_addresses = idev->cnf.max_addresses;
836         tmp_cstamp = ifp->cstamp;
837         tmp_tstamp = ifp->tstamp;
838         spin_unlock_bh(&ifp->lock);
839
840         regen_advance = idev->cnf.regen_max_retry *
841                         idev->cnf.dad_transmits *
842                         idev->nd_parms->retrans_time / HZ;
843         write_unlock(&idev->lock);
844
845         /* A temporary address is created only if this calculated Preferred
846          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
847          * an implementation must not create a temporary address with a zero
848          * Preferred Lifetime.
849          */
850         if (tmp_prefered_lft <= regen_advance) {
851                 in6_ifa_put(ifp);
852                 in6_dev_put(idev);
853                 ret = -1;
854                 goto out;
855         }
856
857         addr_flags = IFA_F_TEMPORARY;
858         /* set in addrconf_prefix_rcv() */
859         if (ifp->flags & IFA_F_OPTIMISTIC)
860                 addr_flags |= IFA_F_OPTIMISTIC;
861
862         ift = !max_addresses ||
863               ipv6_count_addresses(idev) < max_addresses ?
864                 ipv6_add_addr(idev, &addr, tmp_plen,
865                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
866                               addr_flags) : NULL;
867         if (!ift || IS_ERR(ift)) {
868                 in6_ifa_put(ifp);
869                 in6_dev_put(idev);
870                 printk(KERN_INFO
871                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
872                 tmpaddr = &addr;
873                 write_lock(&idev->lock);
874                 goto retry;
875         }
876
877         spin_lock_bh(&ift->lock);
878         ift->ifpub = ifp;
879         ift->valid_lft = tmp_valid_lft;
880         ift->prefered_lft = tmp_prefered_lft;
881         ift->cstamp = tmp_cstamp;
882         ift->tstamp = tmp_tstamp;
883         spin_unlock_bh(&ift->lock);
884
885         addrconf_dad_start(ift, 0);
886         in6_ifa_put(ift);
887         in6_dev_put(idev);
888 out:
889         return ret;
890 }
891 #endif
892
893 /*
894  *      Choose an appropriate source address (RFC3484)
895  */
896 enum {
897         IPV6_SADDR_RULE_INIT = 0,
898         IPV6_SADDR_RULE_LOCAL,
899         IPV6_SADDR_RULE_SCOPE,
900         IPV6_SADDR_RULE_PREFERRED,
901 #ifdef CONFIG_IPV6_MIP6
902         IPV6_SADDR_RULE_HOA,
903 #endif
904         IPV6_SADDR_RULE_OIF,
905         IPV6_SADDR_RULE_LABEL,
906 #ifdef CONFIG_IPV6_PRIVACY
907         IPV6_SADDR_RULE_PRIVACY,
908 #endif
909         IPV6_SADDR_RULE_ORCHID,
910         IPV6_SADDR_RULE_PREFIX,
911         IPV6_SADDR_RULE_MAX
912 };
913
914 struct ipv6_saddr_score {
915         int                     rule;
916         int                     addr_type;
917         struct inet6_ifaddr     *ifa;
918         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
919         int                     scopedist;
920         int                     matchlen;
921 };
922
923 struct ipv6_saddr_dst {
924         struct in6_addr *addr;
925         int ifindex;
926         int scope;
927         int label;
928         unsigned int prefs;
929 };
930
931 static inline int ipv6_saddr_preferred(int type)
932 {
933         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
934                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
935                 return 1;
936         return 0;
937 }
938
939 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
940                                struct ipv6_saddr_dst *dst,
941                                int i)
942 {
943         int ret;
944
945         if (i <= score->rule) {
946                 switch (i) {
947                 case IPV6_SADDR_RULE_SCOPE:
948                         ret = score->scopedist;
949                         break;
950                 case IPV6_SADDR_RULE_PREFIX:
951                         ret = score->matchlen;
952                         break;
953                 default:
954                         ret = !!test_bit(i, score->scorebits);
955                 }
956                 goto out;
957         }
958
959         switch (i) {
960         case IPV6_SADDR_RULE_INIT:
961                 /* Rule 0: remember if hiscore is not ready yet */
962                 ret = !!score->ifa;
963                 break;
964         case IPV6_SADDR_RULE_LOCAL:
965                 /* Rule 1: Prefer same address */
966                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
967                 break;
968         case IPV6_SADDR_RULE_SCOPE:
969                 /* Rule 2: Prefer appropriate scope
970                  *
971                  *      ret
972                  *       ^
973                  *    -1 |  d 15
974                  *    ---+--+-+---> scope
975                  *       |
976                  *       |             d is scope of the destination.
977                  *  B-d  |  \
978                  *       |   \      <- smaller scope is better if
979                  *  B-15 |    \        if scope is enough for destinaion.
980                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
981                  * d-C-1 | /
982                  *       |/         <- greater is better
983                  *   -C  /             if scope is not enough for destination.
984                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
985                  *
986                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
987                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
988                  * Assume B = 0 and we get C > 29.
989                  */
990                 ret = __ipv6_addr_src_scope(score->addr_type);
991                 if (ret >= dst->scope)
992                         ret = -ret;
993                 else
994                         ret -= 128;     /* 30 is enough */
995                 score->scopedist = ret;
996                 break;
997         case IPV6_SADDR_RULE_PREFERRED:
998                 /* Rule 3: Avoid deprecated and optimistic addresses */
999                 ret = ipv6_saddr_preferred(score->addr_type) ||
1000                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1001                 break;
1002 #ifdef CONFIG_IPV6_MIP6
1003         case IPV6_SADDR_RULE_HOA:
1004             {
1005                 /* Rule 4: Prefer home address */
1006                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1007                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1008                 break;
1009             }
1010 #endif
1011         case IPV6_SADDR_RULE_OIF:
1012                 /* Rule 5: Prefer outgoing interface */
1013                 ret = (!dst->ifindex ||
1014                        dst->ifindex == score->ifa->idev->dev->ifindex);
1015                 break;
1016         case IPV6_SADDR_RULE_LABEL:
1017                 /* Rule 6: Prefer matching label */
1018                 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
1019                                       score->ifa->idev->dev->ifindex) == dst->label;
1020                 break;
1021 #ifdef CONFIG_IPV6_PRIVACY
1022         case IPV6_SADDR_RULE_PRIVACY:
1023             {
1024                 /* Rule 7: Prefer public address
1025                  * Note: prefer temprary address if use_tempaddr >= 2
1026                  */
1027                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1028                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1029                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1030                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1031                 break;
1032             }
1033 #endif
1034         case IPV6_SADDR_RULE_ORCHID:
1035                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1036                  *          non-ORCHID vs non-ORCHID
1037                  */
1038                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1039                         ipv6_addr_orchid(dst->addr));
1040                 break;
1041         case IPV6_SADDR_RULE_PREFIX:
1042                 /* Rule 8: Use longest matching prefix */
1043                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1044                                                        dst->addr);
1045                 break;
1046         default:
1047                 ret = 0;
1048         }
1049
1050         if (ret)
1051                 __set_bit(i, score->scorebits);
1052         score->rule = i;
1053 out:
1054         return ret;
1055 }
1056
1057 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1058                        struct in6_addr *daddr, unsigned int prefs,
1059                        struct in6_addr *saddr)
1060 {
1061         struct ipv6_saddr_score scores[2],
1062                                 *score = &scores[0], *hiscore = &scores[1];
1063         struct net *net = dev_net(dst_dev);
1064         struct ipv6_saddr_dst dst;
1065         struct net_device *dev;
1066         int dst_type;
1067
1068         dst_type = __ipv6_addr_type(daddr);
1069         dst.addr = daddr;
1070         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1071         dst.scope = __ipv6_addr_src_scope(dst_type);
1072         dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1073         dst.prefs = prefs;
1074
1075         hiscore->rule = -1;
1076         hiscore->ifa = NULL;
1077
1078         read_lock(&dev_base_lock);
1079         rcu_read_lock();
1080
1081         for_each_netdev(net, dev) {
1082                 struct inet6_dev *idev;
1083
1084                 /* Candidate Source Address (section 4)
1085                  *  - multicast and link-local destination address,
1086                  *    the set of candidate source address MUST only
1087                  *    include addresses assigned to interfaces
1088                  *    belonging to the same link as the outgoing
1089                  *    interface.
1090                  * (- For site-local destination addresses, the
1091                  *    set of candidate source addresses MUST only
1092                  *    include addresses assigned to interfaces
1093                  *    belonging to the same site as the outgoing
1094                  *    interface.)
1095                  */
1096                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1097                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1098                     dst.ifindex && dev->ifindex != dst.ifindex)
1099                         continue;
1100
1101                 idev = __in6_dev_get(dev);
1102                 if (!idev)
1103                         continue;
1104
1105                 read_lock_bh(&idev->lock);
1106                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1107                         int i;
1108
1109                         /*
1110                          * - Tentative Address (RFC2462 section 5.4)
1111                          *  - A tentative address is not considered
1112                          *    "assigned to an interface" in the traditional
1113                          *    sense, unless it is also flagged as optimistic.
1114                          * - Candidate Source Address (section 4)
1115                          *  - In any case, anycast addresses, multicast
1116                          *    addresses, and the unspecified address MUST
1117                          *    NOT be included in a candidate set.
1118                          */
1119                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1120                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1121                                 continue;
1122
1123                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1124
1125                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1126                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1127                                 LIMIT_NETDEBUG(KERN_DEBUG
1128                                                "ADDRCONF: unspecified / multicast address "
1129                                                "assigned as unicast address on %s",
1130                                                dev->name);
1131                                 continue;
1132                         }
1133
1134                         score->rule = -1;
1135                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1136
1137                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1138                                 int minihiscore, miniscore;
1139
1140                                 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1141                                 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1142
1143                                 if (minihiscore > miniscore) {
1144                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1145                                             score->scopedist > 0) {
1146                                                 /*
1147                                                  * special case:
1148                                                  * each remaining entry
1149                                                  * has too small (not enough)
1150                                                  * scope, because ifa entries
1151                                                  * are sorted by their scope
1152                                                  * values.
1153                                                  */
1154                                                 goto try_nextdev;
1155                                         }
1156                                         break;
1157                                 } else if (minihiscore < miniscore) {
1158                                         struct ipv6_saddr_score *tmp;
1159
1160                                         if (hiscore->ifa)
1161                                                 in6_ifa_put(hiscore->ifa);
1162
1163                                         in6_ifa_hold(score->ifa);
1164
1165                                         tmp = hiscore;
1166                                         hiscore = score;
1167                                         score = tmp;
1168
1169                                         /* restore our iterator */
1170                                         score->ifa = hiscore->ifa;
1171
1172                                         break;
1173                                 }
1174                         }
1175                 }
1176 try_nextdev:
1177                 read_unlock_bh(&idev->lock);
1178         }
1179         rcu_read_unlock();
1180         read_unlock(&dev_base_lock);
1181
1182         if (!hiscore->ifa)
1183                 return -EADDRNOTAVAIL;
1184
1185         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1186         in6_ifa_put(hiscore->ifa);
1187         return 0;
1188 }
1189
1190 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1191
1192 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1193                     unsigned char banned_flags)
1194 {
1195         struct inet6_dev *idev;
1196         int err = -EADDRNOTAVAIL;
1197
1198         rcu_read_lock();
1199         if ((idev = __in6_dev_get(dev)) != NULL) {
1200                 struct inet6_ifaddr *ifp;
1201
1202                 read_lock_bh(&idev->lock);
1203                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1204                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1205                                 ipv6_addr_copy(addr, &ifp->addr);
1206                                 err = 0;
1207                                 break;
1208                         }
1209                 }
1210                 read_unlock_bh(&idev->lock);
1211         }
1212         rcu_read_unlock();
1213         return err;
1214 }
1215
1216 static int ipv6_count_addresses(struct inet6_dev *idev)
1217 {
1218         int cnt = 0;
1219         struct inet6_ifaddr *ifp;
1220
1221         read_lock_bh(&idev->lock);
1222         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1223                 cnt++;
1224         read_unlock_bh(&idev->lock);
1225         return cnt;
1226 }
1227
1228 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1229                   struct net_device *dev, int strict)
1230 {
1231         struct inet6_ifaddr * ifp;
1232         u8 hash = ipv6_addr_hash(addr);
1233
1234         read_lock_bh(&addrconf_hash_lock);
1235         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1236                 if (!net_eq(dev_net(ifp->idev->dev), net))
1237                         continue;
1238                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1239                     !(ifp->flags&IFA_F_TENTATIVE)) {
1240                         if (dev == NULL || ifp->idev->dev == dev ||
1241                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1242                                 break;
1243                 }
1244         }
1245         read_unlock_bh(&addrconf_hash_lock);
1246         return ifp != NULL;
1247 }
1248 EXPORT_SYMBOL(ipv6_chk_addr);
1249
1250 static
1251 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1252                        struct net_device *dev)
1253 {
1254         struct inet6_ifaddr * ifp;
1255         u8 hash = ipv6_addr_hash(addr);
1256
1257         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1258                 if (!net_eq(dev_net(ifp->idev->dev), net))
1259                         continue;
1260                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1261                         if (dev == NULL || ifp->idev->dev == dev)
1262                                 break;
1263                 }
1264         }
1265         return ifp != NULL;
1266 }
1267
1268 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1269 {
1270         struct inet6_dev *idev;
1271         struct inet6_ifaddr *ifa;
1272         int     onlink;
1273
1274         onlink = 0;
1275         rcu_read_lock();
1276         idev = __in6_dev_get(dev);
1277         if (idev) {
1278                 read_lock_bh(&idev->lock);
1279                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1280                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1281                                                    ifa->prefix_len);
1282                         if (onlink)
1283                                 break;
1284                 }
1285                 read_unlock_bh(&idev->lock);
1286         }
1287         rcu_read_unlock();
1288         return onlink;
1289 }
1290
1291 EXPORT_SYMBOL(ipv6_chk_prefix);
1292
1293 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1294                                      struct net_device *dev, int strict)
1295 {
1296         struct inet6_ifaddr * ifp;
1297         u8 hash = ipv6_addr_hash(addr);
1298
1299         read_lock_bh(&addrconf_hash_lock);
1300         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1301                 if (!net_eq(dev_net(ifp->idev->dev), net))
1302                         continue;
1303                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1304                         if (dev == NULL || ifp->idev->dev == dev ||
1305                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1306                                 in6_ifa_hold(ifp);
1307                                 break;
1308                         }
1309                 }
1310         }
1311         read_unlock_bh(&addrconf_hash_lock);
1312
1313         return ifp;
1314 }
1315
1316 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1317 {
1318         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1319         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1320         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1321         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1322         int sk_ipv6only = ipv6_only_sock(sk);
1323         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1324         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1325         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1326
1327         if (!sk2_rcv_saddr && !sk_ipv6only)
1328                 return 1;
1329
1330         if (addr_type2 == IPV6_ADDR_ANY &&
1331             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1332                 return 1;
1333
1334         if (addr_type == IPV6_ADDR_ANY &&
1335             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1336                 return 1;
1337
1338         if (sk2_rcv_saddr6 &&
1339             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1340                 return 1;
1341
1342         if (addr_type == IPV6_ADDR_MAPPED &&
1343             !sk2_ipv6only &&
1344             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1345                 return 1;
1346
1347         return 0;
1348 }
1349
1350 /* Gets referenced address, destroys ifaddr */
1351
1352 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1353 {
1354         if (ifp->flags&IFA_F_PERMANENT) {
1355                 spin_lock_bh(&ifp->lock);
1356                 addrconf_del_timer(ifp);
1357                 ifp->flags |= IFA_F_TENTATIVE;
1358                 spin_unlock_bh(&ifp->lock);
1359                 in6_ifa_put(ifp);
1360 #ifdef CONFIG_IPV6_PRIVACY
1361         } else if (ifp->flags&IFA_F_TEMPORARY) {
1362                 struct inet6_ifaddr *ifpub;
1363                 spin_lock_bh(&ifp->lock);
1364                 ifpub = ifp->ifpub;
1365                 if (ifpub) {
1366                         in6_ifa_hold(ifpub);
1367                         spin_unlock_bh(&ifp->lock);
1368                         ipv6_create_tempaddr(ifpub, ifp);
1369                         in6_ifa_put(ifpub);
1370                 } else {
1371                         spin_unlock_bh(&ifp->lock);
1372                 }
1373                 ipv6_del_addr(ifp);
1374 #endif
1375         } else
1376                 ipv6_del_addr(ifp);
1377 }
1378
1379 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1380 {
1381         if (net_ratelimit())
1382                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1383         addrconf_dad_stop(ifp);
1384 }
1385
1386 /* Join to solicited addr multicast group. */
1387
1388 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1389 {
1390         struct in6_addr maddr;
1391
1392         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1393                 return;
1394
1395         addrconf_addr_solict_mult(addr, &maddr);
1396         ipv6_dev_mc_inc(dev, &maddr);
1397 }
1398
1399 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1400 {
1401         struct in6_addr maddr;
1402
1403         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1404                 return;
1405
1406         addrconf_addr_solict_mult(addr, &maddr);
1407         __ipv6_dev_mc_dec(idev, &maddr);
1408 }
1409
1410 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1411 {
1412         struct in6_addr addr;
1413         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1414         if (ipv6_addr_any(&addr))
1415                 return;
1416         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1417 }
1418
1419 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1420 {
1421         struct in6_addr addr;
1422         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1423         if (ipv6_addr_any(&addr))
1424                 return;
1425         __ipv6_dev_ac_dec(ifp->idev, &addr);
1426 }
1427
1428 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1429 {
1430         if (dev->addr_len != ETH_ALEN)
1431                 return -1;
1432         memcpy(eui, dev->dev_addr, 3);
1433         memcpy(eui + 5, dev->dev_addr + 3, 3);
1434
1435         /*
1436          * The zSeries OSA network cards can be shared among various
1437          * OS instances, but the OSA cards have only one MAC address.
1438          * This leads to duplicate address conflicts in conjunction
1439          * with IPv6 if more than one instance uses the same card.
1440          *
1441          * The driver for these cards can deliver a unique 16-bit
1442          * identifier for each instance sharing the same card.  It is
1443          * placed instead of 0xFFFE in the interface identifier.  The
1444          * "u" bit of the interface identifier is not inverted in this
1445          * case.  Hence the resulting interface identifier has local
1446          * scope according to RFC2373.
1447          */
1448         if (dev->dev_id) {
1449                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1450                 eui[4] = dev->dev_id & 0xFF;
1451         } else {
1452                 eui[3] = 0xFF;
1453                 eui[4] = 0xFE;
1454                 eui[0] ^= 2;
1455         }
1456         return 0;
1457 }
1458
1459 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1460 {
1461         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1462         if (dev->addr_len != ARCNET_ALEN)
1463                 return -1;
1464         memset(eui, 0, 7);
1465         eui[7] = *(u8*)dev->dev_addr;
1466         return 0;
1467 }
1468
1469 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1470 {
1471         if (dev->addr_len != INFINIBAND_ALEN)
1472                 return -1;
1473         memcpy(eui, dev->dev_addr + 12, 8);
1474         eui[0] |= 2;
1475         return 0;
1476 }
1477
1478 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1479 {
1480         switch (dev->type) {
1481         case ARPHRD_ETHER:
1482         case ARPHRD_FDDI:
1483         case ARPHRD_IEEE802_TR:
1484                 return addrconf_ifid_eui48(eui, dev);
1485         case ARPHRD_ARCNET:
1486                 return addrconf_ifid_arcnet(eui, dev);
1487         case ARPHRD_INFINIBAND:
1488                 return addrconf_ifid_infiniband(eui, dev);
1489         case ARPHRD_SIT:
1490                 if (dev->priv_flags & IFF_ISATAP)
1491                         return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1492         }
1493         return -1;
1494 }
1495
1496 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1497 {
1498         int err = -1;
1499         struct inet6_ifaddr *ifp;
1500
1501         read_lock_bh(&idev->lock);
1502         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1503                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1504                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1505                         err = 0;
1506                         break;
1507                 }
1508         }
1509         read_unlock_bh(&idev->lock);
1510         return err;
1511 }
1512
1513 #ifdef CONFIG_IPV6_PRIVACY
1514 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1515 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1516 {
1517 regen:
1518         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1519         idev->rndid[0] &= ~0x02;
1520
1521         /*
1522          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1523          * check if generated address is not inappropriate
1524          *
1525          *  - Reserved subnet anycast (RFC 2526)
1526          *      11111101 11....11 1xxxxxxx
1527          *  - ISATAP (RFC4214) 6.1
1528          *      00-00-5E-FE-xx-xx-xx-xx
1529          *  - value 0
1530          *  - XXX: already assigned to an address on the device
1531          */
1532         if (idev->rndid[0] == 0xfd &&
1533             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1534             (idev->rndid[7]&0x80))
1535                 goto regen;
1536         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1537                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1538                         goto regen;
1539                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1540                         goto regen;
1541         }
1542
1543         return 0;
1544 }
1545
1546 static void ipv6_regen_rndid(unsigned long data)
1547 {
1548         struct inet6_dev *idev = (struct inet6_dev *) data;
1549         unsigned long expires;
1550
1551         rcu_read_lock_bh();
1552         write_lock_bh(&idev->lock);
1553
1554         if (idev->dead)
1555                 goto out;
1556
1557         if (__ipv6_regen_rndid(idev) < 0)
1558                 goto out;
1559
1560         expires = jiffies +
1561                 idev->cnf.temp_prefered_lft * HZ -
1562                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1563         if (time_before(expires, jiffies)) {
1564                 printk(KERN_WARNING
1565                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1566                         idev->dev->name);
1567                 goto out;
1568         }
1569
1570         if (!mod_timer(&idev->regen_timer, expires))
1571                 in6_dev_hold(idev);
1572
1573 out:
1574         write_unlock_bh(&idev->lock);
1575         rcu_read_unlock_bh();
1576         in6_dev_put(idev);
1577 }
1578
1579 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1580         int ret = 0;
1581
1582         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1583                 ret = __ipv6_regen_rndid(idev);
1584         return ret;
1585 }
1586 #endif
1587
1588 /*
1589  *      Add prefix route.
1590  */
1591
1592 static void
1593 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1594                       unsigned long expires, u32 flags)
1595 {
1596         struct fib6_config cfg = {
1597                 .fc_table = RT6_TABLE_PREFIX,
1598                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1599                 .fc_ifindex = dev->ifindex,
1600                 .fc_expires = expires,
1601                 .fc_dst_len = plen,
1602                 .fc_flags = RTF_UP | flags,
1603                 .fc_nlinfo.nl_net = dev_net(dev),
1604         };
1605
1606         ipv6_addr_copy(&cfg.fc_dst, pfx);
1607
1608         /* Prevent useless cloning on PtP SIT.
1609            This thing is done here expecting that the whole
1610            class of non-broadcast devices need not cloning.
1611          */
1612 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1613         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1614                 cfg.fc_flags |= RTF_NONEXTHOP;
1615 #endif
1616
1617         ip6_route_add(&cfg);
1618 }
1619
1620 /* Create "default" multicast route to the interface */
1621
1622 static void addrconf_add_mroute(struct net_device *dev)
1623 {
1624         struct fib6_config cfg = {
1625                 .fc_table = RT6_TABLE_LOCAL,
1626                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1627                 .fc_ifindex = dev->ifindex,
1628                 .fc_dst_len = 8,
1629                 .fc_flags = RTF_UP,
1630                 .fc_nlinfo.nl_net = dev_net(dev),
1631         };
1632
1633         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1634
1635         ip6_route_add(&cfg);
1636 }
1637
1638 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1639 static void sit_route_add(struct net_device *dev)
1640 {
1641         struct fib6_config cfg = {
1642                 .fc_table = RT6_TABLE_MAIN,
1643                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1644                 .fc_ifindex = dev->ifindex,
1645                 .fc_dst_len = 96,
1646                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1647                 .fc_nlinfo.nl_net = dev_net(dev),
1648         };
1649
1650         /* prefix length - 96 bits "::d.d.d.d" */
1651         ip6_route_add(&cfg);
1652 }
1653 #endif
1654
1655 static void addrconf_add_lroute(struct net_device *dev)
1656 {
1657         struct in6_addr addr;
1658
1659         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1660         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1661 }
1662
1663 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1664 {
1665         struct inet6_dev *idev;
1666
1667         ASSERT_RTNL();
1668
1669         if ((idev = ipv6_find_idev(dev)) == NULL)
1670                 return NULL;
1671
1672         /* Add default multicast route */
1673         addrconf_add_mroute(dev);
1674
1675         /* Add link local route */
1676         addrconf_add_lroute(dev);
1677         return idev;
1678 }
1679
1680 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1681 {
1682         struct prefix_info *pinfo;
1683         __u32 valid_lft;
1684         __u32 prefered_lft;
1685         int addr_type;
1686         unsigned long rt_expires;
1687         struct inet6_dev *in6_dev;
1688
1689         pinfo = (struct prefix_info *) opt;
1690
1691         if (len < sizeof(struct prefix_info)) {
1692                 ADBG(("addrconf: prefix option too short\n"));
1693                 return;
1694         }
1695
1696         /*
1697          *      Validation checks ([ADDRCONF], page 19)
1698          */
1699
1700         addr_type = ipv6_addr_type(&pinfo->prefix);
1701
1702         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1703                 return;
1704
1705         valid_lft = ntohl(pinfo->valid);
1706         prefered_lft = ntohl(pinfo->prefered);
1707
1708         if (prefered_lft > valid_lft) {
1709                 if (net_ratelimit())
1710                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1711                 return;
1712         }
1713
1714         in6_dev = in6_dev_get(dev);
1715
1716         if (in6_dev == NULL) {
1717                 if (net_ratelimit())
1718                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1719                 return;
1720         }
1721
1722         /*
1723          *      Two things going on here:
1724          *      1) Add routes for on-link prefixes
1725          *      2) Configure prefixes with the auto flag set
1726          */
1727
1728         /* Avoid arithmetic overflow. Really, we could
1729            save rt_expires in seconds, likely valid_lft,
1730            but it would require division in fib gc, that it
1731            not good.
1732          */
1733         if (valid_lft >= 0x7FFFFFFF/HZ)
1734                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1735         else
1736                 rt_expires = valid_lft * HZ;
1737
1738         /*
1739          * We convert this (in jiffies) to clock_t later.
1740          * Avoid arithmetic overflow there as well.
1741          * Overflow can happen only if HZ < USER_HZ.
1742          */
1743         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1744                 rt_expires = 0x7FFFFFFF / USER_HZ;
1745
1746         if (pinfo->onlink) {
1747                 struct rt6_info *rt;
1748                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1749                                 dev->ifindex, 1);
1750
1751                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1752                         if (rt->rt6i_flags&RTF_EXPIRES) {
1753                                 if (valid_lft == 0) {
1754                                         ip6_del_rt(rt);
1755                                         rt = NULL;
1756                                 } else {
1757                                         rt->rt6i_expires = jiffies + rt_expires;
1758                                 }
1759                         }
1760                 } else if (valid_lft) {
1761                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1762                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1763                 }
1764                 if (rt)
1765                         dst_release(&rt->u.dst);
1766         }
1767
1768         /* Try to figure out our local address for this prefix */
1769
1770         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1771                 struct inet6_ifaddr * ifp;
1772                 struct in6_addr addr;
1773                 int create = 0, update_lft = 0;
1774
1775                 if (pinfo->prefix_len == 64) {
1776                         memcpy(&addr, &pinfo->prefix, 8);
1777                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1778                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1779                                 in6_dev_put(in6_dev);
1780                                 return;
1781                         }
1782                         goto ok;
1783                 }
1784                 if (net_ratelimit())
1785                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1786                                pinfo->prefix_len);
1787                 in6_dev_put(in6_dev);
1788                 return;
1789
1790 ok:
1791
1792                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1793
1794                 if (ifp == NULL && valid_lft) {
1795                         int max_addresses = in6_dev->cnf.max_addresses;
1796                         u32 addr_flags = 0;
1797
1798 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1799                         if (in6_dev->cnf.optimistic_dad &&
1800                             !ipv6_devconf.forwarding)
1801                                 addr_flags = IFA_F_OPTIMISTIC;
1802 #endif
1803
1804                         /* Do not allow to create too much of autoconfigured
1805                          * addresses; this would be too easy way to crash kernel.
1806                          */
1807                         if (!max_addresses ||
1808                             ipv6_count_addresses(in6_dev) < max_addresses)
1809                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1810                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1811                                                     addr_flags);
1812
1813                         if (!ifp || IS_ERR(ifp)) {
1814                                 in6_dev_put(in6_dev);
1815                                 return;
1816                         }
1817
1818                         update_lft = create = 1;
1819                         ifp->cstamp = jiffies;
1820                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1821                 }
1822
1823                 if (ifp) {
1824                         int flags;
1825                         unsigned long now;
1826 #ifdef CONFIG_IPV6_PRIVACY
1827                         struct inet6_ifaddr *ift;
1828 #endif
1829                         u32 stored_lft;
1830
1831                         /* update lifetime (RFC2462 5.5.3 e) */
1832                         spin_lock(&ifp->lock);
1833                         now = jiffies;
1834                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1835                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1836                         else
1837                                 stored_lft = 0;
1838                         if (!update_lft && stored_lft) {
1839                                 if (valid_lft > MIN_VALID_LIFETIME ||
1840                                     valid_lft > stored_lft)
1841                                         update_lft = 1;
1842                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1843                                         /* valid_lft <= stored_lft is always true */
1844                                         /* XXX: IPsec */
1845                                         update_lft = 0;
1846                                 } else {
1847                                         valid_lft = MIN_VALID_LIFETIME;
1848                                         if (valid_lft < prefered_lft)
1849                                                 prefered_lft = valid_lft;
1850                                         update_lft = 1;
1851                                 }
1852                         }
1853
1854                         if (update_lft) {
1855                                 ifp->valid_lft = valid_lft;
1856                                 ifp->prefered_lft = prefered_lft;
1857                                 ifp->tstamp = now;
1858                                 flags = ifp->flags;
1859                                 ifp->flags &= ~IFA_F_DEPRECATED;
1860                                 spin_unlock(&ifp->lock);
1861
1862                                 if (!(flags&IFA_F_TENTATIVE))
1863                                         ipv6_ifa_notify(0, ifp);
1864                         } else
1865                                 spin_unlock(&ifp->lock);
1866
1867 #ifdef CONFIG_IPV6_PRIVACY
1868                         read_lock_bh(&in6_dev->lock);
1869                         /* update all temporary addresses in the list */
1870                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1871                                 /*
1872                                  * When adjusting the lifetimes of an existing
1873                                  * temporary address, only lower the lifetimes.
1874                                  * Implementations must not increase the
1875                                  * lifetimes of an existing temporary address
1876                                  * when processing a Prefix Information Option.
1877                                  */
1878                                 if (ifp != ift->ifpub)
1879                                         continue;
1880
1881                                 spin_lock(&ift->lock);
1882                                 flags = ift->flags;
1883                                 if (ift->valid_lft > valid_lft &&
1884                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1885                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1886                                 if (ift->prefered_lft > prefered_lft &&
1887                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1888                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1889                                 spin_unlock(&ift->lock);
1890                                 if (!(flags&IFA_F_TENTATIVE))
1891                                         ipv6_ifa_notify(0, ift);
1892                         }
1893
1894                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1895                                 /*
1896                                  * When a new public address is created as described in [ADDRCONF],
1897                                  * also create a new temporary address.
1898                                  */
1899                                 read_unlock_bh(&in6_dev->lock);
1900                                 ipv6_create_tempaddr(ifp, NULL);
1901                         } else {
1902                                 read_unlock_bh(&in6_dev->lock);
1903                         }
1904 #endif
1905                         in6_ifa_put(ifp);
1906                         addrconf_verify(0);
1907                 }
1908         }
1909         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1910         in6_dev_put(in6_dev);
1911 }
1912
1913 /*
1914  *      Set destination address.
1915  *      Special case for SIT interfaces where we create a new "virtual"
1916  *      device.
1917  */
1918 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1919 {
1920         struct in6_ifreq ireq;
1921         struct net_device *dev;
1922         int err = -EINVAL;
1923
1924         rtnl_lock();
1925
1926         err = -EFAULT;
1927         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1928                 goto err_exit;
1929
1930         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1931
1932         err = -ENODEV;
1933         if (dev == NULL)
1934                 goto err_exit;
1935
1936 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1937         if (dev->type == ARPHRD_SIT) {
1938                 struct ifreq ifr;
1939                 mm_segment_t    oldfs;
1940                 struct ip_tunnel_parm p;
1941
1942                 err = -EADDRNOTAVAIL;
1943                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1944                         goto err_exit;
1945
1946                 memset(&p, 0, sizeof(p));
1947                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1948                 p.iph.saddr = 0;
1949                 p.iph.version = 4;
1950                 p.iph.ihl = 5;
1951                 p.iph.protocol = IPPROTO_IPV6;
1952                 p.iph.ttl = 64;
1953                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1954
1955                 oldfs = get_fs(); set_fs(KERNEL_DS);
1956                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1957                 set_fs(oldfs);
1958
1959                 if (err == 0) {
1960                         err = -ENOBUFS;
1961                         dev = __dev_get_by_name(net, p.name);
1962                         if (!dev)
1963                                 goto err_exit;
1964                         err = dev_open(dev);
1965                 }
1966         }
1967 #endif
1968
1969 err_exit:
1970         rtnl_unlock();
1971         return err;
1972 }
1973
1974 /*
1975  *      Manual configuration of address on an interface
1976  */
1977 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
1978                           int plen, __u8 ifa_flags, __u32 prefered_lft,
1979                           __u32 valid_lft)
1980 {
1981         struct inet6_ifaddr *ifp;
1982         struct inet6_dev *idev;
1983         struct net_device *dev;
1984         int scope;
1985         u32 flags = RTF_EXPIRES;
1986
1987         ASSERT_RTNL();
1988
1989         /* check the lifetime */
1990         if (!valid_lft || prefered_lft > valid_lft)
1991                 return -EINVAL;
1992
1993         dev = __dev_get_by_index(net, ifindex);
1994         if (!dev)
1995                 return -ENODEV;
1996
1997         if ((idev = addrconf_add_dev(dev)) == NULL)
1998                 return -ENOBUFS;
1999
2000         scope = ipv6_addr_scope(pfx);
2001
2002         if (valid_lft == INFINITY_LIFE_TIME) {
2003                 ifa_flags |= IFA_F_PERMANENT;
2004                 flags = 0;
2005         } else if (valid_lft >= 0x7FFFFFFF/HZ)
2006                 valid_lft = 0x7FFFFFFF/HZ;
2007
2008         if (prefered_lft == 0)
2009                 ifa_flags |= IFA_F_DEPRECATED;
2010         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2011                  (prefered_lft != INFINITY_LIFE_TIME))
2012                 prefered_lft = 0x7FFFFFFF/HZ;
2013
2014         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2015
2016         if (!IS_ERR(ifp)) {
2017                 spin_lock_bh(&ifp->lock);
2018                 ifp->valid_lft = valid_lft;
2019                 ifp->prefered_lft = prefered_lft;
2020                 ifp->tstamp = jiffies;
2021                 spin_unlock_bh(&ifp->lock);
2022
2023                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2024                                       jiffies_to_clock_t(valid_lft * HZ), flags);
2025                 /*
2026                  * Note that section 3.1 of RFC 4429 indicates
2027                  * that the Optimistic flag should not be set for
2028                  * manually configured addresses
2029                  */
2030                 addrconf_dad_start(ifp, 0);
2031                 in6_ifa_put(ifp);
2032                 addrconf_verify(0);
2033                 return 0;
2034         }
2035
2036         return PTR_ERR(ifp);
2037 }
2038
2039 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2040                           int plen)
2041 {
2042         struct inet6_ifaddr *ifp;
2043         struct inet6_dev *idev;
2044         struct net_device *dev;
2045
2046         dev = __dev_get_by_index(net, ifindex);
2047         if (!dev)
2048                 return -ENODEV;
2049
2050         if ((idev = __in6_dev_get(dev)) == NULL)
2051                 return -ENXIO;
2052
2053         read_lock_bh(&idev->lock);
2054         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2055                 if (ifp->prefix_len == plen &&
2056                     ipv6_addr_equal(pfx, &ifp->addr)) {
2057                         in6_ifa_hold(ifp);
2058                         read_unlock_bh(&idev->lock);
2059
2060                         ipv6_del_addr(ifp);
2061
2062                         /* If the last address is deleted administratively,
2063                            disable IPv6 on this interface.
2064                          */
2065                         if (idev->addr_list == NULL)
2066                                 addrconf_ifdown(idev->dev, 1);
2067                         return 0;
2068                 }
2069         }
2070         read_unlock_bh(&idev->lock);
2071         return -EADDRNOTAVAIL;
2072 }
2073
2074
2075 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2076 {
2077         struct in6_ifreq ireq;
2078         int err;
2079
2080         if (!capable(CAP_NET_ADMIN))
2081                 return -EPERM;
2082
2083         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2084                 return -EFAULT;
2085
2086         rtnl_lock();
2087         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2088                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2089                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2090         rtnl_unlock();
2091         return err;
2092 }
2093
2094 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2095 {
2096         struct in6_ifreq ireq;
2097         int err;
2098
2099         if (!capable(CAP_NET_ADMIN))
2100                 return -EPERM;
2101
2102         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2103                 return -EFAULT;
2104
2105         rtnl_lock();
2106         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2107                              ireq.ifr6_prefixlen);
2108         rtnl_unlock();
2109         return err;
2110 }
2111
2112 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2113 static void sit_add_v4_addrs(struct inet6_dev *idev)
2114 {
2115         struct inet6_ifaddr * ifp;
2116         struct in6_addr addr;
2117         struct net_device *dev;
2118         struct net *net = dev_net(idev->dev);
2119         int scope;
2120
2121         ASSERT_RTNL();
2122
2123         memset(&addr, 0, sizeof(struct in6_addr));
2124         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2125
2126         if (idev->dev->flags&IFF_POINTOPOINT) {
2127                 addr.s6_addr32[0] = htonl(0xfe800000);
2128                 scope = IFA_LINK;
2129         } else {
2130                 scope = IPV6_ADDR_COMPATv4;
2131         }
2132
2133         if (addr.s6_addr32[3]) {
2134                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2135                 if (!IS_ERR(ifp)) {
2136                         spin_lock_bh(&ifp->lock);
2137                         ifp->flags &= ~IFA_F_TENTATIVE;
2138                         spin_unlock_bh(&ifp->lock);
2139                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2140                         in6_ifa_put(ifp);
2141                 }
2142                 return;
2143         }
2144
2145         for_each_netdev(net, dev) {
2146                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2147                 if (in_dev && (dev->flags & IFF_UP)) {
2148                         struct in_ifaddr * ifa;
2149
2150                         int flag = scope;
2151
2152                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2153                                 int plen;
2154
2155                                 addr.s6_addr32[3] = ifa->ifa_local;
2156
2157                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2158                                         continue;
2159                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2160                                         if (idev->dev->flags&IFF_POINTOPOINT)
2161                                                 continue;
2162                                         flag |= IFA_HOST;
2163                                 }
2164                                 if (idev->dev->flags&IFF_POINTOPOINT)
2165                                         plen = 64;
2166                                 else
2167                                         plen = 96;
2168
2169                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2170                                                     IFA_F_PERMANENT);
2171                                 if (!IS_ERR(ifp)) {
2172                                         spin_lock_bh(&ifp->lock);
2173                                         ifp->flags &= ~IFA_F_TENTATIVE;
2174                                         spin_unlock_bh(&ifp->lock);
2175                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2176                                         in6_ifa_put(ifp);
2177                                 }
2178                         }
2179                 }
2180         }
2181 }
2182 #endif
2183
2184 static void init_loopback(struct net_device *dev)
2185 {
2186         struct inet6_dev  *idev;
2187         struct inet6_ifaddr * ifp;
2188
2189         /* ::1 */
2190
2191         ASSERT_RTNL();
2192
2193         if ((idev = ipv6_find_idev(dev)) == NULL) {
2194                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2195                 return;
2196         }
2197
2198         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2199         if (!IS_ERR(ifp)) {
2200                 spin_lock_bh(&ifp->lock);
2201                 ifp->flags &= ~IFA_F_TENTATIVE;
2202                 spin_unlock_bh(&ifp->lock);
2203                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2204                 in6_ifa_put(ifp);
2205         }
2206 }
2207
2208 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2209 {
2210         struct inet6_ifaddr * ifp;
2211         u32 addr_flags = IFA_F_PERMANENT;
2212
2213 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2214         if (idev->cnf.optimistic_dad &&
2215             !ipv6_devconf.forwarding)
2216                 addr_flags |= IFA_F_OPTIMISTIC;
2217 #endif
2218
2219
2220         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2221         if (!IS_ERR(ifp)) {
2222                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2223                 addrconf_dad_start(ifp, 0);
2224                 in6_ifa_put(ifp);
2225         }
2226 }
2227
2228 static void addrconf_dev_config(struct net_device *dev)
2229 {
2230         struct in6_addr addr;
2231         struct inet6_dev    * idev;
2232
2233         ASSERT_RTNL();
2234
2235         if ((dev->type != ARPHRD_ETHER) &&
2236             (dev->type != ARPHRD_FDDI) &&
2237             (dev->type != ARPHRD_IEEE802_TR) &&
2238             (dev->type != ARPHRD_ARCNET) &&
2239             (dev->type != ARPHRD_INFINIBAND)) {
2240                 /* Alas, we support only Ethernet autoconfiguration. */
2241                 return;
2242         }
2243
2244         idev = addrconf_add_dev(dev);
2245         if (idev == NULL)
2246                 return;
2247
2248         memset(&addr, 0, sizeof(struct in6_addr));
2249         addr.s6_addr32[0] = htonl(0xFE800000);
2250
2251         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2252                 addrconf_add_linklocal(idev, &addr);
2253 }
2254
2255 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2256 static void addrconf_sit_config(struct net_device *dev)
2257 {
2258         struct inet6_dev *idev;
2259
2260         ASSERT_RTNL();
2261
2262         /*
2263          * Configure the tunnel with one of our IPv4
2264          * addresses... we should configure all of
2265          * our v4 addrs in the tunnel
2266          */
2267
2268         if ((idev = ipv6_find_idev(dev)) == NULL) {
2269                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2270                 return;
2271         }
2272
2273         if (dev->priv_flags & IFF_ISATAP) {
2274                 struct in6_addr addr;
2275
2276                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2277                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2278                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2279                         addrconf_add_linklocal(idev, &addr);
2280                 return;
2281         }
2282
2283         sit_add_v4_addrs(idev);
2284
2285         if (dev->flags&IFF_POINTOPOINT) {
2286                 addrconf_add_mroute(dev);
2287                 addrconf_add_lroute(dev);
2288         } else
2289                 sit_route_add(dev);
2290 }
2291 #endif
2292
2293 static inline int
2294 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2295 {
2296         struct in6_addr lladdr;
2297
2298         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2299                 addrconf_add_linklocal(idev, &lladdr);
2300                 return 0;
2301         }
2302         return -1;
2303 }
2304
2305 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2306 {
2307         struct net_device *link_dev;
2308         struct net *net = dev_net(idev->dev);
2309
2310         /* first try to inherit the link-local address from the link device */
2311         if (idev->dev->iflink &&
2312             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2313                 if (!ipv6_inherit_linklocal(idev, link_dev))
2314                         return;
2315         }
2316         /* then try to inherit it from any device */
2317         for_each_netdev(net, link_dev) {
2318                 if (!ipv6_inherit_linklocal(idev, link_dev))
2319                         return;
2320         }
2321         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2322 }
2323
2324 /*
2325  * Autoconfigure tunnel with a link-local address so routing protocols,
2326  * DHCPv6, MLD etc. can be run over the virtual link
2327  */
2328
2329 static void addrconf_ip6_tnl_config(struct net_device *dev)
2330 {
2331         struct inet6_dev *idev;
2332
2333         ASSERT_RTNL();
2334
2335         if ((idev = addrconf_add_dev(dev)) == NULL) {
2336                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2337                 return;
2338         }
2339         ip6_tnl_add_linklocal(idev);
2340 }
2341
2342 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2343                            void * data)
2344 {
2345         struct net_device *dev = (struct net_device *) data;
2346         struct inet6_dev *idev = __in6_dev_get(dev);
2347         int run_pending = 0;
2348         int err;
2349
2350         switch(event) {
2351         case NETDEV_REGISTER:
2352                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2353                         idev = ipv6_add_dev(dev);
2354                         if (!idev)
2355                                 return notifier_from_errno(-ENOMEM);
2356                 }
2357                 break;
2358         case NETDEV_UP:
2359         case NETDEV_CHANGE:
2360                 if (dev->flags & IFF_SLAVE)
2361                         break;
2362
2363                 if (event == NETDEV_UP) {
2364                         if (!addrconf_qdisc_ok(dev)) {
2365                                 /* device is not ready yet. */
2366                                 printk(KERN_INFO
2367                                         "ADDRCONF(NETDEV_UP): %s: "
2368                                         "link is not ready\n",
2369                                         dev->name);
2370                                 break;
2371                         }
2372
2373                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2374                                 idev = ipv6_add_dev(dev);
2375
2376                         if (idev)
2377                                 idev->if_flags |= IF_READY;
2378                 } else {
2379                         if (!addrconf_qdisc_ok(dev)) {
2380                                 /* device is still not ready. */
2381                                 break;
2382                         }
2383
2384                         if (idev) {
2385                                 if (idev->if_flags & IF_READY) {
2386                                         /* device is already configured. */
2387                                         break;
2388                                 }
2389                                 idev->if_flags |= IF_READY;
2390                         }
2391
2392                         printk(KERN_INFO
2393                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2394                                         "link becomes ready\n",
2395                                         dev->name);
2396
2397                         run_pending = 1;
2398                 }
2399
2400                 switch(dev->type) {
2401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2402                 case ARPHRD_SIT:
2403                         addrconf_sit_config(dev);
2404                         break;
2405 #endif
2406                 case ARPHRD_TUNNEL6:
2407                         addrconf_ip6_tnl_config(dev);
2408                         break;
2409                 case ARPHRD_LOOPBACK:
2410                         init_loopback(dev);
2411                         break;
2412
2413                 default:
2414                         addrconf_dev_config(dev);
2415                         break;
2416                 }
2417                 if (idev) {
2418                         if (run_pending)
2419                                 addrconf_dad_run(idev);
2420
2421                         /* If the MTU changed during the interface down, when the
2422                            interface up, the changed MTU must be reflected in the
2423                            idev as well as routers.
2424                          */
2425                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2426                                 rt6_mtu_change(dev, dev->mtu);
2427                                 idev->cnf.mtu6 = dev->mtu;
2428                         }
2429                         idev->tstamp = jiffies;
2430                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2431                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2432                            stop IPv6 on this interface.
2433                          */
2434                         if (dev->mtu < IPV6_MIN_MTU)
2435                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2436                 }
2437                 break;
2438
2439         case NETDEV_CHANGEMTU:
2440                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2441                         rt6_mtu_change(dev, dev->mtu);
2442                         idev->cnf.mtu6 = dev->mtu;
2443                         break;
2444                 }
2445
2446                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2447                         idev = ipv6_add_dev(dev);
2448                         if (idev)
2449                                 break;
2450                 }
2451
2452                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2453
2454         case NETDEV_DOWN:
2455         case NETDEV_UNREGISTER:
2456                 /*
2457                  *      Remove all addresses from this interface.
2458                  */
2459                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2460                 break;
2461
2462         case NETDEV_CHANGENAME:
2463                 if (idev) {
2464                         snmp6_unregister_dev(idev);
2465                         addrconf_sysctl_unregister(idev);
2466                         addrconf_sysctl_register(idev);
2467                         err = snmp6_register_dev(idev);
2468                         if (err)
2469                                 return notifier_from_errno(err);
2470                 }
2471                 break;
2472         }
2473
2474         return NOTIFY_OK;
2475 }
2476
2477 /*
2478  *      addrconf module should be notified of a device going up
2479  */
2480 static struct notifier_block ipv6_dev_notf = {
2481         .notifier_call = addrconf_notify,
2482         .priority = 0
2483 };
2484
2485 static int addrconf_ifdown(struct net_device *dev, int how)
2486 {
2487         struct inet6_dev *idev;
2488         struct inet6_ifaddr *ifa, **bifa;
2489         struct net *net = dev_net(dev);
2490         int i;
2491
2492         ASSERT_RTNL();
2493
2494         if (dev == init_net.loopback_dev && how == 1)
2495                 how = 0;
2496
2497         rt6_ifdown(net, dev);
2498         neigh_ifdown(&nd_tbl, dev);
2499
2500         idev = __in6_dev_get(dev);
2501         if (idev == NULL)
2502                 return -ENODEV;
2503
2504         /* Step 1: remove reference to ipv6 device from parent device.
2505                    Do not dev_put!
2506          */
2507         if (how == 1) {
2508                 idev->dead = 1;
2509
2510                 /* protected by rtnl_lock */
2511                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2512
2513                 /* Step 1.5: remove snmp6 entry */
2514                 snmp6_unregister_dev(idev);
2515
2516         }
2517
2518         /* Step 2: clear hash table */
2519         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2520                 bifa = &inet6_addr_lst[i];
2521
2522                 write_lock_bh(&addrconf_hash_lock);
2523                 while ((ifa = *bifa) != NULL) {
2524                         if (ifa->idev == idev) {
2525                                 *bifa = ifa->lst_next;
2526                                 ifa->lst_next = NULL;
2527                                 addrconf_del_timer(ifa);
2528                                 in6_ifa_put(ifa);
2529                                 continue;
2530                         }
2531                         bifa = &ifa->lst_next;
2532                 }
2533                 write_unlock_bh(&addrconf_hash_lock);
2534         }
2535
2536         write_lock_bh(&idev->lock);
2537
2538         /* Step 3: clear flags for stateless addrconf */
2539         if (how != 1)
2540                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2541
2542         /* Step 4: clear address list */
2543 #ifdef CONFIG_IPV6_PRIVACY
2544         if (how == 1 && del_timer(&idev->regen_timer))
2545                 in6_dev_put(idev);
2546
2547         /* clear tempaddr list */
2548         while ((ifa = idev->tempaddr_list) != NULL) {
2549                 idev->tempaddr_list = ifa->tmp_next;
2550                 ifa->tmp_next = NULL;
2551                 ifa->dead = 1;
2552                 write_unlock_bh(&idev->lock);
2553                 spin_lock_bh(&ifa->lock);
2554
2555                 if (ifa->ifpub) {
2556                         in6_ifa_put(ifa->ifpub);
2557                         ifa->ifpub = NULL;
2558                 }
2559                 spin_unlock_bh(&ifa->lock);
2560                 in6_ifa_put(ifa);
2561                 write_lock_bh(&idev->lock);
2562         }
2563 #endif
2564         while ((ifa = idev->addr_list) != NULL) {
2565                 idev->addr_list = ifa->if_next;
2566                 ifa->if_next = NULL;
2567                 ifa->dead = 1;
2568                 addrconf_del_timer(ifa);
2569                 write_unlock_bh(&idev->lock);
2570
2571                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2572                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2573                 in6_ifa_put(ifa);
2574
2575                 write_lock_bh(&idev->lock);
2576         }
2577         write_unlock_bh(&idev->lock);
2578
2579         /* Step 5: Discard multicast list */
2580
2581         if (how == 1)
2582                 ipv6_mc_destroy_dev(idev);
2583         else
2584                 ipv6_mc_down(idev);
2585
2586         idev->tstamp = jiffies;
2587
2588         /* Shot the device (if unregistered) */
2589
2590         if (how == 1) {
2591                 addrconf_sysctl_unregister(idev);
2592                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2593                 neigh_ifdown(&nd_tbl, dev);
2594                 in6_dev_put(idev);
2595         }
2596         return 0;
2597 }
2598
2599 static void addrconf_rs_timer(unsigned long data)
2600 {
2601         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2602
2603         if (ifp->idev->cnf.forwarding)
2604                 goto out;
2605
2606         if (ifp->idev->if_flags & IF_RA_RCVD) {
2607                 /*
2608                  *      Announcement received after solicitation
2609                  *      was sent
2610                  */
2611                 goto out;
2612         }
2613
2614         spin_lock(&ifp->lock);
2615         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2616                 struct in6_addr all_routers;
2617
2618                 /* The wait after the last probe can be shorter */
2619                 addrconf_mod_timer(ifp, AC_RS,
2620                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2621                                    ifp->idev->cnf.rtr_solicit_delay :
2622                                    ifp->idev->cnf.rtr_solicit_interval);
2623                 spin_unlock(&ifp->lock);
2624
2625                 ipv6_addr_all_routers(&all_routers);
2626
2627                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2628         } else {
2629                 spin_unlock(&ifp->lock);
2630                 /*
2631                  * Note: we do not support deprecated "all on-link"
2632                  * assumption any longer.
2633                  */
2634                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2635                        ifp->idev->dev->name);
2636         }
2637
2638 out:
2639         in6_ifa_put(ifp);
2640 }
2641
2642 /*
2643  *      Duplicate Address Detection
2644  */
2645 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2646 {
2647         unsigned long rand_num;
2648         struct inet6_dev *idev = ifp->idev;
2649
2650         if (ifp->flags & IFA_F_OPTIMISTIC)
2651                 rand_num = 0;
2652         else
2653                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2654
2655         ifp->probes = idev->cnf.dad_transmits;
2656         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2657 }
2658
2659 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2660 {
2661         struct inet6_dev *idev = ifp->idev;
2662         struct net_device *dev = idev->dev;
2663
2664         addrconf_join_solict(dev, &ifp->addr);
2665
2666         net_srandom(ifp->addr.s6_addr32[3]);
2667
2668         read_lock_bh(&idev->lock);
2669         if (ifp->dead)
2670                 goto out;
2671         spin_lock_bh(&ifp->lock);
2672
2673         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2674             !(ifp->flags&IFA_F_TENTATIVE) ||
2675             ifp->flags & IFA_F_NODAD) {
2676                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2677                 spin_unlock_bh(&ifp->lock);
2678                 read_unlock_bh(&idev->lock);
2679
2680                 addrconf_dad_completed(ifp);
2681                 return;
2682         }
2683
2684         if (!(idev->if_flags & IF_READY)) {
2685                 spin_unlock_bh(&ifp->lock);
2686                 read_unlock_bh(&idev->lock);
2687                 /*
2688                  * If the defice is not ready:
2689                  * - keep it tentative if it is a permanent address.
2690                  * - otherwise, kill it.
2691                  */
2692                 in6_ifa_hold(ifp);
2693                 addrconf_dad_stop(ifp);
2694                 return;
2695         }
2696
2697         /*
2698          * Optimistic nodes can start receiving
2699          * Frames right away
2700          */
2701         if(ifp->flags & IFA_F_OPTIMISTIC)
2702                 ip6_ins_rt(ifp->rt);
2703
2704         addrconf_dad_kick(ifp);
2705         spin_unlock_bh(&ifp->lock);
2706 out:
2707         read_unlock_bh(&idev->lock);
2708 }
2709
2710 static void addrconf_dad_timer(unsigned long data)
2711 {
2712         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2713         struct inet6_dev *idev = ifp->idev;
2714         struct in6_addr unspec;
2715         struct in6_addr mcaddr;
2716
2717         read_lock_bh(&idev->lock);
2718         if (idev->dead) {
2719                 read_unlock_bh(&idev->lock);
2720                 goto out;
2721         }
2722         spin_lock_bh(&ifp->lock);
2723         if (ifp->probes == 0) {
2724                 /*
2725                  * DAD was successful
2726                  */
2727
2728                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2729                 spin_unlock_bh(&ifp->lock);
2730                 read_unlock_bh(&idev->lock);
2731
2732                 addrconf_dad_completed(ifp);
2733
2734                 goto out;
2735         }
2736
2737         ifp->probes--;
2738         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2739         spin_unlock_bh(&ifp->lock);
2740         read_unlock_bh(&idev->lock);
2741
2742         /* send a neighbour solicitation for our addr */
2743         memset(&unspec, 0, sizeof(unspec));
2744         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2745         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2746 out:
2747         in6_ifa_put(ifp);
2748 }
2749
2750 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2751 {
2752         struct net_device *     dev = ifp->idev->dev;
2753
2754         /*
2755          *      Configure the address for reception. Now it is valid.
2756          */
2757
2758         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2759
2760         /* If added prefix is link local and forwarding is off,
2761            start sending router solicitations.
2762          */
2763
2764         if (ifp->idev->cnf.forwarding == 0 &&
2765             ifp->idev->cnf.rtr_solicits > 0 &&
2766             (dev->flags&IFF_LOOPBACK) == 0 &&
2767             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2768                 struct in6_addr all_routers;
2769
2770                 ipv6_addr_all_routers(&all_routers);
2771
2772                 /*
2773                  *      If a host as already performed a random delay
2774                  *      [...] as part of DAD [...] there is no need
2775                  *      to delay again before sending the first RS
2776                  */
2777                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2778
2779                 spin_lock_bh(&ifp->lock);
2780                 ifp->probes = 1;
2781                 ifp->idev->if_flags |= IF_RS_SENT;
2782                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2783                 spin_unlock_bh(&ifp->lock);
2784         }
2785 }
2786
2787 static void addrconf_dad_run(struct inet6_dev *idev) {
2788         struct inet6_ifaddr *ifp;
2789
2790         read_lock_bh(&idev->lock);
2791         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2792                 spin_lock_bh(&ifp->lock);
2793                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2794                         spin_unlock_bh(&ifp->lock);
2795                         continue;
2796                 }
2797                 spin_unlock_bh(&ifp->lock);
2798                 addrconf_dad_kick(ifp);
2799         }
2800         read_unlock_bh(&idev->lock);
2801 }
2802
2803 #ifdef CONFIG_PROC_FS
2804 struct if6_iter_state {
2805         struct seq_net_private p;
2806         int bucket;
2807 };
2808
2809 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2810 {
2811         struct inet6_ifaddr *ifa = NULL;
2812         struct if6_iter_state *state = seq->private;
2813         struct net *net = seq_file_net(seq);
2814
2815         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2816                 ifa = inet6_addr_lst[state->bucket];
2817
2818                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2819                         ifa = ifa->lst_next;
2820                 if (ifa)
2821                         break;
2822         }
2823         return ifa;
2824 }
2825
2826 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2827 {
2828         struct if6_iter_state *state = seq->private;
2829         struct net *net = seq_file_net(seq);
2830
2831         ifa = ifa->lst_next;
2832 try_again:
2833         if (ifa) {
2834                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2835                         ifa = ifa->lst_next;
2836                         goto try_again;
2837                 }
2838         }
2839
2840         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2841                 ifa = inet6_addr_lst[state->bucket];
2842                 goto try_again;
2843         }
2844
2845         return ifa;
2846 }
2847
2848 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2849 {
2850         struct inet6_ifaddr *ifa = if6_get_first(seq);
2851
2852         if (ifa)
2853                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2854                         --pos;
2855         return pos ? NULL : ifa;
2856 }
2857
2858 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2859         __acquires(addrconf_hash_lock)
2860 {
2861         read_lock_bh(&addrconf_hash_lock);
2862         return if6_get_idx(seq, *pos);
2863 }
2864
2865 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2866 {
2867         struct inet6_ifaddr *ifa;
2868
2869         ifa = if6_get_next(seq, v);
2870         ++*pos;
2871         return ifa;
2872 }
2873
2874 static void if6_seq_stop(struct seq_file *seq, void *v)
2875         __releases(addrconf_hash_lock)
2876 {
2877         read_unlock_bh(&addrconf_hash_lock);
2878 }
2879
2880 static int if6_seq_show(struct seq_file *seq, void *v)
2881 {
2882         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2883         seq_printf(seq,
2884                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2885                    NIP6(ifp->addr),
2886                    ifp->idev->dev->ifindex,
2887                    ifp->prefix_len,
2888                    ifp->scope,
2889                    ifp->flags,
2890                    ifp->idev->dev->name);
2891         return 0;
2892 }
2893
2894 static const struct seq_operations if6_seq_ops = {
2895         .start  = if6_seq_start,
2896         .next   = if6_seq_next,
2897         .show   = if6_seq_show,
2898         .stop   = if6_seq_stop,
2899 };
2900
2901 static int if6_seq_open(struct inode *inode, struct file *file)
2902 {
2903         return seq_open_net(inode, file, &if6_seq_ops,
2904                             sizeof(struct if6_iter_state));
2905 }
2906
2907 static const struct file_operations if6_fops = {
2908         .owner          = THIS_MODULE,
2909         .open           = if6_seq_open,
2910         .read           = seq_read,
2911         .llseek         = seq_lseek,
2912         .release        = seq_release_net,
2913 };
2914
2915 static int if6_proc_net_init(struct net *net)
2916 {
2917         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2918                 return -ENOMEM;
2919         return 0;
2920 }
2921
2922 static void if6_proc_net_exit(struct net *net)
2923 {
2924        proc_net_remove(net, "if_inet6");
2925 }
2926
2927 static struct pernet_operations if6_proc_net_ops = {
2928        .init = if6_proc_net_init,
2929        .exit = if6_proc_net_exit,
2930 };
2931
2932 int __init if6_proc_init(void)
2933 {
2934         return register_pernet_subsys(&if6_proc_net_ops);
2935 }
2936
2937 void if6_proc_exit(void)
2938 {
2939         unregister_pernet_subsys(&if6_proc_net_ops);
2940 }
2941 #endif  /* CONFIG_PROC_FS */
2942
2943 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2944 /* Check if address is a home address configured on any interface. */
2945 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2946 {
2947         int ret = 0;
2948         struct inet6_ifaddr * ifp;
2949         u8 hash = ipv6_addr_hash(addr);
2950         read_lock_bh(&addrconf_hash_lock);
2951         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2952                 if (!net_eq(dev_net(ifp->idev->dev), net))
2953                         continue;
2954                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2955                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2956                         ret = 1;
2957                         break;
2958                 }
2959         }
2960         read_unlock_bh(&addrconf_hash_lock);
2961         return ret;
2962 }
2963 #endif
2964
2965 /*
2966  *      Periodic address status verification
2967  */
2968
2969 static void addrconf_verify(unsigned long foo)
2970 {
2971         struct inet6_ifaddr *ifp;
2972         unsigned long now, next;
2973         int i;
2974
2975         spin_lock_bh(&addrconf_verify_lock);
2976         now = jiffies;
2977         next = now + ADDR_CHECK_FREQUENCY;
2978
2979         del_timer(&addr_chk_timer);
2980
2981         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2982
2983 restart:
2984                 read_lock(&addrconf_hash_lock);
2985                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2986                         unsigned long age;
2987 #ifdef CONFIG_IPV6_PRIVACY
2988                         unsigned long regen_advance;
2989 #endif
2990
2991                         if (ifp->flags & IFA_F_PERMANENT)
2992                                 continue;
2993
2994                         spin_lock(&ifp->lock);
2995                         age = (now - ifp->tstamp) / HZ;
2996
2997 #ifdef CONFIG_IPV6_PRIVACY
2998                         regen_advance = ifp->idev->cnf.regen_max_retry *
2999                                         ifp->idev->cnf.dad_transmits *
3000                                         ifp->idev->nd_parms->retrans_time / HZ;
3001 #endif
3002
3003                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3004                             age >= ifp->valid_lft) {
3005                                 spin_unlock(&ifp->lock);
3006                                 in6_ifa_hold(ifp);
3007                                 read_unlock(&addrconf_hash_lock);
3008                                 ipv6_del_addr(ifp);
3009                                 goto restart;
3010                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3011                                 spin_unlock(&ifp->lock);
3012                                 continue;
3013                         } else if (age >= ifp->prefered_lft) {
3014                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3015                                 int deprecate = 0;
3016
3017                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3018                                         deprecate = 1;
3019                                         ifp->flags |= IFA_F_DEPRECATED;
3020                                 }
3021
3022                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3023                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3024
3025                                 spin_unlock(&ifp->lock);
3026
3027                                 if (deprecate) {
3028                                         in6_ifa_hold(ifp);
3029                                         read_unlock(&addrconf_hash_lock);
3030
3031                                         ipv6_ifa_notify(0, ifp);
3032                                         in6_ifa_put(ifp);
3033                                         goto restart;
3034                                 }
3035 #ifdef CONFIG_IPV6_PRIVACY
3036                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3037                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3038                                 if (age >= ifp->prefered_lft - regen_advance) {
3039                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3040                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3041                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3042                                         if (!ifp->regen_count && ifpub) {
3043                                                 ifp->regen_count++;
3044                                                 in6_ifa_hold(ifp);
3045                                                 in6_ifa_hold(ifpub);
3046                                                 spin_unlock(&ifp->lock);
3047                                                 read_unlock(&addrconf_hash_lock);
3048                                                 spin_lock(&ifpub->lock);
3049                                                 ifpub->regen_count = 0;
3050                                                 spin_unlock(&ifpub->lock);
3051                                                 ipv6_create_tempaddr(ifpub, ifp);
3052                                                 in6_ifa_put(ifpub);
3053                                                 in6_ifa_put(ifp);
3054                                                 goto restart;
3055                                         }
3056                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3057                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3058                                 spin_unlock(&ifp->lock);
3059 #endif
3060                         } else {
3061                                 /* ifp->prefered_lft <= ifp->valid_lft */
3062                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3063                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3064                                 spin_unlock(&ifp->lock);
3065                         }
3066                 }
3067                 read_unlock(&addrconf_hash_lock);
3068         }
3069
3070         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3071         add_timer(&addr_chk_timer);
3072         spin_unlock_bh(&addrconf_verify_lock);
3073 }
3074
3075 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3076 {
3077         struct in6_addr *pfx = NULL;
3078
3079         if (addr)
3080                 pfx = nla_data(addr);
3081
3082         if (local) {
3083                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3084                         pfx = NULL;
3085                 else
3086                         pfx = nla_data(local);
3087         }
3088
3089         return pfx;
3090 }
3091
3092 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3093         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3094         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3095         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3096 };
3097
3098 static int
3099 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3100 {
3101         struct net *net = sock_net(skb->sk);
3102         struct ifaddrmsg *ifm;
3103         struct nlattr *tb[IFA_MAX+1];
3104         struct in6_addr *pfx;
3105         int err;
3106
3107         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3108         if (err < 0)
3109                 return err;
3110
3111         ifm = nlmsg_data(nlh);
3112         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3113         if (pfx == NULL)
3114                 return -EINVAL;
3115
3116         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3117 }
3118
3119 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3120                              u32 prefered_lft, u32 valid_lft)
3121 {
3122         u32 flags = RTF_EXPIRES;
3123
3124         if (!valid_lft || (prefered_lft > valid_lft))
3125                 return -EINVAL;
3126
3127         if (valid_lft == INFINITY_LIFE_TIME) {
3128                 ifa_flags |= IFA_F_PERMANENT;
3129                 flags = 0;
3130         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3131                 valid_lft = 0x7FFFFFFF/HZ;
3132
3133         if (prefered_lft == 0)
3134                 ifa_flags |= IFA_F_DEPRECATED;
3135         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3136                  (prefered_lft != INFINITY_LIFE_TIME))
3137                 prefered_lft = 0x7FFFFFFF/HZ;
3138
3139         spin_lock_bh(&ifp->lock);
3140         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3141         ifp->tstamp = jiffies;
3142         ifp->valid_lft = valid_lft;
3143         ifp->prefered_lft = prefered_lft;
3144
3145         spin_unlock_bh(&ifp->lock);
3146         if (!(ifp->flags&IFA_F_TENTATIVE))
3147                 ipv6_ifa_notify(0, ifp);
3148
3149         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3150                               jiffies_to_clock_t(valid_lft * HZ), flags);
3151         addrconf_verify(0);
3152
3153         return 0;
3154 }
3155
3156 static int
3157 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3158 {
3159         struct net *net = sock_net(skb->sk);
3160         struct ifaddrmsg *ifm;
3161         struct nlattr *tb[IFA_MAX+1];
3162         struct in6_addr *pfx;
3163         struct inet6_ifaddr *ifa;
3164         struct net_device *dev;
3165         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3166         u8 ifa_flags;
3167         int err;
3168
3169         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3170         if (err < 0)
3171                 return err;
3172
3173         ifm = nlmsg_data(nlh);
3174         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3175         if (pfx == NULL)
3176                 return -EINVAL;
3177
3178         if (tb[IFA_CACHEINFO]) {
3179                 struct ifa_cacheinfo *ci;
3180
3181                 ci = nla_data(tb[IFA_CACHEINFO]);
3182                 valid_lft = ci->ifa_valid;
3183                 preferred_lft = ci->ifa_prefered;
3184         } else {
3185                 preferred_lft = INFINITY_LIFE_TIME;
3186                 valid_lft = INFINITY_LIFE_TIME;
3187         }
3188
3189         dev =  __dev_get_by_index(net, ifm->ifa_index);
3190         if (dev == NULL)
3191                 return -ENODEV;
3192
3193         /* We ignore other flags so far. */
3194         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3195
3196         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3197         if (ifa == NULL) {
3198                 /*
3199                  * It would be best to check for !NLM_F_CREATE here but
3200                  * userspace alreay relies on not having to provide this.
3201                  */
3202                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3203                                       ifm->ifa_prefixlen, ifa_flags,
3204                                       preferred_lft, valid_lft);
3205         }
3206
3207         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3208             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3209                 err = -EEXIST;
3210         else
3211                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3212
3213         in6_ifa_put(ifa);
3214
3215         return err;
3216 }
3217
3218 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3219                           u8 scope, int ifindex)
3220 {
3221         struct ifaddrmsg *ifm;
3222
3223         ifm = nlmsg_data(nlh);
3224         ifm->ifa_family = AF_INET6;
3225         ifm->ifa_prefixlen = prefixlen;
3226         ifm->ifa_flags = flags;
3227         ifm->ifa_scope = scope;
3228         ifm->ifa_index = ifindex;
3229 }
3230
3231 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3232                          unsigned long tstamp, u32 preferred, u32 valid)
3233 {
3234         struct ifa_cacheinfo ci;
3235
3236         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3237                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3238         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3239                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3240         ci.ifa_prefered = preferred;
3241         ci.ifa_valid = valid;
3242
3243         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3244 }
3245
3246 static inline int rt_scope(int ifa_scope)
3247 {
3248         if (ifa_scope & IFA_HOST)
3249                 return RT_SCOPE_HOST;
3250         else if (ifa_scope & IFA_LINK)
3251                 return RT_SCOPE_LINK;
3252         else if (ifa_scope & IFA_SITE)
3253                 return RT_SCOPE_SITE;
3254         else
3255                 return RT_SCOPE_UNIVERSE;
3256 }
3257
3258 static inline int inet6_ifaddr_msgsize(void)
3259 {
3260         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3261                + nla_total_size(16) /* IFA_ADDRESS */
3262                + nla_total_size(sizeof(struct ifa_cacheinfo));
3263 }
3264
3265 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3266                              u32 pid, u32 seq, int event, unsigned int flags)
3267 {
3268         struct nlmsghdr  *nlh;
3269         u32 preferred, valid;
3270
3271         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3272         if (nlh == NULL)
3273                 return -EMSGSIZE;
3274
3275         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3276                       ifa->idev->dev->ifindex);
3277
3278         if (!(ifa->flags&IFA_F_PERMANENT)) {
3279                 preferred = ifa->prefered_lft;
3280                 valid = ifa->valid_lft;
3281                 if (preferred != INFINITY_LIFE_TIME) {
3282                         long tval = (jiffies - ifa->tstamp)/HZ;
3283                         preferred -= tval;
3284                         if (valid != INFINITY_LIFE_TIME)
3285                                 valid -= tval;
3286                 }
3287         } else {
3288                 preferred = INFINITY_LIFE_TIME;
3289                 valid = INFINITY_LIFE_TIME;
3290         }
3291
3292         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3293             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3294                 nlmsg_cancel(skb, nlh);
3295                 return -EMSGSIZE;
3296         }
3297
3298         return nlmsg_end(skb, nlh);
3299 }
3300
3301 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3302                                 u32 pid, u32 seq, int event, u16 flags)
3303 {
3304         struct nlmsghdr  *nlh;
3305         u8 scope = RT_SCOPE_UNIVERSE;
3306         int ifindex = ifmca->idev->dev->ifindex;
3307
3308         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3309                 scope = RT_SCOPE_SITE;
3310
3311         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3312         if (nlh == NULL)
3313                 return -EMSGSIZE;
3314
3315         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3316         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3317             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3318                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3319                 nlmsg_cancel(skb, nlh);
3320                 return -EMSGSIZE;
3321         }
3322
3323         return nlmsg_end(skb, nlh);
3324 }
3325
3326 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3327                                 u32 pid, u32 seq, int event, unsigned int flags)
3328 {
3329         struct nlmsghdr  *nlh;
3330         u8 scope = RT_SCOPE_UNIVERSE;
3331         int ifindex = ifaca->aca_idev->dev->ifindex;
3332
3333         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3334                 scope = RT_SCOPE_SITE;
3335
3336         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3337         if (nlh == NULL)
3338                 return -EMSGSIZE;
3339
3340         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3341         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3342             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3343                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3344                 nlmsg_cancel(skb, nlh);
3345                 return -EMSGSIZE;
3346         }
3347
3348         return nlmsg_end(skb, nlh);
3349 }
3350
3351 enum addr_type_t
3352 {
3353         UNICAST_ADDR,
3354         MULTICAST_ADDR,
3355         ANYCAST_ADDR,
3356 };
3357
3358 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3359                            enum addr_type_t type)
3360 {
3361         int idx, ip_idx;
3362         int s_idx, s_ip_idx;
3363         int err = 1;
3364         struct net_device *dev;
3365         struct inet6_dev *idev = NULL;
3366         struct inet6_ifaddr *ifa;
3367         struct ifmcaddr6 *ifmca;
3368         struct ifacaddr6 *ifaca;
3369         struct net *net = sock_net(skb->sk);
3370
3371         s_idx = cb->args[0];
3372         s_ip_idx = ip_idx = cb->args[1];
3373
3374         idx = 0;
3375         for_each_netdev(net, dev) {
3376                 if (idx < s_idx)
3377                         goto cont;
3378                 if (idx > s_idx)
3379                         s_ip_idx = 0;
3380                 ip_idx = 0;
3381                 if ((idev = in6_dev_get(dev)) == NULL)
3382                         goto cont;
3383                 read_lock_bh(&idev->lock);
3384                 switch (type) {
3385                 case UNICAST_ADDR:
3386                         /* unicast address incl. temp addr */
3387                         for (ifa = idev->addr_list; ifa;
3388                              ifa = ifa->if_next, ip_idx++) {
3389                                 if (ip_idx < s_ip_idx)
3390                                         continue;
3391                                 err = inet6_fill_ifaddr(skb, ifa,
3392                                                         NETLINK_CB(cb->skb).pid,
3393                                                         cb->nlh->nlmsg_seq,
3394                                                         RTM_NEWADDR,
3395                                                         NLM_F_MULTI);
3396                         }
3397                         break;
3398                 case MULTICAST_ADDR:
3399                         /* multicast address */
3400                         for (ifmca = idev->mc_list; ifmca;
3401                              ifmca = ifmca->next, ip_idx++) {
3402                                 if (ip_idx < s_ip_idx)
3403                                         continue;
3404                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3405                                                           NETLINK_CB(cb->skb).pid,
3406                                                           cb->nlh->nlmsg_seq,
3407                                                           RTM_GETMULTICAST,
3408                                                           NLM_F_MULTI);
3409                         }
3410                         break;
3411                 case ANYCAST_ADDR:
3412                         /* anycast address */
3413                         for (ifaca = idev->ac_list; ifaca;
3414                              ifaca = ifaca->aca_next, ip_idx++) {
3415                                 if (ip_idx < s_ip_idx)
3416                                         continue;
3417                                 err = inet6_fill_ifacaddr(skb, ifaca,
3418                                                           NETLINK_CB(cb->skb).pid,
3419                                                           cb->nlh->nlmsg_seq,
3420                                                           RTM_GETANYCAST,
3421                                                           NLM_F_MULTI);
3422                         }
3423                         break;
3424                 default:
3425                         break;
3426                 }
3427                 read_unlock_bh(&idev->lock);
3428                 in6_dev_put(idev);
3429
3430                 if (err <= 0)
3431                         break;
3432 cont:
3433                 idx++;
3434         }
3435         cb->args[0] = idx;
3436         cb->args[1] = ip_idx;
3437         return skb->len;
3438 }
3439
3440 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3441 {
3442         enum addr_type_t type = UNICAST_ADDR;
3443
3444         return inet6_dump_addr(skb, cb, type);
3445 }
3446
3447 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3448 {
3449         enum addr_type_t type = MULTICAST_ADDR;
3450
3451         return inet6_dump_addr(skb, cb, type);
3452 }
3453
3454
3455 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3456 {
3457         enum addr_type_t type = ANYCAST_ADDR;
3458
3459         return inet6_dump_addr(skb, cb, type);
3460 }
3461
3462 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3463                              void *arg)
3464 {
3465         struct net *net = sock_net(in_skb->sk);
3466         struct ifaddrmsg *ifm;
3467         struct nlattr *tb[IFA_MAX+1];
3468         struct in6_addr *addr = NULL;
3469         struct net_device *dev = NULL;
3470         struct inet6_ifaddr *ifa;
3471         struct sk_buff *skb;
3472         int err;
3473
3474         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3475         if (err < 0)
3476                 goto errout;
3477
3478         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3479         if (addr == NULL) {
3480                 err = -EINVAL;
3481                 goto errout;
3482         }
3483
3484         ifm = nlmsg_data(nlh);
3485         if (ifm->ifa_index)
3486                 dev = __dev_get_by_index(net, ifm->ifa_index);
3487
3488         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3489                 err = -EADDRNOTAVAIL;
3490                 goto errout;
3491         }
3492
3493         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3494                 err = -ENOBUFS;
3495                 goto errout_ifa;
3496         }
3497
3498         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3499                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3500         if (err < 0) {
3501                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3502                 WARN_ON(err == -EMSGSIZE);
3503                 kfree_skb(skb);
3504                 goto errout_ifa;
3505         }
3506         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3507 errout_ifa:
3508         in6_ifa_put(ifa);
3509 errout:
3510         return err;
3511 }
3512
3513 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3514 {
3515         struct sk_buff *skb;
3516         struct net *net = dev_net(ifa->idev->dev);
3517         int err = -ENOBUFS;
3518
3519         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3520         if (skb == NULL)
3521                 goto errout;
3522
3523         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3524         if (err < 0) {
3525                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3526                 WARN_ON(err == -EMSGSIZE);
3527                 kfree_skb(skb);
3528                 goto errout;
3529         }
3530         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3531 errout:
3532         if (err < 0)
3533                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3534 }
3535
3536 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3537                                 __s32 *array, int bytes)
3538 {
3539         BUG_ON(bytes < (DEVCONF_MAX * 4));
3540
3541         memset(array, 0, bytes);
3542         array[DEVCONF_FORWARDING] = cnf->forwarding;
3543         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3544         array[DEVCONF_MTU6] = cnf->mtu6;
3545         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3546         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3547         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3548         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3549         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3550         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3551         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3552         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3553 #ifdef CONFIG_IPV6_PRIVACY
3554         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3555         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3556         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3557         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3558         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3559 #endif
3560         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3561         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3562         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3563 #ifdef CONFIG_IPV6_ROUTER_PREF
3564         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3565         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3566 #ifdef CONFIG_IPV6_ROUTE_INFO
3567         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3568 #endif
3569 #endif
3570         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3571         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3572 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3573         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3574 #endif
3575 #ifdef CONFIG_IPV6_MROUTE
3576         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3577 #endif
3578 }
3579
3580 static inline size_t inet6_if_nlmsg_size(void)
3581 {
3582         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3583                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3584                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3585                + nla_total_size(4) /* IFLA_MTU */
3586                + nla_total_size(4) /* IFLA_LINK */
3587                + nla_total_size( /* IFLA_PROTINFO */
3588                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3589                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3590                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3591                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3592                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3593                  );
3594 }
3595
3596 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3597                                       int bytes)
3598 {
3599         int i;
3600         int pad = bytes - sizeof(u64) * items;
3601         BUG_ON(pad < 0);
3602
3603         /* Use put_unaligned() because stats may not be aligned for u64. */
3604         put_unaligned(items, &stats[0]);
3605         for (i = 1; i < items; i++)
3606                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3607
3608         memset(&stats[items], 0, pad);
3609 }
3610
3611 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3612                              int bytes)
3613 {
3614         switch(attrtype) {
3615         case IFLA_INET6_STATS:
3616                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3617                 break;
3618         case IFLA_INET6_ICMP6STATS:
3619                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3620                 break;
3621         }
3622 }
3623
3624 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3625                              u32 pid, u32 seq, int event, unsigned int flags)
3626 {
3627         struct net_device *dev = idev->dev;
3628         struct nlattr *nla;
3629         struct ifinfomsg *hdr;
3630         struct nlmsghdr *nlh;
3631         void *protoinfo;
3632         struct ifla_cacheinfo ci;
3633
3634         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3635         if (nlh == NULL)
3636                 return -EMSGSIZE;
3637
3638         hdr = nlmsg_data(nlh);
3639         hdr->ifi_family = AF_INET6;
3640         hdr->__ifi_pad = 0;
3641         hdr->ifi_type = dev->type;
3642         hdr->ifi_index = dev->ifindex;
3643         hdr->ifi_flags = dev_get_flags(dev);
3644         hdr->ifi_change = 0;
3645
3646         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3647
3648         if (dev->addr_len)
3649                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3650
3651         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3652         if (dev->ifindex != dev->iflink)
3653                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3654
3655         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3656         if (protoinfo == NULL)
3657                 goto nla_put_failure;
3658
3659         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3660
3661         ci.max_reasm_len = IPV6_MAXPLEN;
3662         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3663                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3664         ci.reachable_time = idev->nd_parms->reachable_time;
3665         ci.retrans_time = idev->nd_parms->retrans_time;
3666         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3667
3668         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3669         if (nla == NULL)
3670                 goto nla_put_failure;
3671         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3672
3673         /* XXX - MC not implemented */
3674
3675         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3676         if (nla == NULL)
3677                 goto nla_put_failure;
3678         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3679
3680         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3681         if (nla == NULL)
3682                 goto nla_put_failure;
3683         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3684
3685         nla_nest_end(skb, protoinfo);
3686         return nlmsg_end(skb, nlh);
3687
3688 nla_put_failure:
3689         nlmsg_cancel(skb, nlh);
3690         return -EMSGSIZE;
3691 }
3692
3693 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3694 {
3695         struct net *net = sock_net(skb->sk);
3696         int idx, err;
3697         int s_idx = cb->args[0];
3698         struct net_device *dev;
3699         struct inet6_dev *idev;
3700
3701         read_lock(&dev_base_lock);
3702         idx = 0;
3703         for_each_netdev(net, dev) {
3704                 if (idx < s_idx)
3705                         goto cont;
3706                 if ((idev = in6_dev_get(dev)) == NULL)
3707                         goto cont;
3708                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3709                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3710                 in6_dev_put(idev);
3711                 if (err <= 0)
3712                         break;
3713 cont:
3714                 idx++;
3715         }
3716         read_unlock(&dev_base_lock);
3717         cb->args[0] = idx;
3718
3719         return skb->len;
3720 }
3721
3722 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3723 {
3724         struct sk_buff *skb;
3725         struct net *net = dev_net(idev->dev);
3726         int err = -ENOBUFS;
3727
3728         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3729         if (skb == NULL)
3730                 goto errout;
3731
3732         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3733         if (err < 0) {
3734                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3735                 WARN_ON(err == -EMSGSIZE);
3736                 kfree_skb(skb);
3737                 goto errout;
3738         }
3739         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3740 errout:
3741         if (err < 0)
3742                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3743 }
3744
3745 static inline size_t inet6_prefix_nlmsg_size(void)
3746 {
3747         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3748                + nla_total_size(sizeof(struct in6_addr))
3749                + nla_total_size(sizeof(struct prefix_cacheinfo));
3750 }
3751
3752 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3753                              struct prefix_info *pinfo, u32 pid, u32 seq,
3754                              int event, unsigned int flags)
3755 {
3756         struct prefixmsg *pmsg;
3757         struct nlmsghdr *nlh;
3758         struct prefix_cacheinfo ci;
3759
3760         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3761         if (nlh == NULL)
3762                 return -EMSGSIZE;
3763
3764         pmsg = nlmsg_data(nlh);
3765         pmsg->prefix_family = AF_INET6;
3766         pmsg->prefix_pad1 = 0;
3767         pmsg->prefix_pad2 = 0;
3768         pmsg->prefix_ifindex = idev->dev->ifindex;
3769         pmsg->prefix_len = pinfo->prefix_len;
3770         pmsg->prefix_type = pinfo->type;
3771         pmsg->prefix_pad3 = 0;
3772         pmsg->prefix_flags = 0;
3773         if (pinfo->onlink)
3774                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3775         if (pinfo->autoconf)
3776                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3777
3778         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3779
3780         ci.preferred_time = ntohl(pinfo->prefered);
3781         ci.valid_time = ntohl(pinfo->valid);
3782         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3783
3784         return nlmsg_end(skb, nlh);
3785
3786 nla_put_failure:
3787         nlmsg_cancel(skb, nlh);
3788         return -EMSGSIZE;
3789 }
3790
3791 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3792                          struct prefix_info *pinfo)
3793 {
3794         struct sk_buff *skb;
3795         struct net *net = dev_net(idev->dev);
3796         int err = -ENOBUFS;
3797
3798         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3799         if (skb == NULL)
3800                 goto errout;
3801
3802         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3803         if (err < 0) {
3804                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3805                 WARN_ON(err == -EMSGSIZE);
3806                 kfree_skb(skb);
3807                 goto errout;
3808         }
3809         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3810 errout:
3811         if (err < 0)
3812                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3813 }
3814
3815 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3816 {
3817         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3818
3819         switch (event) {
3820         case RTM_NEWADDR:
3821                 /*
3822                  * If the address was optimistic
3823                  * we inserted the route at the start of
3824                  * our DAD process, so we don't need
3825                  * to do it again
3826                  */
3827                 if (!(ifp->rt->rt6i_node))
3828                         ip6_ins_rt(ifp->rt);
3829                 if (ifp->idev->cnf.forwarding)
3830                         addrconf_join_anycast(ifp);
3831                 break;
3832         case RTM_DELADDR:
3833                 if (ifp->idev->cnf.forwarding)
3834                         addrconf_leave_anycast(ifp);
3835                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3836                 dst_hold(&ifp->rt->u.dst);
3837                 if (ip6_del_rt(ifp->rt))
3838                         dst_free(&ifp->rt->u.dst);
3839                 break;
3840         }
3841 }
3842
3843 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3844 {
3845         rcu_read_lock_bh();
3846         if (likely(ifp->idev->dead == 0))
3847                 __ipv6_ifa_notify(event, ifp);
3848         rcu_read_unlock_bh();
3849 }
3850
3851 #ifdef CONFIG_SYSCTL
3852
3853 static
3854 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3855                            void __user *buffer, size_t *lenp, loff_t *ppos)
3856 {
3857         int *valp = ctl->data;
3858         int val = *valp;
3859         int ret;
3860
3861         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3862
3863         if (write)
3864                 addrconf_fixup_forwarding(ctl, valp, val);
3865         return ret;
3866 }
3867
3868 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3869                                             int __user *name, int nlen,
3870                                             void __user *oldval,
3871                                             size_t __user *oldlenp,
3872                                             void __user *newval, size_t newlen)
3873 {
3874         int *valp = table->data;
3875         int val = *valp;
3876         int new;
3877
3878         if (!newval || !newlen)
3879                 return 0;
3880         if (newlen != sizeof(int))
3881                 return -EINVAL;
3882         if (get_user(new, (int __user *)newval))
3883                 return -EFAULT;
3884         if (new == *valp)
3885                 return 0;
3886         if (oldval && oldlenp) {
3887                 size_t len;
3888                 if (get_user(len, oldlenp))
3889                         return -EFAULT;
3890                 if (len) {
3891                         if (len > table->maxlen)
3892                                 len = table->maxlen;
3893                         if (copy_to_user(oldval, valp, len))
3894                                 return -EFAULT;
3895                         if (put_user(len, oldlenp))
3896                                 return -EFAULT;
3897                 }
3898         }
3899
3900         *valp = new;
3901         addrconf_fixup_forwarding(table, valp, val);
3902         return 1;
3903 }
3904
3905 static struct addrconf_sysctl_table
3906 {
3907         struct ctl_table_header *sysctl_header;
3908         ctl_table addrconf_vars[DEVCONF_MAX+1];
3909         char *dev_name;
3910 } addrconf_sysctl __read_mostly = {
3911         .sysctl_header = NULL,
3912         .addrconf_vars = {
3913                 {
3914                         .ctl_name       =       NET_IPV6_FORWARDING,
3915                         .procname       =       "forwarding",
3916                         .data           =       &ipv6_devconf.forwarding,
3917                         .maxlen         =       sizeof(int),
3918                         .mode           =       0644,
3919                         .proc_handler   =       &addrconf_sysctl_forward,
3920                         .strategy       =       &addrconf_sysctl_forward_strategy,
3921                 },
3922                 {
3923                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3924                         .procname       =       "hop_limit",
3925                         .data           =       &ipv6_devconf.hop_limit,
3926                         .maxlen         =       sizeof(int),
3927                         .mode           =       0644,
3928                         .proc_handler   =       proc_dointvec,
3929                 },
3930                 {
3931                         .ctl_name       =       NET_IPV6_MTU,
3932                         .procname       =       "mtu",
3933                         .data           =       &ipv6_devconf.mtu6,
3934                         .maxlen         =       sizeof(int),
3935                         .mode           =       0644,
3936                         .proc_handler   =       &proc_dointvec,
3937                 },
3938                 {
3939                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3940                         .procname       =       "accept_ra",
3941                         .data           =       &ipv6_devconf.accept_ra,
3942                         .maxlen         =       sizeof(int),
3943                         .mode           =       0644,
3944                         .proc_handler   =       &proc_dointvec,
3945                 },
3946                 {
3947                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3948                         .procname       =       "accept_redirects",
3949                         .data           =       &ipv6_devconf.accept_redirects,
3950                         .maxlen         =       sizeof(int),
3951                         .mode           =       0644,
3952                         .proc_handler   =       &proc_dointvec,
3953                 },
3954                 {
3955                         .ctl_name       =       NET_IPV6_AUTOCONF,
3956                         .procname       =       "autoconf",
3957                         .data           =       &ipv6_devconf.autoconf,
3958                         .maxlen         =       sizeof(int),
3959                         .mode           =       0644,
3960                         .proc_handler   =       &proc_dointvec,
3961                 },
3962                 {
3963                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3964                         .procname       =       "dad_transmits",
3965                         .data           =       &ipv6_devconf.dad_transmits,
3966                         .maxlen         =       sizeof(int),
3967                         .mode           =       0644,
3968                         .proc_handler   =       &proc_dointvec,
3969                 },
3970                 {
3971                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3972                         .procname       =       "router_solicitations",
3973                         .data           =       &ipv6_devconf.rtr_solicits,
3974                         .maxlen         =       sizeof(int),
3975                         .mode           =       0644,
3976                         .proc_handler   =       &proc_dointvec,
3977                 },
3978                 {
3979                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3980                         .procname       =       "router_solicitation_interval",
3981                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3982                         .maxlen         =       sizeof(int),
3983                         .mode           =       0644,
3984                         .proc_handler   =       &proc_dointvec_jiffies,
3985                         .strategy       =       &sysctl_jiffies,
3986                 },
3987                 {
3988                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3989                         .procname       =       "router_solicitation_delay",
3990                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3991                         .maxlen         =       sizeof(int),
3992                         .mode           =       0644,
3993                         .proc_handler   =       &proc_dointvec_jiffies,
3994                         .strategy       =       &sysctl_jiffies,
3995                 },
3996                 {
3997                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3998                         .procname       =       "force_mld_version",
3999                         .data           =       &ipv6_devconf.force_mld_version,
4000                         .maxlen         =       sizeof(int),
4001                         .mode           =       0644,
4002                         .proc_handler   =       &proc_dointvec,
4003                 },
4004 #ifdef CONFIG_IPV6_PRIVACY
4005                 {
4006                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4007                         .procname       =       "use_tempaddr",
4008                         .data           =       &ipv6_devconf.use_tempaddr,
4009                         .maxlen         =       sizeof(int),
4010                         .mode           =       0644,
4011                         .proc_handler   =       &proc_dointvec,
4012                 },
4013                 {
4014                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4015                         .procname       =       "temp_valid_lft",
4016                         .data           =       &ipv6_devconf.temp_valid_lft,
4017                         .maxlen         =       sizeof(int),
4018                         .mode           =       0644,
4019                         .proc_handler   =       &proc_dointvec,
4020                 },
4021                 {
4022                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4023                         .procname       =       "temp_prefered_lft",
4024                         .data           =       &ipv6_devconf.temp_prefered_lft,
4025                         .maxlen         =       sizeof(int),
4026                         .mode           =       0644,
4027                         .proc_handler   =       &proc_dointvec,
4028                 },
4029                 {
4030                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4031                         .procname       =       "regen_max_retry",
4032                         .data           =       &ipv6_devconf.regen_max_retry,
4033                         .maxlen         =       sizeof(int),
4034                         .mode           =       0644,
4035                         .proc_handler   =       &proc_dointvec,
4036                 },
4037                 {
4038                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4039                         .procname       =       "max_desync_factor",
4040                         .data           =       &ipv6_devconf.max_desync_factor,
4041                         .maxlen         =       sizeof(int),
4042                         .mode           =       0644,
4043                         .proc_handler   =       &proc_dointvec,
4044                 },
4045 #endif
4046                 {
4047                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4048                         .procname       =       "max_addresses",
4049                         .data           =       &ipv6_devconf.max_addresses,
4050                         .maxlen         =       sizeof(int),
4051                         .mode           =       0644,
4052                         .proc_handler   =       &proc_dointvec,
4053                 },
4054                 {
4055                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4056                         .procname       =       "accept_ra_defrtr",
4057                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4058                         .maxlen         =       sizeof(int),
4059                         .mode           =       0644,
4060                         .proc_handler   =       &proc_dointvec,
4061                 },
4062                 {
4063                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4064                         .procname       =       "accept_ra_pinfo",
4065                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4066                         .maxlen         =       sizeof(int),
4067                         .mode           =       0644,
4068                         .proc_handler   =       &proc_dointvec,
4069                 },
4070 #ifdef CONFIG_IPV6_ROUTER_PREF
4071                 {
4072                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4073                         .procname       =       "accept_ra_rtr_pref",
4074                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4075                         .maxlen         =       sizeof(int),
4076                         .mode           =       0644,
4077                         .proc_handler   =       &proc_dointvec,
4078                 },
4079                 {
4080                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4081                         .procname       =       "router_probe_interval",
4082                         .data           =       &ipv6_devconf.rtr_probe_interval,
4083                         .maxlen         =       sizeof(int),
4084                         .mode           =       0644,
4085                         .proc_handler   =       &proc_dointvec_jiffies,
4086                         .strategy       =       &sysctl_jiffies,
4087                 },
4088 #ifdef CONFIG_IPV6_ROUTE_INFO
4089                 {
4090                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4091                         .procname       =       "accept_ra_rt_info_max_plen",
4092                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4093                         .maxlen         =       sizeof(int),
4094                         .mode           =       0644,
4095                         .proc_handler   =       &proc_dointvec,
4096                 },
4097 #endif
4098 #endif
4099                 {
4100                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4101                         .procname       =       "proxy_ndp",
4102                         .data           =       &ipv6_devconf.proxy_ndp,
4103                         .maxlen         =       sizeof(int),
4104                         .mode           =       0644,
4105                         .proc_handler   =       &proc_dointvec,
4106                 },
4107                 {
4108                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4109                         .procname       =       "accept_source_route",
4110                         .data           =       &ipv6_devconf.accept_source_route,
4111                         .maxlen         =       sizeof(int),
4112                         .mode           =       0644,
4113                         .proc_handler   =       &proc_dointvec,
4114                 },
4115 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4116                 {
4117                         .ctl_name       =       CTL_UNNUMBERED,
4118                         .procname       =       "optimistic_dad",
4119                         .data           =       &ipv6_devconf.optimistic_dad,
4120                         .maxlen         =       sizeof(int),
4121                         .mode           =       0644,
4122                         .proc_handler   =       &proc_dointvec,
4123
4124                 },
4125 #endif
4126 #ifdef CONFIG_IPV6_MROUTE
4127                 {
4128                         .ctl_name       =       CTL_UNNUMBERED,
4129                         .procname       =       "mc_forwarding",
4130                         .data           =       &ipv6_devconf.mc_forwarding,
4131                         .maxlen         =       sizeof(int),
4132                         .mode           =       0644,
4133                         .proc_handler   =       &proc_dointvec,
4134                 },
4135 #endif
4136                 {
4137                         .ctl_name       =       0,      /* sentinel */
4138                 }
4139         },
4140 };
4141
4142 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4143                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4144 {
4145         int i;
4146         struct addrconf_sysctl_table *t;
4147
4148 #define ADDRCONF_CTL_PATH_DEV   3
4149
4150         struct ctl_path addrconf_ctl_path[] = {
4151                 { .procname = "net", .ctl_name = CTL_NET, },
4152                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4153                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4154                 { /* to be set */ },
4155                 { },
4156         };
4157
4158
4159         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4160         if (t == NULL)
4161                 goto out;
4162
4163         for (i=0; t->addrconf_vars[i].data; i++) {
4164                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4165                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4166                 t->addrconf_vars[i].extra2 = net;
4167         }
4168
4169         /*
4170          * Make a copy of dev_name, because '.procname' is regarded as const
4171          * by sysctl and we wouldn't want anyone to change it under our feet
4172          * (see SIOCSIFNAME).
4173          */
4174         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4175         if (!t->dev_name)
4176                 goto free;
4177
4178         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4179         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4180
4181         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4182                         t->addrconf_vars);
4183         if (t->sysctl_header == NULL)
4184                 goto free_procname;
4185
4186         p->sysctl = t;
4187         return 0;
4188
4189 free_procname:
4190         kfree(t->dev_name);
4191 free:
4192         kfree(t);
4193 out:
4194         return -ENOBUFS;
4195 }
4196
4197 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4198 {
4199         struct addrconf_sysctl_table *t;
4200
4201         if (p->sysctl == NULL)
4202                 return;
4203
4204         t = p->sysctl;
4205         p->sysctl = NULL;
4206         unregister_sysctl_table(t->sysctl_header);
4207         kfree(t->dev_name);
4208         kfree(t);
4209 }
4210
4211 static void addrconf_sysctl_register(struct inet6_dev *idev)
4212 {
4213         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4214                               NET_IPV6_NEIGH, "ipv6",
4215                               &ndisc_ifinfo_sysctl_change,
4216                               NULL);
4217         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4218                         idev->dev->ifindex, idev, &idev->cnf);
4219 }
4220
4221 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4222 {
4223         __addrconf_sysctl_unregister(&idev->cnf);
4224         neigh_sysctl_unregister(idev->nd_parms);
4225 }
4226
4227
4228 #endif
4229
4230 static int addrconf_init_net(struct net *net)
4231 {
4232         int err;
4233         struct ipv6_devconf *all, *dflt;
4234
4235         err = -ENOMEM;
4236         all = &ipv6_devconf;
4237         dflt = &ipv6_devconf_dflt;
4238
4239         if (net != &init_net) {
4240                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4241                 if (all == NULL)
4242                         goto err_alloc_all;
4243
4244                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4245                 if (dflt == NULL)
4246                         goto err_alloc_dflt;
4247         }
4248
4249         net->ipv6.devconf_all = all;
4250         net->ipv6.devconf_dflt = dflt;
4251
4252 #ifdef CONFIG_SYSCTL
4253         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4254                         NULL, all);
4255         if (err < 0)
4256                 goto err_reg_all;
4257
4258         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4259                         NULL, dflt);
4260         if (err < 0)
4261                 goto err_reg_dflt;
4262 #endif
4263         return 0;
4264
4265 #ifdef CONFIG_SYSCTL
4266 err_reg_dflt:
4267         __addrconf_sysctl_unregister(all);
4268 err_reg_all:
4269         kfree(dflt);
4270 #endif
4271 err_alloc_dflt:
4272         kfree(all);
4273 err_alloc_all:
4274         return err;
4275 }
4276
4277 static void addrconf_exit_net(struct net *net)
4278 {
4279 #ifdef CONFIG_SYSCTL
4280         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4281         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4282 #endif
4283         if (net != &init_net) {
4284                 kfree(net->ipv6.devconf_dflt);
4285                 kfree(net->ipv6.devconf_all);
4286         }
4287 }
4288
4289 static struct pernet_operations addrconf_ops = {
4290         .init = addrconf_init_net,
4291         .exit = addrconf_exit_net,
4292 };
4293
4294 /*
4295  *      Device notifier
4296  */
4297
4298 int register_inet6addr_notifier(struct notifier_block *nb)
4299 {
4300         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4301 }
4302
4303 EXPORT_SYMBOL(register_inet6addr_notifier);
4304
4305 int unregister_inet6addr_notifier(struct notifier_block *nb)
4306 {
4307         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4308 }
4309
4310 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4311
4312
4313 static int addrconf_net_init(struct net *net)
4314 {
4315         return 0;
4316 }
4317
4318 static void addrconf_net_exit(struct net *net)
4319 {
4320         struct net_device *dev;
4321
4322         rtnl_lock();
4323         /* clean dev list */
4324         for_each_netdev(net, dev) {
4325                 if (__in6_dev_get(dev) == NULL)
4326                         continue;
4327                 addrconf_ifdown(dev, 1);
4328         }
4329         addrconf_ifdown(net->loopback_dev, 2);
4330         rtnl_unlock();
4331 }
4332
4333 static struct pernet_operations addrconf_net_ops = {
4334         .init = addrconf_net_init,
4335         .exit = addrconf_net_exit,
4336 };
4337
4338 /*
4339  *      Init / cleanup code
4340  */
4341
4342 int __init addrconf_init(void)
4343 {
4344         int err;
4345
4346         if ((err = ipv6_addr_label_init()) < 0) {
4347                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4348                         err);
4349                 return err;
4350         }
4351
4352         register_pernet_subsys(&addrconf_ops);
4353
4354         /* The addrconf netdev notifier requires that loopback_dev
4355          * has it's ipv6 private information allocated and setup
4356          * before it can bring up and give link-local addresses
4357          * to other devices which are up.
4358          *
4359          * Unfortunately, loopback_dev is not necessarily the first
4360          * entry in the global dev_base list of net devices.  In fact,
4361          * it is likely to be the very last entry on that list.
4362          * So this causes the notifier registry below to try and
4363          * give link-local addresses to all devices besides loopback_dev
4364          * first, then loopback_dev, which cases all the non-loopback_dev
4365          * devices to fail to get a link-local address.
4366          *
4367          * So, as a temporary fix, allocate the ipv6 structure for
4368          * loopback_dev first by hand.
4369          * Longer term, all of the dependencies ipv6 has upon the loopback
4370          * device and it being up should be removed.
4371          */
4372         rtnl_lock();
4373         if (!ipv6_add_dev(init_net.loopback_dev))
4374                 err = -ENOMEM;
4375         rtnl_unlock();
4376         if (err)
4377                 goto errlo;
4378
4379         err = register_pernet_device(&addrconf_net_ops);
4380         if (err)
4381                 return err;
4382
4383         register_netdevice_notifier(&ipv6_dev_notf);
4384
4385         addrconf_verify(0);
4386
4387         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4388         if (err < 0)
4389                 goto errout;
4390
4391         /* Only the first call to __rtnl_register can fail */
4392         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4393         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4394         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4395         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4396         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4397
4398         ipv6_addr_label_rtnl_register();
4399
4400         return 0;
4401 errout:
4402         unregister_netdevice_notifier(&ipv6_dev_notf);
4403 errlo:
4404         unregister_pernet_subsys(&addrconf_ops);
4405
4406         return err;
4407 }
4408
4409 void addrconf_cleanup(void)
4410 {
4411         struct inet6_ifaddr *ifa;
4412         int i;
4413
4414         unregister_netdevice_notifier(&ipv6_dev_notf);
4415         unregister_pernet_device(&addrconf_net_ops);
4416
4417         unregister_pernet_subsys(&addrconf_ops);
4418
4419         rtnl_lock();
4420
4421         /*
4422          *      Check hash table.
4423          */
4424         write_lock_bh(&addrconf_hash_lock);
4425         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4426                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4427                         struct inet6_ifaddr *bifa;
4428
4429                         bifa = ifa;
4430                         ifa = ifa->lst_next;
4431                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4432                         /* Do not free it; something is wrong.
4433                            Now we can investigate it with debugger.
4434                          */
4435                 }
4436         }
4437         write_unlock_bh(&addrconf_hash_lock);
4438
4439         del_timer(&addr_chk_timer);
4440         rtnl_unlock();
4441
4442         unregister_pernet_subsys(&addrconf_net_ops);
4443 }