[NET] NETNS: Omit sock->sk_net without CONFIG_NET_NS.
[linux-3.10.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16
17 /*
18  *      Changes:
19  *
20  *      Janos Farkas                    :       delete timer on ifdown
21  *      <chexum@bankinf.banki.hu>
22  *      Andi Kleen                      :       kill double kfree on module
23  *                                              unload.
24  *      Maciej W. Rozycki               :       FDDI support
25  *      sekiya@USAGI                    :       Don't send too many RS
26  *                                              packets.
27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
28  *                                              packets.
29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
30  *                                              address validation timer.
31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
32  *                                              support.
33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
34  *                                              address on a same interface.
35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
37  *                                              seq_file.
38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
39  *                                              selection; consider scope,
40  *                                              status etc.
41  */
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99
100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
108 {
109 }
110
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 {
113 }
114 #endif
115
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
120
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
123
124 static int ipv6_count_addresses(struct inet6_dev *idev);
125
126 /*
127  *      Configured unicast address hash table
128  */
129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
131
132 static void addrconf_verify(unsigned long);
133
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
136
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139
140 static int addrconf_ifdown(struct net_device *dev, int how);
141
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
149
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151                                 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153                               struct net_device *dev);
154
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156
157 struct ipv6_devconf ipv6_devconf __read_mostly = {
158         .forwarding             = 0,
159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
160         .mtu6                   = IPV6_MIN_MTU,
161         .accept_ra              = 1,
162         .accept_redirects       = 1,
163         .autoconf               = 1,
164         .force_mld_version      = 0,
165         .dad_transmits          = 1,
166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170         .use_tempaddr           = 0,
171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
173         .regen_max_retry        = REGEN_MAX_RETRY,
174         .max_desync_factor      = MAX_DESYNC_FACTOR,
175 #endif
176         .max_addresses          = IPV6_MAX_ADDRESSES,
177         .accept_ra_defrtr       = 1,
178         .accept_ra_pinfo        = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180         .accept_ra_rtr_pref     = 1,
181         .rtr_probe_interval     = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183         .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186         .proxy_ndp              = 0,
187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
188 };
189
190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
191         .forwarding             = 0,
192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
193         .mtu6                   = IPV6_MIN_MTU,
194         .accept_ra              = 1,
195         .accept_redirects       = 1,
196         .autoconf               = 1,
197         .dad_transmits          = 1,
198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
201 #ifdef CONFIG_IPV6_PRIVACY
202         .use_tempaddr           = 0,
203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
205         .regen_max_retry        = REGEN_MAX_RETRY,
206         .max_desync_factor      = MAX_DESYNC_FACTOR,
207 #endif
208         .max_addresses          = IPV6_MAX_ADDRESSES,
209         .accept_ra_defrtr       = 1,
210         .accept_ra_pinfo        = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212         .accept_ra_rtr_pref     = 1,
213         .rtr_probe_interval     = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215         .accept_ra_rt_info_max_plen = 0,
216 #endif
217 #endif
218         .proxy_ndp              = 0,
219         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
220 };
221
222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
225
226 /* Check if a valid qdisc is available */
227 static inline int addrconf_qdisc_ok(struct net_device *dev)
228 {
229         return (dev->qdisc != &noop_qdisc);
230 }
231
232 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
233 {
234         if (del_timer(&ifp->timer))
235                 __in6_ifa_put(ifp);
236 }
237
238 enum addrconf_timer_t
239 {
240         AC_NONE,
241         AC_DAD,
242         AC_RS,
243 };
244
245 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
246                                enum addrconf_timer_t what,
247                                unsigned long when)
248 {
249         if (!del_timer(&ifp->timer))
250                 in6_ifa_hold(ifp);
251
252         switch (what) {
253         case AC_DAD:
254                 ifp->timer.function = addrconf_dad_timer;
255                 break;
256         case AC_RS:
257                 ifp->timer.function = addrconf_rs_timer;
258                 break;
259         default:;
260         }
261         ifp->timer.expires = jiffies + when;
262         add_timer(&ifp->timer);
263 }
264
265 static int snmp6_alloc_dev(struct inet6_dev *idev)
266 {
267         if (snmp_mib_init((void **)idev->stats.ipv6,
268                           sizeof(struct ipstats_mib)) < 0)
269                 goto err_ip;
270         if (snmp_mib_init((void **)idev->stats.icmpv6,
271                           sizeof(struct icmpv6_mib)) < 0)
272                 goto err_icmp;
273         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
274                           sizeof(struct icmpv6msg_mib)) < 0)
275                 goto err_icmpmsg;
276
277         return 0;
278
279 err_icmpmsg:
280         snmp_mib_free((void **)idev->stats.icmpv6);
281 err_icmp:
282         snmp_mib_free((void **)idev->stats.ipv6);
283 err_ip:
284         return -ENOMEM;
285 }
286
287 static void snmp6_free_dev(struct inet6_dev *idev)
288 {
289         snmp_mib_free((void **)idev->stats.icmpv6msg);
290         snmp_mib_free((void **)idev->stats.icmpv6);
291         snmp_mib_free((void **)idev->stats.ipv6);
292 }
293
294 /* Nobody refers to this device, we may destroy it. */
295
296 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
297 {
298         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
299         kfree(idev);
300 }
301
302 void in6_dev_finish_destroy(struct inet6_dev *idev)
303 {
304         struct net_device *dev = idev->dev;
305         BUG_TRAP(idev->addr_list==NULL);
306         BUG_TRAP(idev->mc_list==NULL);
307 #ifdef NET_REFCNT_DEBUG
308         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
309 #endif
310         dev_put(dev);
311         if (!idev->dead) {
312                 printk("Freeing alive inet6 device %p\n", idev);
313                 return;
314         }
315         snmp6_free_dev(idev);
316         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
317 }
318
319 EXPORT_SYMBOL(in6_dev_finish_destroy);
320
321 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
322 {
323         struct inet6_dev *ndev;
324         struct in6_addr maddr;
325
326         ASSERT_RTNL();
327
328         if (dev->mtu < IPV6_MIN_MTU)
329                 return NULL;
330
331         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
332
333         if (ndev == NULL)
334                 return NULL;
335
336         rwlock_init(&ndev->lock);
337         ndev->dev = dev;
338         memcpy(&ndev->cnf, dev_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 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
416 {
417         struct inet6_dev *idev;
418
419         ASSERT_RTNL();
420
421         if ((idev = __in6_dev_get(dev)) == NULL) {
422                 if ((idev = ipv6_add_dev(dev)) == NULL)
423                         return NULL;
424         }
425
426         if (dev->flags&IFF_UP)
427                 ipv6_mc_up(idev);
428         return idev;
429 }
430
431 #ifdef CONFIG_SYSCTL
432 static void dev_forward_change(struct inet6_dev *idev)
433 {
434         struct net_device *dev;
435         struct inet6_ifaddr *ifa;
436         struct in6_addr addr;
437
438         if (!idev)
439                 return;
440         dev = idev->dev;
441         if (dev && (dev->flags & IFF_MULTICAST)) {
442                 ipv6_addr_all_routers(&addr);
443
444                 if (idev->cnf.forwarding)
445                         ipv6_dev_mc_inc(dev, &addr);
446                 else
447                         ipv6_dev_mc_dec(dev, &addr);
448         }
449         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
450                 if (ifa->flags&IFA_F_TENTATIVE)
451                         continue;
452                 if (idev->cnf.forwarding)
453                         addrconf_join_anycast(ifa);
454                 else
455                         addrconf_leave_anycast(ifa);
456         }
457 }
458
459
460 static void addrconf_forward_change(struct net *net, __s32 newf)
461 {
462         struct net_device *dev;
463         struct inet6_dev *idev;
464
465         read_lock(&dev_base_lock);
466         for_each_netdev(net, dev) {
467                 rcu_read_lock();
468                 idev = __in6_dev_get(dev);
469                 if (idev) {
470                         int changed = (!idev->cnf.forwarding) ^ (!newf);
471                         idev->cnf.forwarding = newf;
472                         if (changed)
473                                 dev_forward_change(idev);
474                 }
475                 rcu_read_unlock();
476         }
477         read_unlock(&dev_base_lock);
478 }
479
480 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
481 {
482         struct net *net;
483
484         net = (struct net *)table->extra2;
485         if (p == &net->ipv6.devconf_dflt->forwarding)
486                 return;
487
488         if (p == &net->ipv6.devconf_all->forwarding) {
489                 __s32 newf = net->ipv6.devconf_all->forwarding;
490                 net->ipv6.devconf_dflt->forwarding = newf;
491                 addrconf_forward_change(net, newf);
492         } else if ((!*p) ^ (!old))
493                 dev_forward_change((struct inet6_dev *)table->extra1);
494
495         if (*p)
496                 rt6_purge_dflt_routers(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         int tmp_plen;
780         int ret = 0;
781         int max_addresses;
782         u32 addr_flags;
783
784         write_lock(&idev->lock);
785         if (ift) {
786                 spin_lock_bh(&ift->lock);
787                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
788                 spin_unlock_bh(&ift->lock);
789                 tmpaddr = &addr;
790         } else {
791                 tmpaddr = NULL;
792         }
793 retry:
794         in6_dev_hold(idev);
795         if (idev->cnf.use_tempaddr <= 0) {
796                 write_unlock(&idev->lock);
797                 printk(KERN_INFO
798                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
799                 in6_dev_put(idev);
800                 ret = -1;
801                 goto out;
802         }
803         spin_lock_bh(&ifp->lock);
804         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
805                 idev->cnf.use_tempaddr = -1;    /*XXX*/
806                 spin_unlock_bh(&ifp->lock);
807                 write_unlock(&idev->lock);
808                 printk(KERN_WARNING
809                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
810                 in6_dev_put(idev);
811                 ret = -1;
812                 goto out;
813         }
814         in6_ifa_hold(ifp);
815         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
816         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
817                 spin_unlock_bh(&ifp->lock);
818                 write_unlock(&idev->lock);
819                 printk(KERN_WARNING
820                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
821                 in6_ifa_put(ifp);
822                 in6_dev_put(idev);
823                 ret = -1;
824                 goto out;
825         }
826         memcpy(&addr.s6_addr[8], idev->rndid, 8);
827         tmp_valid_lft = min_t(__u32,
828                               ifp->valid_lft,
829                               idev->cnf.temp_valid_lft);
830         tmp_prefered_lft = min_t(__u32,
831                                  ifp->prefered_lft,
832                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
833         tmp_plen = ifp->prefix_len;
834         max_addresses = idev->cnf.max_addresses;
835         tmp_cstamp = ifp->cstamp;
836         tmp_tstamp = ifp->tstamp;
837         spin_unlock_bh(&ifp->lock);
838
839         write_unlock(&idev->lock);
840
841         addr_flags = IFA_F_TEMPORARY;
842         /* set in addrconf_prefix_rcv() */
843         if (ifp->flags & IFA_F_OPTIMISTIC)
844                 addr_flags |= IFA_F_OPTIMISTIC;
845
846         ift = !max_addresses ||
847               ipv6_count_addresses(idev) < max_addresses ?
848                 ipv6_add_addr(idev, &addr, tmp_plen,
849                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
850                               addr_flags) : NULL;
851         if (!ift || IS_ERR(ift)) {
852                 in6_ifa_put(ifp);
853                 in6_dev_put(idev);
854                 printk(KERN_INFO
855                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
856                 tmpaddr = &addr;
857                 write_lock(&idev->lock);
858                 goto retry;
859         }
860
861         spin_lock_bh(&ift->lock);
862         ift->ifpub = ifp;
863         ift->valid_lft = tmp_valid_lft;
864         ift->prefered_lft = tmp_prefered_lft;
865         ift->cstamp = tmp_cstamp;
866         ift->tstamp = tmp_tstamp;
867         spin_unlock_bh(&ift->lock);
868
869         addrconf_dad_start(ift, 0);
870         in6_ifa_put(ift);
871         in6_dev_put(idev);
872 out:
873         return ret;
874 }
875 #endif
876
877 /*
878  *      Choose an appropriate source address (RFC3484)
879  */
880 enum {
881         IPV6_SADDR_RULE_INIT = 0,
882         IPV6_SADDR_RULE_LOCAL,
883         IPV6_SADDR_RULE_SCOPE,
884         IPV6_SADDR_RULE_PREFERRED,
885 #ifdef CONFIG_IPV6_MIP6
886         IPV6_SADDR_RULE_HOA,
887 #endif
888         IPV6_SADDR_RULE_OIF,
889         IPV6_SADDR_RULE_LABEL,
890 #ifdef CONFIG_IPV6_PRIVACY
891         IPV6_SADDR_RULE_PRIVACY,
892 #endif
893         IPV6_SADDR_RULE_ORCHID,
894         IPV6_SADDR_RULE_PREFIX,
895         IPV6_SADDR_RULE_MAX
896 };
897
898 struct ipv6_saddr_score {
899         int                     rule;
900         int                     addr_type;
901         struct inet6_ifaddr     *ifa;
902         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
903         int                     scopedist;
904         int                     matchlen;
905 };
906
907 struct ipv6_saddr_dst {
908         struct in6_addr *addr;
909         int ifindex;
910         int scope;
911         int label;
912         unsigned int prefs;
913 };
914
915 static inline int ipv6_saddr_preferred(int type)
916 {
917         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
918                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
919                 return 1;
920         return 0;
921 }
922
923 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
924                                struct ipv6_saddr_dst *dst,
925                                int i)
926 {
927         int ret;
928
929         if (i <= score->rule) {
930                 switch (i) {
931                 case IPV6_SADDR_RULE_SCOPE:
932                         ret = score->scopedist;
933                         break;
934                 case IPV6_SADDR_RULE_PREFIX:
935                         ret = score->matchlen;
936                         break;
937                 default:
938                         ret = !!test_bit(i, score->scorebits);
939                 }
940                 goto out;
941         }
942
943         switch (i) {
944         case IPV6_SADDR_RULE_INIT:
945                 /* Rule 0: remember if hiscore is not ready yet */
946                 ret = !!score->ifa;
947                 break;
948         case IPV6_SADDR_RULE_LOCAL:
949                 /* Rule 1: Prefer same address */
950                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
951                 break;
952         case IPV6_SADDR_RULE_SCOPE:
953                 /* Rule 2: Prefer appropriate scope
954                  *
955                  *      ret
956                  *       ^
957                  *    -1 |  d 15
958                  *    ---+--+-+---> scope
959                  *       |
960                  *       |             d is scope of the destination.
961                  *  B-d  |  \
962                  *       |   \      <- smaller scope is better if
963                  *  B-15 |    \        if scope is enough for destinaion.
964                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
965                  * d-C-1 | /
966                  *       |/         <- greater is better
967                  *   -C  /             if scope is not enough for destination.
968                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
969                  *
970                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
971                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
972                  * Assume B = 0 and we get C > 29.
973                  */
974                 ret = __ipv6_addr_src_scope(score->addr_type);
975                 if (ret >= dst->scope)
976                         ret = -ret;
977                 else
978                         ret -= 128;     /* 30 is enough */
979                 score->scopedist = ret;
980                 break;
981         case IPV6_SADDR_RULE_PREFERRED:
982                 /* Rule 3: Avoid deprecated and optimistic addresses */
983                 ret = ipv6_saddr_preferred(score->addr_type) ||
984                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
985                 break;
986 #ifdef CONFIG_IPV6_MIP6
987         case IPV6_SADDR_RULE_HOA:
988             {
989                 /* Rule 4: Prefer home address */
990                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
991                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
992                 break;
993             }
994 #endif
995         case IPV6_SADDR_RULE_OIF:
996                 /* Rule 5: Prefer outgoing interface */
997                 ret = (!dst->ifindex ||
998                        dst->ifindex == score->ifa->idev->dev->ifindex);
999                 break;
1000         case IPV6_SADDR_RULE_LABEL:
1001                 /* Rule 6: Prefer matching label */
1002                 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
1003                                       score->ifa->idev->dev->ifindex) == dst->label;
1004                 break;
1005 #ifdef CONFIG_IPV6_PRIVACY
1006         case IPV6_SADDR_RULE_PRIVACY:
1007             {
1008                 /* Rule 7: Prefer public address
1009                  * Note: prefer temprary address if use_tempaddr >= 2
1010                  */
1011                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1012                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1013                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1014                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1015                 break;
1016             }
1017 #endif
1018         case IPV6_SADDR_RULE_ORCHID:
1019                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1020                  *          non-ORCHID vs non-ORCHID
1021                  */
1022                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1023                         ipv6_addr_orchid(dst->addr));
1024                 break;
1025         case IPV6_SADDR_RULE_PREFIX:
1026                 /* Rule 8: Use longest matching prefix */
1027                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1028                                                        dst->addr);
1029                 break;
1030         default:
1031                 ret = 0;
1032         }
1033
1034         if (ret)
1035                 __set_bit(i, score->scorebits);
1036         score->rule = i;
1037 out:
1038         return ret;
1039 }
1040
1041 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1042                        struct in6_addr *daddr, unsigned int prefs,
1043                        struct in6_addr *saddr)
1044 {
1045         struct ipv6_saddr_score scores[2],
1046                                 *score = &scores[0], *hiscore = &scores[1];
1047         struct net *net = dev_net(dst_dev);
1048         struct ipv6_saddr_dst dst;
1049         struct net_device *dev;
1050         int dst_type;
1051
1052         dst_type = __ipv6_addr_type(daddr);
1053         dst.addr = daddr;
1054         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1055         dst.scope = __ipv6_addr_src_scope(dst_type);
1056         dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1057         dst.prefs = prefs;
1058
1059         hiscore->rule = -1;
1060         hiscore->ifa = NULL;
1061
1062         read_lock(&dev_base_lock);
1063         rcu_read_lock();
1064
1065         for_each_netdev(net, dev) {
1066                 struct inet6_dev *idev;
1067
1068                 /* Candidate Source Address (section 4)
1069                  *  - multicast and link-local destination address,
1070                  *    the set of candidate source address MUST only
1071                  *    include addresses assigned to interfaces
1072                  *    belonging to the same link as the outgoing
1073                  *    interface.
1074                  * (- For site-local destination addresses, the
1075                  *    set of candidate source addresses MUST only
1076                  *    include addresses assigned to interfaces
1077                  *    belonging to the same site as the outgoing
1078                  *    interface.)
1079                  */
1080                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1081                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1082                     dst.ifindex && dev->ifindex != dst.ifindex)
1083                         continue;
1084
1085                 idev = __in6_dev_get(dev);
1086                 if (!idev)
1087                         continue;
1088
1089                 read_lock_bh(&idev->lock);
1090                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1091                         int i;
1092
1093                         /*
1094                          * - Tentative Address (RFC2462 section 5.4)
1095                          *  - A tentative address is not considered
1096                          *    "assigned to an interface" in the traditional
1097                          *    sense, unless it is also flagged as optimistic.
1098                          * - Candidate Source Address (section 4)
1099                          *  - In any case, anycast addresses, multicast
1100                          *    addresses, and the unspecified address MUST
1101                          *    NOT be included in a candidate set.
1102                          */
1103                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1104                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1105                                 continue;
1106
1107                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1108
1109                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1110                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1111                                 LIMIT_NETDEBUG(KERN_DEBUG
1112                                                "ADDRCONF: unspecified / multicast address "
1113                                                "assigned as unicast address on %s",
1114                                                dev->name);
1115                                 continue;
1116                         }
1117
1118                         score->rule = -1;
1119                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1120
1121                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1122                                 int minihiscore, miniscore;
1123
1124                                 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1125                                 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1126
1127                                 if (minihiscore > miniscore) {
1128                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1129                                             score->scopedist > 0) {
1130                                                 /*
1131                                                  * special case:
1132                                                  * each remaining entry
1133                                                  * has too small (not enough)
1134                                                  * scope, because ifa entries
1135                                                  * are sorted by their scope
1136                                                  * values.
1137                                                  */
1138                                                 goto try_nextdev;
1139                                         }
1140                                         break;
1141                                 } else if (minihiscore < miniscore) {
1142                                         struct ipv6_saddr_score *tmp;
1143
1144                                         if (hiscore->ifa)
1145                                                 in6_ifa_put(hiscore->ifa);
1146
1147                                         in6_ifa_hold(score->ifa);
1148
1149                                         tmp = hiscore;
1150                                         hiscore = score;
1151                                         score = tmp;
1152
1153                                         /* restore our iterator */
1154                                         score->ifa = hiscore->ifa;
1155
1156                                         break;
1157                                 }
1158                         }
1159                 }
1160 try_nextdev:
1161                 read_unlock_bh(&idev->lock);
1162         }
1163         rcu_read_unlock();
1164         read_unlock(&dev_base_lock);
1165
1166         if (!hiscore->ifa)
1167                 return -EADDRNOTAVAIL;
1168
1169         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1170         in6_ifa_put(hiscore->ifa);
1171         return 0;
1172 }
1173
1174 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1175
1176 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1177                     unsigned char banned_flags)
1178 {
1179         struct inet6_dev *idev;
1180         int err = -EADDRNOTAVAIL;
1181
1182         rcu_read_lock();
1183         if ((idev = __in6_dev_get(dev)) != NULL) {
1184                 struct inet6_ifaddr *ifp;
1185
1186                 read_lock_bh(&idev->lock);
1187                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1188                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1189                                 ipv6_addr_copy(addr, &ifp->addr);
1190                                 err = 0;
1191                                 break;
1192                         }
1193                 }
1194                 read_unlock_bh(&idev->lock);
1195         }
1196         rcu_read_unlock();
1197         return err;
1198 }
1199
1200 static int ipv6_count_addresses(struct inet6_dev *idev)
1201 {
1202         int cnt = 0;
1203         struct inet6_ifaddr *ifp;
1204
1205         read_lock_bh(&idev->lock);
1206         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1207                 cnt++;
1208         read_unlock_bh(&idev->lock);
1209         return cnt;
1210 }
1211
1212 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1213                   struct net_device *dev, int strict)
1214 {
1215         struct inet6_ifaddr * ifp;
1216         u8 hash = ipv6_addr_hash(addr);
1217
1218         read_lock_bh(&addrconf_hash_lock);
1219         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1220                 if (dev_net(ifp->idev->dev) != net)
1221                         continue;
1222                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1223                     !(ifp->flags&IFA_F_TENTATIVE)) {
1224                         if (dev == NULL || ifp->idev->dev == dev ||
1225                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1226                                 break;
1227                 }
1228         }
1229         read_unlock_bh(&addrconf_hash_lock);
1230         return ifp != NULL;
1231 }
1232 EXPORT_SYMBOL(ipv6_chk_addr);
1233
1234 static
1235 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1236                        struct net_device *dev)
1237 {
1238         struct inet6_ifaddr * ifp;
1239         u8 hash = ipv6_addr_hash(addr);
1240
1241         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1242                 if (dev_net(ifp->idev->dev) != net)
1243                         continue;
1244                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1245                         if (dev == NULL || ifp->idev->dev == dev)
1246                                 break;
1247                 }
1248         }
1249         return ifp != NULL;
1250 }
1251
1252 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1253                                      struct net_device *dev, int strict)
1254 {
1255         struct inet6_ifaddr * ifp;
1256         u8 hash = ipv6_addr_hash(addr);
1257
1258         read_lock_bh(&addrconf_hash_lock);
1259         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1260                 if (dev_net(ifp->idev->dev) != net)
1261                         continue;
1262                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1263                         if (dev == NULL || ifp->idev->dev == dev ||
1264                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1265                                 in6_ifa_hold(ifp);
1266                                 break;
1267                         }
1268                 }
1269         }
1270         read_unlock_bh(&addrconf_hash_lock);
1271
1272         return ifp;
1273 }
1274
1275 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1276 {
1277         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1278         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1279         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1280         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1281         int sk_ipv6only = ipv6_only_sock(sk);
1282         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1283         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1284         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1285
1286         if (!sk2_rcv_saddr && !sk_ipv6only)
1287                 return 1;
1288
1289         if (addr_type2 == IPV6_ADDR_ANY &&
1290             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1291                 return 1;
1292
1293         if (addr_type == IPV6_ADDR_ANY &&
1294             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1295                 return 1;
1296
1297         if (sk2_rcv_saddr6 &&
1298             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1299                 return 1;
1300
1301         if (addr_type == IPV6_ADDR_MAPPED &&
1302             !sk2_ipv6only &&
1303             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1304                 return 1;
1305
1306         return 0;
1307 }
1308
1309 /* Gets referenced address, destroys ifaddr */
1310
1311 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1312 {
1313         if (ifp->flags&IFA_F_PERMANENT) {
1314                 spin_lock_bh(&ifp->lock);
1315                 addrconf_del_timer(ifp);
1316                 ifp->flags |= IFA_F_TENTATIVE;
1317                 spin_unlock_bh(&ifp->lock);
1318                 in6_ifa_put(ifp);
1319 #ifdef CONFIG_IPV6_PRIVACY
1320         } else if (ifp->flags&IFA_F_TEMPORARY) {
1321                 struct inet6_ifaddr *ifpub;
1322                 spin_lock_bh(&ifp->lock);
1323                 ifpub = ifp->ifpub;
1324                 if (ifpub) {
1325                         in6_ifa_hold(ifpub);
1326                         spin_unlock_bh(&ifp->lock);
1327                         ipv6_create_tempaddr(ifpub, ifp);
1328                         in6_ifa_put(ifpub);
1329                 } else {
1330                         spin_unlock_bh(&ifp->lock);
1331                 }
1332                 ipv6_del_addr(ifp);
1333 #endif
1334         } else
1335                 ipv6_del_addr(ifp);
1336 }
1337
1338 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1339 {
1340         if (net_ratelimit())
1341                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1342         addrconf_dad_stop(ifp);
1343 }
1344
1345 /* Join to solicited addr multicast group. */
1346
1347 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1348 {
1349         struct in6_addr maddr;
1350
1351         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1352                 return;
1353
1354         addrconf_addr_solict_mult(addr, &maddr);
1355         ipv6_dev_mc_inc(dev, &maddr);
1356 }
1357
1358 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1359 {
1360         struct in6_addr maddr;
1361
1362         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1363                 return;
1364
1365         addrconf_addr_solict_mult(addr, &maddr);
1366         __ipv6_dev_mc_dec(idev, &maddr);
1367 }
1368
1369 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1370 {
1371         struct in6_addr addr;
1372         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1373         if (ipv6_addr_any(&addr))
1374                 return;
1375         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1376 }
1377
1378 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1379 {
1380         struct in6_addr addr;
1381         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1382         if (ipv6_addr_any(&addr))
1383                 return;
1384         __ipv6_dev_ac_dec(ifp->idev, &addr);
1385 }
1386
1387 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1388 {
1389         if (dev->addr_len != ETH_ALEN)
1390                 return -1;
1391         memcpy(eui, dev->dev_addr, 3);
1392         memcpy(eui + 5, dev->dev_addr + 3, 3);
1393
1394         /*
1395          * The zSeries OSA network cards can be shared among various
1396          * OS instances, but the OSA cards have only one MAC address.
1397          * This leads to duplicate address conflicts in conjunction
1398          * with IPv6 if more than one instance uses the same card.
1399          *
1400          * The driver for these cards can deliver a unique 16-bit
1401          * identifier for each instance sharing the same card.  It is
1402          * placed instead of 0xFFFE in the interface identifier.  The
1403          * "u" bit of the interface identifier is not inverted in this
1404          * case.  Hence the resulting interface identifier has local
1405          * scope according to RFC2373.
1406          */
1407         if (dev->dev_id) {
1408                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1409                 eui[4] = dev->dev_id & 0xFF;
1410         } else {
1411                 eui[3] = 0xFF;
1412                 eui[4] = 0xFE;
1413                 eui[0] ^= 2;
1414         }
1415         return 0;
1416 }
1417
1418 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1419 {
1420         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1421         if (dev->addr_len != ARCNET_ALEN)
1422                 return -1;
1423         memset(eui, 0, 7);
1424         eui[7] = *(u8*)dev->dev_addr;
1425         return 0;
1426 }
1427
1428 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1429 {
1430         if (dev->addr_len != INFINIBAND_ALEN)
1431                 return -1;
1432         memcpy(eui, dev->dev_addr + 12, 8);
1433         eui[0] |= 2;
1434         return 0;
1435 }
1436
1437 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1438 {
1439         switch (dev->type) {
1440         case ARPHRD_ETHER:
1441         case ARPHRD_FDDI:
1442         case ARPHRD_IEEE802_TR:
1443                 return addrconf_ifid_eui48(eui, dev);
1444         case ARPHRD_ARCNET:
1445                 return addrconf_ifid_arcnet(eui, dev);
1446         case ARPHRD_INFINIBAND:
1447                 return addrconf_ifid_infiniband(eui, dev);
1448         case ARPHRD_SIT:
1449                 if (dev->priv_flags & IFF_ISATAP)
1450                         return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1451         }
1452         return -1;
1453 }
1454
1455 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1456 {
1457         int err = -1;
1458         struct inet6_ifaddr *ifp;
1459
1460         read_lock_bh(&idev->lock);
1461         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1462                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1463                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1464                         err = 0;
1465                         break;
1466                 }
1467         }
1468         read_unlock_bh(&idev->lock);
1469         return err;
1470 }
1471
1472 #ifdef CONFIG_IPV6_PRIVACY
1473 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1474 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1475 {
1476 regen:
1477         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1478         idev->rndid[0] &= ~0x02;
1479
1480         /*
1481          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1482          * check if generated address is not inappropriate
1483          *
1484          *  - Reserved subnet anycast (RFC 2526)
1485          *      11111101 11....11 1xxxxxxx
1486          *  - ISATAP (RFC4214) 6.1
1487          *      00-00-5E-FE-xx-xx-xx-xx
1488          *  - value 0
1489          *  - XXX: already assigned to an address on the device
1490          */
1491         if (idev->rndid[0] == 0xfd &&
1492             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1493             (idev->rndid[7]&0x80))
1494                 goto regen;
1495         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1496                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1497                         goto regen;
1498                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1499                         goto regen;
1500         }
1501
1502         return 0;
1503 }
1504
1505 static void ipv6_regen_rndid(unsigned long data)
1506 {
1507         struct inet6_dev *idev = (struct inet6_dev *) data;
1508         unsigned long expires;
1509
1510         rcu_read_lock_bh();
1511         write_lock_bh(&idev->lock);
1512
1513         if (idev->dead)
1514                 goto out;
1515
1516         if (__ipv6_regen_rndid(idev) < 0)
1517                 goto out;
1518
1519         expires = jiffies +
1520                 idev->cnf.temp_prefered_lft * HZ -
1521                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1522         if (time_before(expires, jiffies)) {
1523                 printk(KERN_WARNING
1524                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1525                         idev->dev->name);
1526                 goto out;
1527         }
1528
1529         if (!mod_timer(&idev->regen_timer, expires))
1530                 in6_dev_hold(idev);
1531
1532 out:
1533         write_unlock_bh(&idev->lock);
1534         rcu_read_unlock_bh();
1535         in6_dev_put(idev);
1536 }
1537
1538 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1539         int ret = 0;
1540
1541         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1542                 ret = __ipv6_regen_rndid(idev);
1543         return ret;
1544 }
1545 #endif
1546
1547 /*
1548  *      Add prefix route.
1549  */
1550
1551 static void
1552 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1553                       unsigned long expires, u32 flags)
1554 {
1555         struct fib6_config cfg = {
1556                 .fc_table = RT6_TABLE_PREFIX,
1557                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1558                 .fc_ifindex = dev->ifindex,
1559                 .fc_expires = expires,
1560                 .fc_dst_len = plen,
1561                 .fc_flags = RTF_UP | flags,
1562                 .fc_nlinfo.nl_net = dev_net(dev),
1563         };
1564
1565         ipv6_addr_copy(&cfg.fc_dst, pfx);
1566
1567         /* Prevent useless cloning on PtP SIT.
1568            This thing is done here expecting that the whole
1569            class of non-broadcast devices need not cloning.
1570          */
1571 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1572         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1573                 cfg.fc_flags |= RTF_NONEXTHOP;
1574 #endif
1575
1576         ip6_route_add(&cfg);
1577 }
1578
1579 /* Create "default" multicast route to the interface */
1580
1581 static void addrconf_add_mroute(struct net_device *dev)
1582 {
1583         struct fib6_config cfg = {
1584                 .fc_table = RT6_TABLE_LOCAL,
1585                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1586                 .fc_ifindex = dev->ifindex,
1587                 .fc_dst_len = 8,
1588                 .fc_flags = RTF_UP,
1589                 .fc_nlinfo.nl_net = dev_net(dev),
1590         };
1591
1592         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1593
1594         ip6_route_add(&cfg);
1595 }
1596
1597 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1598 static void sit_route_add(struct net_device *dev)
1599 {
1600         struct fib6_config cfg = {
1601                 .fc_table = RT6_TABLE_MAIN,
1602                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1603                 .fc_ifindex = dev->ifindex,
1604                 .fc_dst_len = 96,
1605                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1606                 .fc_nlinfo.nl_net = dev_net(dev),
1607         };
1608
1609         /* prefix length - 96 bits "::d.d.d.d" */
1610         ip6_route_add(&cfg);
1611 }
1612 #endif
1613
1614 static void addrconf_add_lroute(struct net_device *dev)
1615 {
1616         struct in6_addr addr;
1617
1618         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1619         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1620 }
1621
1622 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1623 {
1624         struct inet6_dev *idev;
1625
1626         ASSERT_RTNL();
1627
1628         if ((idev = ipv6_find_idev(dev)) == NULL)
1629                 return NULL;
1630
1631         /* Add default multicast route */
1632         addrconf_add_mroute(dev);
1633
1634         /* Add link local route */
1635         addrconf_add_lroute(dev);
1636         return idev;
1637 }
1638
1639 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1640 {
1641         struct prefix_info *pinfo;
1642         __u32 valid_lft;
1643         __u32 prefered_lft;
1644         int addr_type;
1645         unsigned long rt_expires;
1646         struct inet6_dev *in6_dev;
1647
1648         pinfo = (struct prefix_info *) opt;
1649
1650         if (len < sizeof(struct prefix_info)) {
1651                 ADBG(("addrconf: prefix option too short\n"));
1652                 return;
1653         }
1654
1655         /*
1656          *      Validation checks ([ADDRCONF], page 19)
1657          */
1658
1659         addr_type = ipv6_addr_type(&pinfo->prefix);
1660
1661         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1662                 return;
1663
1664         valid_lft = ntohl(pinfo->valid);
1665         prefered_lft = ntohl(pinfo->prefered);
1666
1667         if (prefered_lft > valid_lft) {
1668                 if (net_ratelimit())
1669                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1670                 return;
1671         }
1672
1673         in6_dev = in6_dev_get(dev);
1674
1675         if (in6_dev == NULL) {
1676                 if (net_ratelimit())
1677                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1678                 return;
1679         }
1680
1681         /*
1682          *      Two things going on here:
1683          *      1) Add routes for on-link prefixes
1684          *      2) Configure prefixes with the auto flag set
1685          */
1686
1687         /* Avoid arithmetic overflow. Really, we could
1688            save rt_expires in seconds, likely valid_lft,
1689            but it would require division in fib gc, that it
1690            not good.
1691          */
1692         if (valid_lft >= 0x7FFFFFFF/HZ)
1693                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1694         else
1695                 rt_expires = valid_lft * HZ;
1696
1697         /*
1698          * We convert this (in jiffies) to clock_t later.
1699          * Avoid arithmetic overflow there as well.
1700          * Overflow can happen only if HZ < USER_HZ.
1701          */
1702         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1703                 rt_expires = 0x7FFFFFFF / USER_HZ;
1704
1705         if (pinfo->onlink) {
1706                 struct rt6_info *rt;
1707                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1708                                 dev->ifindex, 1);
1709
1710                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1711                         if (rt->rt6i_flags&RTF_EXPIRES) {
1712                                 if (valid_lft == 0) {
1713                                         ip6_del_rt(rt);
1714                                         rt = NULL;
1715                                 } else {
1716                                         rt->rt6i_expires = jiffies + rt_expires;
1717                                 }
1718                         }
1719                 } else if (valid_lft) {
1720                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1721                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1722                 }
1723                 if (rt)
1724                         dst_release(&rt->u.dst);
1725         }
1726
1727         /* Try to figure out our local address for this prefix */
1728
1729         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1730                 struct inet6_ifaddr * ifp;
1731                 struct in6_addr addr;
1732                 int create = 0, update_lft = 0;
1733
1734                 if (pinfo->prefix_len == 64) {
1735                         memcpy(&addr, &pinfo->prefix, 8);
1736                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1737                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1738                                 in6_dev_put(in6_dev);
1739                                 return;
1740                         }
1741                         goto ok;
1742                 }
1743                 if (net_ratelimit())
1744                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1745                                pinfo->prefix_len);
1746                 in6_dev_put(in6_dev);
1747                 return;
1748
1749 ok:
1750
1751                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1752
1753                 if (ifp == NULL && valid_lft) {
1754                         int max_addresses = in6_dev->cnf.max_addresses;
1755                         u32 addr_flags = 0;
1756
1757 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1758                         if (in6_dev->cnf.optimistic_dad &&
1759                             !ipv6_devconf.forwarding)
1760                                 addr_flags = IFA_F_OPTIMISTIC;
1761 #endif
1762
1763                         /* Do not allow to create too much of autoconfigured
1764                          * addresses; this would be too easy way to crash kernel.
1765                          */
1766                         if (!max_addresses ||
1767                             ipv6_count_addresses(in6_dev) < max_addresses)
1768                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1769                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1770                                                     addr_flags);
1771
1772                         if (!ifp || IS_ERR(ifp)) {
1773                                 in6_dev_put(in6_dev);
1774                                 return;
1775                         }
1776
1777                         update_lft = create = 1;
1778                         ifp->cstamp = jiffies;
1779                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1780                 }
1781
1782                 if (ifp) {
1783                         int flags;
1784                         unsigned long now;
1785 #ifdef CONFIG_IPV6_PRIVACY
1786                         struct inet6_ifaddr *ift;
1787 #endif
1788                         u32 stored_lft;
1789
1790                         /* update lifetime (RFC2462 5.5.3 e) */
1791                         spin_lock(&ifp->lock);
1792                         now = jiffies;
1793                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1794                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1795                         else
1796                                 stored_lft = 0;
1797                         if (!update_lft && stored_lft) {
1798                                 if (valid_lft > MIN_VALID_LIFETIME ||
1799                                     valid_lft > stored_lft)
1800                                         update_lft = 1;
1801                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1802                                         /* valid_lft <= stored_lft is always true */
1803                                         /* XXX: IPsec */
1804                                         update_lft = 0;
1805                                 } else {
1806                                         valid_lft = MIN_VALID_LIFETIME;
1807                                         if (valid_lft < prefered_lft)
1808                                                 prefered_lft = valid_lft;
1809                                         update_lft = 1;
1810                                 }
1811                         }
1812
1813                         if (update_lft) {
1814                                 ifp->valid_lft = valid_lft;
1815                                 ifp->prefered_lft = prefered_lft;
1816                                 ifp->tstamp = now;
1817                                 flags = ifp->flags;
1818                                 ifp->flags &= ~IFA_F_DEPRECATED;
1819                                 spin_unlock(&ifp->lock);
1820
1821                                 if (!(flags&IFA_F_TENTATIVE))
1822                                         ipv6_ifa_notify(0, ifp);
1823                         } else
1824                                 spin_unlock(&ifp->lock);
1825
1826 #ifdef CONFIG_IPV6_PRIVACY
1827                         read_lock_bh(&in6_dev->lock);
1828                         /* update all temporary addresses in the list */
1829                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1830                                 /*
1831                                  * When adjusting the lifetimes of an existing
1832                                  * temporary address, only lower the lifetimes.
1833                                  * Implementations must not increase the
1834                                  * lifetimes of an existing temporary address
1835                                  * when processing a Prefix Information Option.
1836                                  */
1837                                 spin_lock(&ift->lock);
1838                                 flags = ift->flags;
1839                                 if (ift->valid_lft > valid_lft &&
1840                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1841                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1842                                 if (ift->prefered_lft > prefered_lft &&
1843                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1844                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1845                                 spin_unlock(&ift->lock);
1846                                 if (!(flags&IFA_F_TENTATIVE))
1847                                         ipv6_ifa_notify(0, ift);
1848                         }
1849
1850                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1851                                 /*
1852                                  * When a new public address is created as described in [ADDRCONF],
1853                                  * also create a new temporary address.
1854                                  */
1855                                 read_unlock_bh(&in6_dev->lock);
1856                                 ipv6_create_tempaddr(ifp, NULL);
1857                         } else {
1858                                 read_unlock_bh(&in6_dev->lock);
1859                         }
1860 #endif
1861                         in6_ifa_put(ifp);
1862                         addrconf_verify(0);
1863                 }
1864         }
1865         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1866         in6_dev_put(in6_dev);
1867 }
1868
1869 /*
1870  *      Set destination address.
1871  *      Special case for SIT interfaces where we create a new "virtual"
1872  *      device.
1873  */
1874 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1875 {
1876         struct in6_ifreq ireq;
1877         struct net_device *dev;
1878         int err = -EINVAL;
1879
1880         rtnl_lock();
1881
1882         err = -EFAULT;
1883         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1884                 goto err_exit;
1885
1886         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1887
1888         err = -ENODEV;
1889         if (dev == NULL)
1890                 goto err_exit;
1891
1892 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1893         if (dev->type == ARPHRD_SIT) {
1894                 struct ifreq ifr;
1895                 mm_segment_t    oldfs;
1896                 struct ip_tunnel_parm p;
1897
1898                 err = -EADDRNOTAVAIL;
1899                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1900                         goto err_exit;
1901
1902                 memset(&p, 0, sizeof(p));
1903                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1904                 p.iph.saddr = 0;
1905                 p.iph.version = 4;
1906                 p.iph.ihl = 5;
1907                 p.iph.protocol = IPPROTO_IPV6;
1908                 p.iph.ttl = 64;
1909                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1910
1911                 oldfs = get_fs(); set_fs(KERNEL_DS);
1912                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1913                 set_fs(oldfs);
1914
1915                 if (err == 0) {
1916                         err = -ENOBUFS;
1917                         dev = __dev_get_by_name(net, p.name);
1918                         if (!dev)
1919                                 goto err_exit;
1920                         err = dev_open(dev);
1921                 }
1922         }
1923 #endif
1924
1925 err_exit:
1926         rtnl_unlock();
1927         return err;
1928 }
1929
1930 /*
1931  *      Manual configuration of address on an interface
1932  */
1933 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
1934                           int plen, __u8 ifa_flags, __u32 prefered_lft,
1935                           __u32 valid_lft)
1936 {
1937         struct inet6_ifaddr *ifp;
1938         struct inet6_dev *idev;
1939         struct net_device *dev;
1940         int scope;
1941         u32 flags = RTF_EXPIRES;
1942
1943         ASSERT_RTNL();
1944
1945         /* check the lifetime */
1946         if (!valid_lft || prefered_lft > valid_lft)
1947                 return -EINVAL;
1948
1949         dev = __dev_get_by_index(net, ifindex);
1950         if (!dev)
1951                 return -ENODEV;
1952
1953         if ((idev = addrconf_add_dev(dev)) == NULL)
1954                 return -ENOBUFS;
1955
1956         scope = ipv6_addr_scope(pfx);
1957
1958         if (valid_lft == INFINITY_LIFE_TIME) {
1959                 ifa_flags |= IFA_F_PERMANENT;
1960                 flags = 0;
1961         } else if (valid_lft >= 0x7FFFFFFF/HZ)
1962                 valid_lft = 0x7FFFFFFF/HZ;
1963
1964         if (prefered_lft == 0)
1965                 ifa_flags |= IFA_F_DEPRECATED;
1966         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1967                  (prefered_lft != INFINITY_LIFE_TIME))
1968                 prefered_lft = 0x7FFFFFFF/HZ;
1969
1970         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1971
1972         if (!IS_ERR(ifp)) {
1973                 spin_lock_bh(&ifp->lock);
1974                 ifp->valid_lft = valid_lft;
1975                 ifp->prefered_lft = prefered_lft;
1976                 ifp->tstamp = jiffies;
1977                 spin_unlock_bh(&ifp->lock);
1978
1979                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1980                                       jiffies_to_clock_t(valid_lft * HZ), flags);
1981                 /*
1982                  * Note that section 3.1 of RFC 4429 indicates
1983                  * that the Optimistic flag should not be set for
1984                  * manually configured addresses
1985                  */
1986                 addrconf_dad_start(ifp, 0);
1987                 in6_ifa_put(ifp);
1988                 addrconf_verify(0);
1989                 return 0;
1990         }
1991
1992         return PTR_ERR(ifp);
1993 }
1994
1995 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
1996                           int plen)
1997 {
1998         struct inet6_ifaddr *ifp;
1999         struct inet6_dev *idev;
2000         struct net_device *dev;
2001
2002         dev = __dev_get_by_index(net, ifindex);
2003         if (!dev)
2004                 return -ENODEV;
2005
2006         if ((idev = __in6_dev_get(dev)) == NULL)
2007                 return -ENXIO;
2008
2009         read_lock_bh(&idev->lock);
2010         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2011                 if (ifp->prefix_len == plen &&
2012                     ipv6_addr_equal(pfx, &ifp->addr)) {
2013                         in6_ifa_hold(ifp);
2014                         read_unlock_bh(&idev->lock);
2015
2016                         ipv6_del_addr(ifp);
2017
2018                         /* If the last address is deleted administratively,
2019                            disable IPv6 on this interface.
2020                          */
2021                         if (idev->addr_list == NULL)
2022                                 addrconf_ifdown(idev->dev, 1);
2023                         return 0;
2024                 }
2025         }
2026         read_unlock_bh(&idev->lock);
2027         return -EADDRNOTAVAIL;
2028 }
2029
2030
2031 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2032 {
2033         struct in6_ifreq ireq;
2034         int err;
2035
2036         if (!capable(CAP_NET_ADMIN))
2037                 return -EPERM;
2038
2039         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2040                 return -EFAULT;
2041
2042         rtnl_lock();
2043         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2044                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2045                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2046         rtnl_unlock();
2047         return err;
2048 }
2049
2050 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2051 {
2052         struct in6_ifreq ireq;
2053         int err;
2054
2055         if (!capable(CAP_NET_ADMIN))
2056                 return -EPERM;
2057
2058         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2059                 return -EFAULT;
2060
2061         rtnl_lock();
2062         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2063                              ireq.ifr6_prefixlen);
2064         rtnl_unlock();
2065         return err;
2066 }
2067
2068 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2069 static void sit_add_v4_addrs(struct inet6_dev *idev)
2070 {
2071         struct inet6_ifaddr * ifp;
2072         struct in6_addr addr;
2073         struct net_device *dev;
2074         struct net *net = dev_net(idev->dev);
2075         int scope;
2076
2077         ASSERT_RTNL();
2078
2079         memset(&addr, 0, sizeof(struct in6_addr));
2080         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2081
2082         if (idev->dev->flags&IFF_POINTOPOINT) {
2083                 addr.s6_addr32[0] = htonl(0xfe800000);
2084                 scope = IFA_LINK;
2085         } else {
2086                 scope = IPV6_ADDR_COMPATv4;
2087         }
2088
2089         if (addr.s6_addr32[3]) {
2090                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2091                 if (!IS_ERR(ifp)) {
2092                         spin_lock_bh(&ifp->lock);
2093                         ifp->flags &= ~IFA_F_TENTATIVE;
2094                         spin_unlock_bh(&ifp->lock);
2095                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2096                         in6_ifa_put(ifp);
2097                 }
2098                 return;
2099         }
2100
2101         for_each_netdev(net, dev) {
2102                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2103                 if (in_dev && (dev->flags & IFF_UP)) {
2104                         struct in_ifaddr * ifa;
2105
2106                         int flag = scope;
2107
2108                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2109                                 int plen;
2110
2111                                 addr.s6_addr32[3] = ifa->ifa_local;
2112
2113                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2114                                         continue;
2115                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2116                                         if (idev->dev->flags&IFF_POINTOPOINT)
2117                                                 continue;
2118                                         flag |= IFA_HOST;
2119                                 }
2120                                 if (idev->dev->flags&IFF_POINTOPOINT)
2121                                         plen = 64;
2122                                 else
2123                                         plen = 96;
2124
2125                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2126                                                     IFA_F_PERMANENT);
2127                                 if (!IS_ERR(ifp)) {
2128                                         spin_lock_bh(&ifp->lock);
2129                                         ifp->flags &= ~IFA_F_TENTATIVE;
2130                                         spin_unlock_bh(&ifp->lock);
2131                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2132                                         in6_ifa_put(ifp);
2133                                 }
2134                         }
2135                 }
2136         }
2137 }
2138 #endif
2139
2140 static void init_loopback(struct net_device *dev)
2141 {
2142         struct inet6_dev  *idev;
2143         struct inet6_ifaddr * ifp;
2144
2145         /* ::1 */
2146
2147         ASSERT_RTNL();
2148
2149         if ((idev = ipv6_find_idev(dev)) == NULL) {
2150                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2151                 return;
2152         }
2153
2154         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2155         if (!IS_ERR(ifp)) {
2156                 spin_lock_bh(&ifp->lock);
2157                 ifp->flags &= ~IFA_F_TENTATIVE;
2158                 spin_unlock_bh(&ifp->lock);
2159                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2160                 in6_ifa_put(ifp);
2161         }
2162 }
2163
2164 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2165 {
2166         struct inet6_ifaddr * ifp;
2167         u32 addr_flags = IFA_F_PERMANENT;
2168
2169 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2170         if (idev->cnf.optimistic_dad &&
2171             !ipv6_devconf.forwarding)
2172                 addr_flags |= IFA_F_OPTIMISTIC;
2173 #endif
2174
2175
2176         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2177         if (!IS_ERR(ifp)) {
2178                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2179                 addrconf_dad_start(ifp, 0);
2180                 in6_ifa_put(ifp);
2181         }
2182 }
2183
2184 static void addrconf_dev_config(struct net_device *dev)
2185 {
2186         struct in6_addr addr;
2187         struct inet6_dev    * idev;
2188
2189         ASSERT_RTNL();
2190
2191         if ((dev->type != ARPHRD_ETHER) &&
2192             (dev->type != ARPHRD_FDDI) &&
2193             (dev->type != ARPHRD_IEEE802_TR) &&
2194             (dev->type != ARPHRD_ARCNET) &&
2195             (dev->type != ARPHRD_INFINIBAND)) {
2196                 /* Alas, we support only Ethernet autoconfiguration. */
2197                 return;
2198         }
2199
2200         idev = addrconf_add_dev(dev);
2201         if (idev == NULL)
2202                 return;
2203
2204         memset(&addr, 0, sizeof(struct in6_addr));
2205         addr.s6_addr32[0] = htonl(0xFE800000);
2206
2207         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2208                 addrconf_add_linklocal(idev, &addr);
2209 }
2210
2211 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2212 static void addrconf_sit_config(struct net_device *dev)
2213 {
2214         struct inet6_dev *idev;
2215
2216         ASSERT_RTNL();
2217
2218         /*
2219          * Configure the tunnel with one of our IPv4
2220          * addresses... we should configure all of
2221          * our v4 addrs in the tunnel
2222          */
2223
2224         if ((idev = ipv6_find_idev(dev)) == NULL) {
2225                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2226                 return;
2227         }
2228
2229         if (dev->priv_flags & IFF_ISATAP) {
2230                 struct in6_addr addr;
2231
2232                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2233                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2234                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2235                         addrconf_add_linklocal(idev, &addr);
2236                 return;
2237         }
2238
2239         sit_add_v4_addrs(idev);
2240
2241         if (dev->flags&IFF_POINTOPOINT) {
2242                 addrconf_add_mroute(dev);
2243                 addrconf_add_lroute(dev);
2244         } else
2245                 sit_route_add(dev);
2246 }
2247 #endif
2248
2249 static inline int
2250 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2251 {
2252         struct in6_addr lladdr;
2253
2254         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2255                 addrconf_add_linklocal(idev, &lladdr);
2256                 return 0;
2257         }
2258         return -1;
2259 }
2260
2261 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2262 {
2263         struct net_device *link_dev;
2264         struct net *net = dev_net(idev->dev);
2265
2266         /* first try to inherit the link-local address from the link device */
2267         if (idev->dev->iflink &&
2268             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2269                 if (!ipv6_inherit_linklocal(idev, link_dev))
2270                         return;
2271         }
2272         /* then try to inherit it from any device */
2273         for_each_netdev(net, link_dev) {
2274                 if (!ipv6_inherit_linklocal(idev, link_dev))
2275                         return;
2276         }
2277         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2278 }
2279
2280 /*
2281  * Autoconfigure tunnel with a link-local address so routing protocols,
2282  * DHCPv6, MLD etc. can be run over the virtual link
2283  */
2284
2285 static void addrconf_ip6_tnl_config(struct net_device *dev)
2286 {
2287         struct inet6_dev *idev;
2288
2289         ASSERT_RTNL();
2290
2291         if ((idev = addrconf_add_dev(dev)) == NULL) {
2292                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2293                 return;
2294         }
2295         ip6_tnl_add_linklocal(idev);
2296 }
2297
2298 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2299                            void * data)
2300 {
2301         struct net_device *dev = (struct net_device *) data;
2302         struct inet6_dev *idev = __in6_dev_get(dev);
2303         int run_pending = 0;
2304         int err;
2305
2306         switch(event) {
2307         case NETDEV_REGISTER:
2308                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2309                         idev = ipv6_add_dev(dev);
2310                         if (!idev)
2311                                 return notifier_from_errno(-ENOMEM);
2312                 }
2313                 break;
2314         case NETDEV_UP:
2315         case NETDEV_CHANGE:
2316                 if (dev->flags & IFF_SLAVE)
2317                         break;
2318
2319                 if (event == NETDEV_UP) {
2320                         if (!addrconf_qdisc_ok(dev)) {
2321                                 /* device is not ready yet. */
2322                                 printk(KERN_INFO
2323                                         "ADDRCONF(NETDEV_UP): %s: "
2324                                         "link is not ready\n",
2325                                         dev->name);
2326                                 break;
2327                         }
2328
2329                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2330                                 idev = ipv6_add_dev(dev);
2331
2332                         if (idev)
2333                                 idev->if_flags |= IF_READY;
2334                 } else {
2335                         if (!addrconf_qdisc_ok(dev)) {
2336                                 /* device is still not ready. */
2337                                 break;
2338                         }
2339
2340                         if (idev) {
2341                                 if (idev->if_flags & IF_READY) {
2342                                         /* device is already configured. */
2343                                         break;
2344                                 }
2345                                 idev->if_flags |= IF_READY;
2346                         }
2347
2348                         printk(KERN_INFO
2349                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2350                                         "link becomes ready\n",
2351                                         dev->name);
2352
2353                         run_pending = 1;
2354                 }
2355
2356                 switch(dev->type) {
2357 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2358                 case ARPHRD_SIT:
2359                         addrconf_sit_config(dev);
2360                         break;
2361 #endif
2362                 case ARPHRD_TUNNEL6:
2363                         addrconf_ip6_tnl_config(dev);
2364                         break;
2365                 case ARPHRD_LOOPBACK:
2366                         init_loopback(dev);
2367                         break;
2368
2369                 default:
2370                         addrconf_dev_config(dev);
2371                         break;
2372                 }
2373                 if (idev) {
2374                         if (run_pending)
2375                                 addrconf_dad_run(idev);
2376
2377                         /* If the MTU changed during the interface down, when the
2378                            interface up, the changed MTU must be reflected in the
2379                            idev as well as routers.
2380                          */
2381                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2382                                 rt6_mtu_change(dev, dev->mtu);
2383                                 idev->cnf.mtu6 = dev->mtu;
2384                         }
2385                         idev->tstamp = jiffies;
2386                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2387                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2388                            stop IPv6 on this interface.
2389                          */
2390                         if (dev->mtu < IPV6_MIN_MTU)
2391                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2392                 }
2393                 break;
2394
2395         case NETDEV_CHANGEMTU:
2396                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2397                         rt6_mtu_change(dev, dev->mtu);
2398                         idev->cnf.mtu6 = dev->mtu;
2399                         break;
2400                 }
2401
2402                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2403                         idev = ipv6_add_dev(dev);
2404                         if (idev)
2405                                 break;
2406                 }
2407
2408                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2409
2410         case NETDEV_DOWN:
2411         case NETDEV_UNREGISTER:
2412                 /*
2413                  *      Remove all addresses from this interface.
2414                  */
2415                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2416                 break;
2417
2418         case NETDEV_CHANGENAME:
2419                 if (idev) {
2420                         snmp6_unregister_dev(idev);
2421                         addrconf_sysctl_unregister(idev);
2422                         addrconf_sysctl_register(idev);
2423                         err = snmp6_register_dev(idev);
2424                         if (err)
2425                                 return notifier_from_errno(err);
2426                 }
2427                 break;
2428         }
2429
2430         return NOTIFY_OK;
2431 }
2432
2433 /*
2434  *      addrconf module should be notified of a device going up
2435  */
2436 static struct notifier_block ipv6_dev_notf = {
2437         .notifier_call = addrconf_notify,
2438         .priority = 0
2439 };
2440
2441 static int addrconf_ifdown(struct net_device *dev, int how)
2442 {
2443         struct inet6_dev *idev;
2444         struct inet6_ifaddr *ifa, **bifa;
2445         struct net *net = dev_net(dev);
2446         int i;
2447
2448         ASSERT_RTNL();
2449
2450         if (dev == init_net.loopback_dev && how == 1)
2451                 how = 0;
2452
2453         rt6_ifdown(net, dev);
2454         neigh_ifdown(&nd_tbl, dev);
2455
2456         idev = __in6_dev_get(dev);
2457         if (idev == NULL)
2458                 return -ENODEV;
2459
2460         /* Step 1: remove reference to ipv6 device from parent device.
2461                    Do not dev_put!
2462          */
2463         if (how == 1) {
2464                 idev->dead = 1;
2465
2466                 /* protected by rtnl_lock */
2467                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2468
2469                 /* Step 1.5: remove snmp6 entry */
2470                 snmp6_unregister_dev(idev);
2471
2472         }
2473
2474         /* Step 2: clear hash table */
2475         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2476                 bifa = &inet6_addr_lst[i];
2477
2478                 write_lock_bh(&addrconf_hash_lock);
2479                 while ((ifa = *bifa) != NULL) {
2480                         if (ifa->idev == idev) {
2481                                 *bifa = ifa->lst_next;
2482                                 ifa->lst_next = NULL;
2483                                 addrconf_del_timer(ifa);
2484                                 in6_ifa_put(ifa);
2485                                 continue;
2486                         }
2487                         bifa = &ifa->lst_next;
2488                 }
2489                 write_unlock_bh(&addrconf_hash_lock);
2490         }
2491
2492         write_lock_bh(&idev->lock);
2493
2494         /* Step 3: clear flags for stateless addrconf */
2495         if (how != 1)
2496                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2497
2498         /* Step 4: clear address list */
2499 #ifdef CONFIG_IPV6_PRIVACY
2500         if (how == 1 && del_timer(&idev->regen_timer))
2501                 in6_dev_put(idev);
2502
2503         /* clear tempaddr list */
2504         while ((ifa = idev->tempaddr_list) != NULL) {
2505                 idev->tempaddr_list = ifa->tmp_next;
2506                 ifa->tmp_next = NULL;
2507                 ifa->dead = 1;
2508                 write_unlock_bh(&idev->lock);
2509                 spin_lock_bh(&ifa->lock);
2510
2511                 if (ifa->ifpub) {
2512                         in6_ifa_put(ifa->ifpub);
2513                         ifa->ifpub = NULL;
2514                 }
2515                 spin_unlock_bh(&ifa->lock);
2516                 in6_ifa_put(ifa);
2517                 write_lock_bh(&idev->lock);
2518         }
2519 #endif
2520         while ((ifa = idev->addr_list) != NULL) {
2521                 idev->addr_list = ifa->if_next;
2522                 ifa->if_next = NULL;
2523                 ifa->dead = 1;
2524                 addrconf_del_timer(ifa);
2525                 write_unlock_bh(&idev->lock);
2526
2527                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2528                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2529                 in6_ifa_put(ifa);
2530
2531                 write_lock_bh(&idev->lock);
2532         }
2533         write_unlock_bh(&idev->lock);
2534
2535         /* Step 5: Discard multicast list */
2536
2537         if (how == 1)
2538                 ipv6_mc_destroy_dev(idev);
2539         else
2540                 ipv6_mc_down(idev);
2541
2542         idev->tstamp = jiffies;
2543
2544         /* Shot the device (if unregistered) */
2545
2546         if (how == 1) {
2547                 addrconf_sysctl_unregister(idev);
2548                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2549                 neigh_ifdown(&nd_tbl, dev);
2550                 in6_dev_put(idev);
2551         }
2552         return 0;
2553 }
2554
2555 static void addrconf_rs_timer(unsigned long data)
2556 {
2557         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2558
2559         if (ifp->idev->cnf.forwarding)
2560                 goto out;
2561
2562         if (ifp->idev->if_flags & IF_RA_RCVD) {
2563                 /*
2564                  *      Announcement received after solicitation
2565                  *      was sent
2566                  */
2567                 goto out;
2568         }
2569
2570         spin_lock(&ifp->lock);
2571         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2572                 struct in6_addr all_routers;
2573
2574                 /* The wait after the last probe can be shorter */
2575                 addrconf_mod_timer(ifp, AC_RS,
2576                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2577                                    ifp->idev->cnf.rtr_solicit_delay :
2578                                    ifp->idev->cnf.rtr_solicit_interval);
2579                 spin_unlock(&ifp->lock);
2580
2581                 ipv6_addr_all_routers(&all_routers);
2582
2583                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2584         } else {
2585                 spin_unlock(&ifp->lock);
2586                 /*
2587                  * Note: we do not support deprecated "all on-link"
2588                  * assumption any longer.
2589                  */
2590                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2591                        ifp->idev->dev->name);
2592         }
2593
2594 out:
2595         in6_ifa_put(ifp);
2596 }
2597
2598 /*
2599  *      Duplicate Address Detection
2600  */
2601 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2602 {
2603         unsigned long rand_num;
2604         struct inet6_dev *idev = ifp->idev;
2605
2606         if (ifp->flags & IFA_F_OPTIMISTIC)
2607                 rand_num = 0;
2608         else
2609                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2610
2611         ifp->probes = idev->cnf.dad_transmits;
2612         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2613 }
2614
2615 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2616 {
2617         struct inet6_dev *idev = ifp->idev;
2618         struct net_device *dev = idev->dev;
2619
2620         addrconf_join_solict(dev, &ifp->addr);
2621
2622         net_srandom(ifp->addr.s6_addr32[3]);
2623
2624         read_lock_bh(&idev->lock);
2625         if (ifp->dead)
2626                 goto out;
2627         spin_lock_bh(&ifp->lock);
2628
2629         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2630             !(ifp->flags&IFA_F_TENTATIVE) ||
2631             ifp->flags & IFA_F_NODAD) {
2632                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2633                 spin_unlock_bh(&ifp->lock);
2634                 read_unlock_bh(&idev->lock);
2635
2636                 addrconf_dad_completed(ifp);
2637                 return;
2638         }
2639
2640         if (!(idev->if_flags & IF_READY)) {
2641                 spin_unlock_bh(&ifp->lock);
2642                 read_unlock_bh(&idev->lock);
2643                 /*
2644                  * If the defice is not ready:
2645                  * - keep it tentative if it is a permanent address.
2646                  * - otherwise, kill it.
2647                  */
2648                 in6_ifa_hold(ifp);
2649                 addrconf_dad_stop(ifp);
2650                 return;
2651         }
2652
2653         /*
2654          * Optimistic nodes can start receiving
2655          * Frames right away
2656          */
2657         if(ifp->flags & IFA_F_OPTIMISTIC)
2658                 ip6_ins_rt(ifp->rt);
2659
2660         addrconf_dad_kick(ifp);
2661         spin_unlock_bh(&ifp->lock);
2662 out:
2663         read_unlock_bh(&idev->lock);
2664 }
2665
2666 static void addrconf_dad_timer(unsigned long data)
2667 {
2668         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2669         struct inet6_dev *idev = ifp->idev;
2670         struct in6_addr unspec;
2671         struct in6_addr mcaddr;
2672
2673         read_lock_bh(&idev->lock);
2674         if (idev->dead) {
2675                 read_unlock_bh(&idev->lock);
2676                 goto out;
2677         }
2678         spin_lock_bh(&ifp->lock);
2679         if (ifp->probes == 0) {
2680                 /*
2681                  * DAD was successful
2682                  */
2683
2684                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2685                 spin_unlock_bh(&ifp->lock);
2686                 read_unlock_bh(&idev->lock);
2687
2688                 addrconf_dad_completed(ifp);
2689
2690                 goto out;
2691         }
2692
2693         ifp->probes--;
2694         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2695         spin_unlock_bh(&ifp->lock);
2696         read_unlock_bh(&idev->lock);
2697
2698         /* send a neighbour solicitation for our addr */
2699         memset(&unspec, 0, sizeof(unspec));
2700         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2701         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2702 out:
2703         in6_ifa_put(ifp);
2704 }
2705
2706 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2707 {
2708         struct net_device *     dev = ifp->idev->dev;
2709
2710         /*
2711          *      Configure the address for reception. Now it is valid.
2712          */
2713
2714         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2715
2716         /* If added prefix is link local and forwarding is off,
2717            start sending router solicitations.
2718          */
2719
2720         if (ifp->idev->cnf.forwarding == 0 &&
2721             ifp->idev->cnf.rtr_solicits > 0 &&
2722             (dev->flags&IFF_LOOPBACK) == 0 &&
2723             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2724                 struct in6_addr all_routers;
2725
2726                 ipv6_addr_all_routers(&all_routers);
2727
2728                 /*
2729                  *      If a host as already performed a random delay
2730                  *      [...] as part of DAD [...] there is no need
2731                  *      to delay again before sending the first RS
2732                  */
2733                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2734
2735                 spin_lock_bh(&ifp->lock);
2736                 ifp->probes = 1;
2737                 ifp->idev->if_flags |= IF_RS_SENT;
2738                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2739                 spin_unlock_bh(&ifp->lock);
2740         }
2741 }
2742
2743 static void addrconf_dad_run(struct inet6_dev *idev) {
2744         struct inet6_ifaddr *ifp;
2745
2746         read_lock_bh(&idev->lock);
2747         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2748                 spin_lock_bh(&ifp->lock);
2749                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2750                         spin_unlock_bh(&ifp->lock);
2751                         continue;
2752                 }
2753                 spin_unlock_bh(&ifp->lock);
2754                 addrconf_dad_kick(ifp);
2755         }
2756         read_unlock_bh(&idev->lock);
2757 }
2758
2759 #ifdef CONFIG_PROC_FS
2760 struct if6_iter_state {
2761         struct seq_net_private p;
2762         int bucket;
2763 };
2764
2765 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2766 {
2767         struct inet6_ifaddr *ifa = NULL;
2768         struct if6_iter_state *state = seq->private;
2769         struct net *net = state->p.net;
2770
2771         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2772                 ifa = inet6_addr_lst[state->bucket];
2773
2774                 while (ifa && dev_net(ifa->idev->dev) != net)
2775                         ifa = ifa->lst_next;
2776                 if (ifa)
2777                         break;
2778         }
2779         return ifa;
2780 }
2781
2782 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2783 {
2784         struct if6_iter_state *state = seq->private;
2785         struct net *net = state->p.net;
2786
2787         ifa = ifa->lst_next;
2788 try_again:
2789         if (ifa) {
2790                 if (dev_net(ifa->idev->dev) != net) {
2791                         ifa = ifa->lst_next;
2792                         goto try_again;
2793                 }
2794         }
2795
2796         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2797                 ifa = inet6_addr_lst[state->bucket];
2798                 goto try_again;
2799         }
2800
2801         return ifa;
2802 }
2803
2804 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2805 {
2806         struct inet6_ifaddr *ifa = if6_get_first(seq);
2807
2808         if (ifa)
2809                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2810                         --pos;
2811         return pos ? NULL : ifa;
2812 }
2813
2814 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2815         __acquires(addrconf_hash_lock)
2816 {
2817         read_lock_bh(&addrconf_hash_lock);
2818         return if6_get_idx(seq, *pos);
2819 }
2820
2821 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2822 {
2823         struct inet6_ifaddr *ifa;
2824
2825         ifa = if6_get_next(seq, v);
2826         ++*pos;
2827         return ifa;
2828 }
2829
2830 static void if6_seq_stop(struct seq_file *seq, void *v)
2831         __releases(addrconf_hash_lock)
2832 {
2833         read_unlock_bh(&addrconf_hash_lock);
2834 }
2835
2836 static int if6_seq_show(struct seq_file *seq, void *v)
2837 {
2838         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2839         seq_printf(seq,
2840                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2841                    NIP6(ifp->addr),
2842                    ifp->idev->dev->ifindex,
2843                    ifp->prefix_len,
2844                    ifp->scope,
2845                    ifp->flags,
2846                    ifp->idev->dev->name);
2847         return 0;
2848 }
2849
2850 static const struct seq_operations if6_seq_ops = {
2851         .start  = if6_seq_start,
2852         .next   = if6_seq_next,
2853         .show   = if6_seq_show,
2854         .stop   = if6_seq_stop,
2855 };
2856
2857 static int if6_seq_open(struct inode *inode, struct file *file)
2858 {
2859         return seq_open_net(inode, file, &if6_seq_ops,
2860                             sizeof(struct if6_iter_state));
2861 }
2862
2863 static const struct file_operations if6_fops = {
2864         .owner          = THIS_MODULE,
2865         .open           = if6_seq_open,
2866         .read           = seq_read,
2867         .llseek         = seq_lseek,
2868         .release        = seq_release_net,
2869 };
2870
2871 static int if6_proc_net_init(struct net *net)
2872 {
2873         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2874                 return -ENOMEM;
2875         return 0;
2876 }
2877
2878 static void if6_proc_net_exit(struct net *net)
2879 {
2880        proc_net_remove(net, "if_inet6");
2881 }
2882
2883 static struct pernet_operations if6_proc_net_ops = {
2884        .init = if6_proc_net_init,
2885        .exit = if6_proc_net_exit,
2886 };
2887
2888 int __init if6_proc_init(void)
2889 {
2890         return register_pernet_subsys(&if6_proc_net_ops);
2891 }
2892
2893 void if6_proc_exit(void)
2894 {
2895         unregister_pernet_subsys(&if6_proc_net_ops);
2896 }
2897 #endif  /* CONFIG_PROC_FS */
2898
2899 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2900 /* Check if address is a home address configured on any interface. */
2901 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2902 {
2903         int ret = 0;
2904         struct inet6_ifaddr * ifp;
2905         u8 hash = ipv6_addr_hash(addr);
2906         read_lock_bh(&addrconf_hash_lock);
2907         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2908                 if (dev_net(ifp->idev->dev) != net)
2909                         continue;
2910                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2911                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2912                         ret = 1;
2913                         break;
2914                 }
2915         }
2916         read_unlock_bh(&addrconf_hash_lock);
2917         return ret;
2918 }
2919 #endif
2920
2921 /*
2922  *      Periodic address status verification
2923  */
2924
2925 static void addrconf_verify(unsigned long foo)
2926 {
2927         struct inet6_ifaddr *ifp;
2928         unsigned long now, next;
2929         int i;
2930
2931         spin_lock_bh(&addrconf_verify_lock);
2932         now = jiffies;
2933         next = now + ADDR_CHECK_FREQUENCY;
2934
2935         del_timer(&addr_chk_timer);
2936
2937         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2938
2939 restart:
2940                 read_lock(&addrconf_hash_lock);
2941                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2942                         unsigned long age;
2943 #ifdef CONFIG_IPV6_PRIVACY
2944                         unsigned long regen_advance;
2945 #endif
2946
2947                         if (ifp->flags & IFA_F_PERMANENT)
2948                                 continue;
2949
2950                         spin_lock(&ifp->lock);
2951                         age = (now - ifp->tstamp) / HZ;
2952
2953 #ifdef CONFIG_IPV6_PRIVACY
2954                         regen_advance = ifp->idev->cnf.regen_max_retry *
2955                                         ifp->idev->cnf.dad_transmits *
2956                                         ifp->idev->nd_parms->retrans_time / HZ;
2957 #endif
2958
2959                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2960                             age >= ifp->valid_lft) {
2961                                 spin_unlock(&ifp->lock);
2962                                 in6_ifa_hold(ifp);
2963                                 read_unlock(&addrconf_hash_lock);
2964                                 ipv6_del_addr(ifp);
2965                                 goto restart;
2966                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2967                                 spin_unlock(&ifp->lock);
2968                                 continue;
2969                         } else if (age >= ifp->prefered_lft) {
2970                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2971                                 int deprecate = 0;
2972
2973                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2974                                         deprecate = 1;
2975                                         ifp->flags |= IFA_F_DEPRECATED;
2976                                 }
2977
2978                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2979                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2980
2981                                 spin_unlock(&ifp->lock);
2982
2983                                 if (deprecate) {
2984                                         in6_ifa_hold(ifp);
2985                                         read_unlock(&addrconf_hash_lock);
2986
2987                                         ipv6_ifa_notify(0, ifp);
2988                                         in6_ifa_put(ifp);
2989                                         goto restart;
2990                                 }
2991 #ifdef CONFIG_IPV6_PRIVACY
2992                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2993                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2994                                 if (age >= ifp->prefered_lft - regen_advance) {
2995                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2996                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2997                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2998                                         if (!ifp->regen_count && ifpub) {
2999                                                 ifp->regen_count++;
3000                                                 in6_ifa_hold(ifp);
3001                                                 in6_ifa_hold(ifpub);
3002                                                 spin_unlock(&ifp->lock);
3003                                                 read_unlock(&addrconf_hash_lock);
3004                                                 spin_lock(&ifpub->lock);
3005                                                 ifpub->regen_count = 0;
3006                                                 spin_unlock(&ifpub->lock);
3007                                                 ipv6_create_tempaddr(ifpub, ifp);
3008                                                 in6_ifa_put(ifpub);
3009                                                 in6_ifa_put(ifp);
3010                                                 goto restart;
3011                                         }
3012                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3013                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3014                                 spin_unlock(&ifp->lock);
3015 #endif
3016                         } else {
3017                                 /* ifp->prefered_lft <= ifp->valid_lft */
3018                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3019                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3020                                 spin_unlock(&ifp->lock);
3021                         }
3022                 }
3023                 read_unlock(&addrconf_hash_lock);
3024         }
3025
3026         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3027         add_timer(&addr_chk_timer);
3028         spin_unlock_bh(&addrconf_verify_lock);
3029 }
3030
3031 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3032 {
3033         struct in6_addr *pfx = NULL;
3034
3035         if (addr)
3036                 pfx = nla_data(addr);
3037
3038         if (local) {
3039                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3040                         pfx = NULL;
3041                 else
3042                         pfx = nla_data(local);
3043         }
3044
3045         return pfx;
3046 }
3047
3048 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3049         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3050         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3051         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3052 };
3053
3054 static int
3055 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3056 {
3057         struct net *net = sock_net(skb->sk);
3058         struct ifaddrmsg *ifm;
3059         struct nlattr *tb[IFA_MAX+1];
3060         struct in6_addr *pfx;
3061         int err;
3062
3063         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3064         if (err < 0)
3065                 return err;
3066
3067         ifm = nlmsg_data(nlh);
3068         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3069         if (pfx == NULL)
3070                 return -EINVAL;
3071
3072         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3073 }
3074
3075 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3076                              u32 prefered_lft, u32 valid_lft)
3077 {
3078         u32 flags = RTF_EXPIRES;
3079
3080         if (!valid_lft || (prefered_lft > valid_lft))
3081                 return -EINVAL;
3082
3083         if (valid_lft == INFINITY_LIFE_TIME) {
3084                 ifa_flags |= IFA_F_PERMANENT;
3085                 flags = 0;
3086         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3087                 valid_lft = 0x7FFFFFFF/HZ;
3088
3089         if (prefered_lft == 0)
3090                 ifa_flags |= IFA_F_DEPRECATED;
3091         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3092                  (prefered_lft != INFINITY_LIFE_TIME))
3093                 prefered_lft = 0x7FFFFFFF/HZ;
3094
3095         spin_lock_bh(&ifp->lock);
3096         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3097         ifp->tstamp = jiffies;
3098         ifp->valid_lft = valid_lft;
3099         ifp->prefered_lft = prefered_lft;
3100
3101         spin_unlock_bh(&ifp->lock);
3102         if (!(ifp->flags&IFA_F_TENTATIVE))
3103                 ipv6_ifa_notify(0, ifp);
3104
3105         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3106                               jiffies_to_clock_t(valid_lft * HZ), flags);
3107         addrconf_verify(0);
3108
3109         return 0;
3110 }
3111
3112 static int
3113 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3114 {
3115         struct net *net = sock_net(skb->sk);
3116         struct ifaddrmsg *ifm;
3117         struct nlattr *tb[IFA_MAX+1];
3118         struct in6_addr *pfx;
3119         struct inet6_ifaddr *ifa;
3120         struct net_device *dev;
3121         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3122         u8 ifa_flags;
3123         int err;
3124
3125         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3126         if (err < 0)
3127                 return err;
3128
3129         ifm = nlmsg_data(nlh);
3130         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3131         if (pfx == NULL)
3132                 return -EINVAL;
3133
3134         if (tb[IFA_CACHEINFO]) {
3135                 struct ifa_cacheinfo *ci;
3136
3137                 ci = nla_data(tb[IFA_CACHEINFO]);
3138                 valid_lft = ci->ifa_valid;
3139                 preferred_lft = ci->ifa_prefered;
3140         } else {
3141                 preferred_lft = INFINITY_LIFE_TIME;
3142                 valid_lft = INFINITY_LIFE_TIME;
3143         }
3144
3145         dev =  __dev_get_by_index(net, ifm->ifa_index);
3146         if (dev == NULL)
3147                 return -ENODEV;
3148
3149         /* We ignore other flags so far. */
3150         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3151
3152         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3153         if (ifa == NULL) {
3154                 /*
3155                  * It would be best to check for !NLM_F_CREATE here but
3156                  * userspace alreay relies on not having to provide this.
3157                  */
3158                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3159                                       ifm->ifa_prefixlen, ifa_flags,
3160                                       preferred_lft, valid_lft);
3161         }
3162
3163         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3164             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3165                 err = -EEXIST;
3166         else
3167                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3168
3169         in6_ifa_put(ifa);
3170
3171         return err;
3172 }
3173
3174 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3175                           u8 scope, int ifindex)
3176 {
3177         struct ifaddrmsg *ifm;
3178
3179         ifm = nlmsg_data(nlh);
3180         ifm->ifa_family = AF_INET6;
3181         ifm->ifa_prefixlen = prefixlen;
3182         ifm->ifa_flags = flags;
3183         ifm->ifa_scope = scope;
3184         ifm->ifa_index = ifindex;
3185 }
3186
3187 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3188                          unsigned long tstamp, u32 preferred, u32 valid)
3189 {
3190         struct ifa_cacheinfo ci;
3191
3192         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3193                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3194         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3195                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3196         ci.ifa_prefered = preferred;
3197         ci.ifa_valid = valid;
3198
3199         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3200 }
3201
3202 static inline int rt_scope(int ifa_scope)
3203 {
3204         if (ifa_scope & IFA_HOST)
3205                 return RT_SCOPE_HOST;
3206         else if (ifa_scope & IFA_LINK)
3207                 return RT_SCOPE_LINK;
3208         else if (ifa_scope & IFA_SITE)
3209                 return RT_SCOPE_SITE;
3210         else
3211                 return RT_SCOPE_UNIVERSE;
3212 }
3213
3214 static inline int inet6_ifaddr_msgsize(void)
3215 {
3216         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3217                + nla_total_size(16) /* IFA_ADDRESS */
3218                + nla_total_size(sizeof(struct ifa_cacheinfo));
3219 }
3220
3221 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3222                              u32 pid, u32 seq, int event, unsigned int flags)
3223 {
3224         struct nlmsghdr  *nlh;
3225         u32 preferred, valid;
3226
3227         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3228         if (nlh == NULL)
3229                 return -EMSGSIZE;
3230
3231         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3232                       ifa->idev->dev->ifindex);
3233
3234         if (!(ifa->flags&IFA_F_PERMANENT)) {
3235                 preferred = ifa->prefered_lft;
3236                 valid = ifa->valid_lft;
3237                 if (preferred != INFINITY_LIFE_TIME) {
3238                         long tval = (jiffies - ifa->tstamp)/HZ;
3239                         preferred -= tval;
3240                         if (valid != INFINITY_LIFE_TIME)
3241                                 valid -= tval;
3242                 }
3243         } else {
3244                 preferred = INFINITY_LIFE_TIME;
3245                 valid = INFINITY_LIFE_TIME;
3246         }
3247
3248         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3249             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3250                 nlmsg_cancel(skb, nlh);
3251                 return -EMSGSIZE;
3252         }
3253
3254         return nlmsg_end(skb, nlh);
3255 }
3256
3257 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3258                                 u32 pid, u32 seq, int event, u16 flags)
3259 {
3260         struct nlmsghdr  *nlh;
3261         u8 scope = RT_SCOPE_UNIVERSE;
3262         int ifindex = ifmca->idev->dev->ifindex;
3263
3264         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3265                 scope = RT_SCOPE_SITE;
3266
3267         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3268         if (nlh == NULL)
3269                 return -EMSGSIZE;
3270
3271         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3272         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3273             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3274                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3275                 nlmsg_cancel(skb, nlh);
3276                 return -EMSGSIZE;
3277         }
3278
3279         return nlmsg_end(skb, nlh);
3280 }
3281
3282 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3283                                 u32 pid, u32 seq, int event, unsigned int flags)
3284 {
3285         struct nlmsghdr  *nlh;
3286         u8 scope = RT_SCOPE_UNIVERSE;
3287         int ifindex = ifaca->aca_idev->dev->ifindex;
3288
3289         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3290                 scope = RT_SCOPE_SITE;
3291
3292         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3293         if (nlh == NULL)
3294                 return -EMSGSIZE;
3295
3296         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3297         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3298             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3299                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3300                 nlmsg_cancel(skb, nlh);
3301                 return -EMSGSIZE;
3302         }
3303
3304         return nlmsg_end(skb, nlh);
3305 }
3306
3307 enum addr_type_t
3308 {
3309         UNICAST_ADDR,
3310         MULTICAST_ADDR,
3311         ANYCAST_ADDR,
3312 };
3313
3314 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3315                            enum addr_type_t type)
3316 {
3317         int idx, ip_idx;
3318         int s_idx, s_ip_idx;
3319         int err = 1;
3320         struct net_device *dev;
3321         struct inet6_dev *idev = NULL;
3322         struct inet6_ifaddr *ifa;
3323         struct ifmcaddr6 *ifmca;
3324         struct ifacaddr6 *ifaca;
3325         struct net *net = sock_net(skb->sk);
3326
3327         s_idx = cb->args[0];
3328         s_ip_idx = ip_idx = cb->args[1];
3329
3330         idx = 0;
3331         for_each_netdev(net, dev) {
3332                 if (idx < s_idx)
3333                         goto cont;
3334                 if (idx > s_idx)
3335                         s_ip_idx = 0;
3336                 ip_idx = 0;
3337                 if ((idev = in6_dev_get(dev)) == NULL)
3338                         goto cont;
3339                 read_lock_bh(&idev->lock);
3340                 switch (type) {
3341                 case UNICAST_ADDR:
3342                         /* unicast address incl. temp addr */
3343                         for (ifa = idev->addr_list; ifa;
3344                              ifa = ifa->if_next, ip_idx++) {
3345                                 if (ip_idx < s_ip_idx)
3346                                         continue;
3347                                 err = inet6_fill_ifaddr(skb, ifa,
3348                                                         NETLINK_CB(cb->skb).pid,
3349                                                         cb->nlh->nlmsg_seq,
3350                                                         RTM_NEWADDR,
3351                                                         NLM_F_MULTI);
3352                         }
3353                         break;
3354                 case MULTICAST_ADDR:
3355                         /* multicast address */
3356                         for (ifmca = idev->mc_list; ifmca;
3357                              ifmca = ifmca->next, ip_idx++) {
3358                                 if (ip_idx < s_ip_idx)
3359                                         continue;
3360                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3361                                                           NETLINK_CB(cb->skb).pid,
3362                                                           cb->nlh->nlmsg_seq,
3363                                                           RTM_GETMULTICAST,
3364                                                           NLM_F_MULTI);
3365                         }
3366                         break;
3367                 case ANYCAST_ADDR:
3368                         /* anycast address */
3369                         for (ifaca = idev->ac_list; ifaca;
3370                              ifaca = ifaca->aca_next, ip_idx++) {
3371                                 if (ip_idx < s_ip_idx)
3372                                         continue;
3373                                 err = inet6_fill_ifacaddr(skb, ifaca,
3374                                                           NETLINK_CB(cb->skb).pid,
3375                                                           cb->nlh->nlmsg_seq,
3376                                                           RTM_GETANYCAST,
3377                                                           NLM_F_MULTI);
3378                         }
3379                         break;
3380                 default:
3381                         break;
3382                 }
3383                 read_unlock_bh(&idev->lock);
3384                 in6_dev_put(idev);
3385
3386                 if (err <= 0)
3387                         break;
3388 cont:
3389                 idx++;
3390         }
3391         cb->args[0] = idx;
3392         cb->args[1] = ip_idx;
3393         return skb->len;
3394 }
3395
3396 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3397 {
3398         enum addr_type_t type = UNICAST_ADDR;
3399
3400         return inet6_dump_addr(skb, cb, type);
3401 }
3402
3403 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3404 {
3405         enum addr_type_t type = MULTICAST_ADDR;
3406
3407         return inet6_dump_addr(skb, cb, type);
3408 }
3409
3410
3411 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3412 {
3413         enum addr_type_t type = ANYCAST_ADDR;
3414
3415         return inet6_dump_addr(skb, cb, type);
3416 }
3417
3418 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3419                              void *arg)
3420 {
3421         struct net *net = sock_net(in_skb->sk);
3422         struct ifaddrmsg *ifm;
3423         struct nlattr *tb[IFA_MAX+1];
3424         struct in6_addr *addr = NULL;
3425         struct net_device *dev = NULL;
3426         struct inet6_ifaddr *ifa;
3427         struct sk_buff *skb;
3428         int err;
3429
3430         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3431         if (err < 0)
3432                 goto errout;
3433
3434         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3435         if (addr == NULL) {
3436                 err = -EINVAL;
3437                 goto errout;
3438         }
3439
3440         ifm = nlmsg_data(nlh);
3441         if (ifm->ifa_index)
3442                 dev = __dev_get_by_index(net, ifm->ifa_index);
3443
3444         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3445                 err = -EADDRNOTAVAIL;
3446                 goto errout;
3447         }
3448
3449         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3450                 err = -ENOBUFS;
3451                 goto errout_ifa;
3452         }
3453
3454         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3455                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3456         if (err < 0) {
3457                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3458                 WARN_ON(err == -EMSGSIZE);
3459                 kfree_skb(skb);
3460                 goto errout_ifa;
3461         }
3462         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3463 errout_ifa:
3464         in6_ifa_put(ifa);
3465 errout:
3466         return err;
3467 }
3468
3469 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3470 {
3471         struct sk_buff *skb;
3472         struct net *net = dev_net(ifa->idev->dev);
3473         int err = -ENOBUFS;
3474
3475         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3476         if (skb == NULL)
3477                 goto errout;
3478
3479         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3480         if (err < 0) {
3481                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3482                 WARN_ON(err == -EMSGSIZE);
3483                 kfree_skb(skb);
3484                 goto errout;
3485         }
3486         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3487 errout:
3488         if (err < 0)
3489                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3490 }
3491
3492 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3493                                 __s32 *array, int bytes)
3494 {
3495         BUG_ON(bytes < (DEVCONF_MAX * 4));
3496
3497         memset(array, 0, bytes);
3498         array[DEVCONF_FORWARDING] = cnf->forwarding;
3499         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3500         array[DEVCONF_MTU6] = cnf->mtu6;
3501         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3502         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3503         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3504         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3505         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3506         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3507         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3508         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3509 #ifdef CONFIG_IPV6_PRIVACY
3510         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3511         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3512         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3513         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3514         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3515 #endif
3516         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3517         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3518         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3519 #ifdef CONFIG_IPV6_ROUTER_PREF
3520         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3521         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3522 #ifdef CONFIG_IPV6_ROUTE_INFO
3523         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3524 #endif
3525 #endif
3526         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3527         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3528 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3529         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3530 #endif
3531 }
3532
3533 static inline size_t inet6_if_nlmsg_size(void)
3534 {
3535         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3536                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3537                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3538                + nla_total_size(4) /* IFLA_MTU */
3539                + nla_total_size(4) /* IFLA_LINK */
3540                + nla_total_size( /* IFLA_PROTINFO */
3541                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3542                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3543                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3544                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3545                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3546                  );
3547 }
3548
3549 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3550                                       int bytes)
3551 {
3552         int i;
3553         int pad = bytes - sizeof(u64) * items;
3554         BUG_ON(pad < 0);
3555
3556         /* Use put_unaligned() because stats may not be aligned for u64. */
3557         put_unaligned(items, &stats[0]);
3558         for (i = 1; i < items; i++)
3559                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3560
3561         memset(&stats[items], 0, pad);
3562 }
3563
3564 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3565                              int bytes)
3566 {
3567         switch(attrtype) {
3568         case IFLA_INET6_STATS:
3569                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3570                 break;
3571         case IFLA_INET6_ICMP6STATS:
3572                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3573                 break;
3574         }
3575 }
3576
3577 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3578                              u32 pid, u32 seq, int event, unsigned int flags)
3579 {
3580         struct net_device *dev = idev->dev;
3581         struct nlattr *nla;
3582         struct ifinfomsg *hdr;
3583         struct nlmsghdr *nlh;
3584         void *protoinfo;
3585         struct ifla_cacheinfo ci;
3586
3587         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3588         if (nlh == NULL)
3589                 return -EMSGSIZE;
3590
3591         hdr = nlmsg_data(nlh);
3592         hdr->ifi_family = AF_INET6;
3593         hdr->__ifi_pad = 0;
3594         hdr->ifi_type = dev->type;
3595         hdr->ifi_index = dev->ifindex;
3596         hdr->ifi_flags = dev_get_flags(dev);
3597         hdr->ifi_change = 0;
3598
3599         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3600
3601         if (dev->addr_len)
3602                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3603
3604         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3605         if (dev->ifindex != dev->iflink)
3606                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3607
3608         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3609         if (protoinfo == NULL)
3610                 goto nla_put_failure;
3611
3612         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3613
3614         ci.max_reasm_len = IPV6_MAXPLEN;
3615         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3616                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3617         ci.reachable_time = idev->nd_parms->reachable_time;
3618         ci.retrans_time = idev->nd_parms->retrans_time;
3619         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3620
3621         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3622         if (nla == NULL)
3623                 goto nla_put_failure;
3624         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3625
3626         /* XXX - MC not implemented */
3627
3628         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3629         if (nla == NULL)
3630                 goto nla_put_failure;
3631         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3632
3633         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3634         if (nla == NULL)
3635                 goto nla_put_failure;
3636         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3637
3638         nla_nest_end(skb, protoinfo);
3639         return nlmsg_end(skb, nlh);
3640
3641 nla_put_failure:
3642         nlmsg_cancel(skb, nlh);
3643         return -EMSGSIZE;
3644 }
3645
3646 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3647 {
3648         struct net *net = sock_net(skb->sk);
3649         int idx, err;
3650         int s_idx = cb->args[0];
3651         struct net_device *dev;
3652         struct inet6_dev *idev;
3653
3654         read_lock(&dev_base_lock);
3655         idx = 0;
3656         for_each_netdev(net, dev) {
3657                 if (idx < s_idx)
3658                         goto cont;
3659                 if ((idev = in6_dev_get(dev)) == NULL)
3660                         goto cont;
3661                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3662                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3663                 in6_dev_put(idev);
3664                 if (err <= 0)
3665                         break;
3666 cont:
3667                 idx++;
3668         }
3669         read_unlock(&dev_base_lock);
3670         cb->args[0] = idx;
3671
3672         return skb->len;
3673 }
3674
3675 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3676 {
3677         struct sk_buff *skb;
3678         struct net *net = dev_net(idev->dev);
3679         int err = -ENOBUFS;
3680
3681         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3682         if (skb == NULL)
3683                 goto errout;
3684
3685         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3686         if (err < 0) {
3687                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3688                 WARN_ON(err == -EMSGSIZE);
3689                 kfree_skb(skb);
3690                 goto errout;
3691         }
3692         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3693 errout:
3694         if (err < 0)
3695                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3696 }
3697
3698 static inline size_t inet6_prefix_nlmsg_size(void)
3699 {
3700         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3701                + nla_total_size(sizeof(struct in6_addr))
3702                + nla_total_size(sizeof(struct prefix_cacheinfo));
3703 }
3704
3705 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3706                              struct prefix_info *pinfo, u32 pid, u32 seq,
3707                              int event, unsigned int flags)
3708 {
3709         struct prefixmsg *pmsg;
3710         struct nlmsghdr *nlh;
3711         struct prefix_cacheinfo ci;
3712
3713         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3714         if (nlh == NULL)
3715                 return -EMSGSIZE;
3716
3717         pmsg = nlmsg_data(nlh);
3718         pmsg->prefix_family = AF_INET6;
3719         pmsg->prefix_pad1 = 0;
3720         pmsg->prefix_pad2 = 0;
3721         pmsg->prefix_ifindex = idev->dev->ifindex;
3722         pmsg->prefix_len = pinfo->prefix_len;
3723         pmsg->prefix_type = pinfo->type;
3724         pmsg->prefix_pad3 = 0;
3725         pmsg->prefix_flags = 0;
3726         if (pinfo->onlink)
3727                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3728         if (pinfo->autoconf)
3729                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3730
3731         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3732
3733         ci.preferred_time = ntohl(pinfo->prefered);
3734         ci.valid_time = ntohl(pinfo->valid);
3735         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3736
3737         return nlmsg_end(skb, nlh);
3738
3739 nla_put_failure:
3740         nlmsg_cancel(skb, nlh);
3741         return -EMSGSIZE;
3742 }
3743
3744 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3745                          struct prefix_info *pinfo)
3746 {
3747         struct sk_buff *skb;
3748         struct net *net = dev_net(idev->dev);
3749         int err = -ENOBUFS;
3750
3751         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3752         if (skb == NULL)
3753                 goto errout;
3754
3755         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3756         if (err < 0) {
3757                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3758                 WARN_ON(err == -EMSGSIZE);
3759                 kfree_skb(skb);
3760                 goto errout;
3761         }
3762         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3763 errout:
3764         if (err < 0)
3765                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3766 }
3767
3768 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3769 {
3770         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3771
3772         switch (event) {
3773         case RTM_NEWADDR:
3774                 /*
3775                  * If the address was optimistic
3776                  * we inserted the route at the start of
3777                  * our DAD process, so we don't need
3778                  * to do it again
3779                  */
3780                 if (!(ifp->rt->rt6i_node))
3781                         ip6_ins_rt(ifp->rt);
3782                 if (ifp->idev->cnf.forwarding)
3783                         addrconf_join_anycast(ifp);
3784                 break;
3785         case RTM_DELADDR:
3786                 if (ifp->idev->cnf.forwarding)
3787                         addrconf_leave_anycast(ifp);
3788                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3789                 dst_hold(&ifp->rt->u.dst);
3790                 if (ip6_del_rt(ifp->rt))
3791                         dst_free(&ifp->rt->u.dst);
3792                 break;
3793         }
3794 }
3795
3796 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3797 {
3798         rcu_read_lock_bh();
3799         if (likely(ifp->idev->dead == 0))
3800                 __ipv6_ifa_notify(event, ifp);
3801         rcu_read_unlock_bh();
3802 }
3803
3804 #ifdef CONFIG_SYSCTL
3805
3806 static
3807 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3808                            void __user *buffer, size_t *lenp, loff_t *ppos)
3809 {
3810         int *valp = ctl->data;
3811         int val = *valp;
3812         int ret;
3813
3814         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3815
3816         if (write)
3817                 addrconf_fixup_forwarding(ctl, valp, val);
3818         return ret;
3819 }
3820
3821 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3822                                             int __user *name, int nlen,
3823                                             void __user *oldval,
3824                                             size_t __user *oldlenp,
3825                                             void __user *newval, size_t newlen)
3826 {
3827         int *valp = table->data;
3828         int val = *valp;
3829         int new;
3830
3831         if (!newval || !newlen)
3832                 return 0;
3833         if (newlen != sizeof(int))
3834                 return -EINVAL;
3835         if (get_user(new, (int __user *)newval))
3836                 return -EFAULT;
3837         if (new == *valp)
3838                 return 0;
3839         if (oldval && oldlenp) {
3840                 size_t len;
3841                 if (get_user(len, oldlenp))
3842                         return -EFAULT;
3843                 if (len) {
3844                         if (len > table->maxlen)
3845                                 len = table->maxlen;
3846                         if (copy_to_user(oldval, valp, len))
3847                                 return -EFAULT;
3848                         if (put_user(len, oldlenp))
3849                                 return -EFAULT;
3850                 }
3851         }
3852
3853         *valp = new;
3854         addrconf_fixup_forwarding(table, valp, val);
3855         return 1;
3856 }
3857
3858 static struct addrconf_sysctl_table
3859 {
3860         struct ctl_table_header *sysctl_header;
3861         ctl_table addrconf_vars[__NET_IPV6_MAX];
3862         char *dev_name;
3863 } addrconf_sysctl __read_mostly = {
3864         .sysctl_header = NULL,
3865         .addrconf_vars = {
3866                 {
3867                         .ctl_name       =       NET_IPV6_FORWARDING,
3868                         .procname       =       "forwarding",
3869                         .data           =       &ipv6_devconf.forwarding,
3870                         .maxlen         =       sizeof(int),
3871                         .mode           =       0644,
3872                         .proc_handler   =       &addrconf_sysctl_forward,
3873                         .strategy       =       &addrconf_sysctl_forward_strategy,
3874                 },
3875                 {
3876                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3877                         .procname       =       "hop_limit",
3878                         .data           =       &ipv6_devconf.hop_limit,
3879                         .maxlen         =       sizeof(int),
3880                         .mode           =       0644,
3881                         .proc_handler   =       proc_dointvec,
3882                 },
3883                 {
3884                         .ctl_name       =       NET_IPV6_MTU,
3885                         .procname       =       "mtu",
3886                         .data           =       &ipv6_devconf.mtu6,
3887                         .maxlen         =       sizeof(int),
3888                         .mode           =       0644,
3889                         .proc_handler   =       &proc_dointvec,
3890                 },
3891                 {
3892                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3893                         .procname       =       "accept_ra",
3894                         .data           =       &ipv6_devconf.accept_ra,
3895                         .maxlen         =       sizeof(int),
3896                         .mode           =       0644,
3897                         .proc_handler   =       &proc_dointvec,
3898                 },
3899                 {
3900                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3901                         .procname       =       "accept_redirects",
3902                         .data           =       &ipv6_devconf.accept_redirects,
3903                         .maxlen         =       sizeof(int),
3904                         .mode           =       0644,
3905                         .proc_handler   =       &proc_dointvec,
3906                 },
3907                 {
3908                         .ctl_name       =       NET_IPV6_AUTOCONF,
3909                         .procname       =       "autoconf",
3910                         .data           =       &ipv6_devconf.autoconf,
3911                         .maxlen         =       sizeof(int),
3912                         .mode           =       0644,
3913                         .proc_handler   =       &proc_dointvec,
3914                 },
3915                 {
3916                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3917                         .procname       =       "dad_transmits",
3918                         .data           =       &ipv6_devconf.dad_transmits,
3919                         .maxlen         =       sizeof(int),
3920                         .mode           =       0644,
3921                         .proc_handler   =       &proc_dointvec,
3922                 },
3923                 {
3924                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3925                         .procname       =       "router_solicitations",
3926                         .data           =       &ipv6_devconf.rtr_solicits,
3927                         .maxlen         =       sizeof(int),
3928                         .mode           =       0644,
3929                         .proc_handler   =       &proc_dointvec,
3930                 },
3931                 {
3932                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3933                         .procname       =       "router_solicitation_interval",
3934                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3935                         .maxlen         =       sizeof(int),
3936                         .mode           =       0644,
3937                         .proc_handler   =       &proc_dointvec_jiffies,
3938                         .strategy       =       &sysctl_jiffies,
3939                 },
3940                 {
3941                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3942                         .procname       =       "router_solicitation_delay",
3943                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3944                         .maxlen         =       sizeof(int),
3945                         .mode           =       0644,
3946                         .proc_handler   =       &proc_dointvec_jiffies,
3947                         .strategy       =       &sysctl_jiffies,
3948                 },
3949                 {
3950                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3951                         .procname       =       "force_mld_version",
3952                         .data           =       &ipv6_devconf.force_mld_version,
3953                         .maxlen         =       sizeof(int),
3954                         .mode           =       0644,
3955                         .proc_handler   =       &proc_dointvec,
3956                 },
3957 #ifdef CONFIG_IPV6_PRIVACY
3958                 {
3959                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3960                         .procname       =       "use_tempaddr",
3961                         .data           =       &ipv6_devconf.use_tempaddr,
3962                         .maxlen         =       sizeof(int),
3963                         .mode           =       0644,
3964                         .proc_handler   =       &proc_dointvec,
3965                 },
3966                 {
3967                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3968                         .procname       =       "temp_valid_lft",
3969                         .data           =       &ipv6_devconf.temp_valid_lft,
3970                         .maxlen         =       sizeof(int),
3971                         .mode           =       0644,
3972                         .proc_handler   =       &proc_dointvec,
3973                 },
3974                 {
3975                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3976                         .procname       =       "temp_prefered_lft",
3977                         .data           =       &ipv6_devconf.temp_prefered_lft,
3978                         .maxlen         =       sizeof(int),
3979                         .mode           =       0644,
3980                         .proc_handler   =       &proc_dointvec,
3981                 },
3982                 {
3983                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3984                         .procname       =       "regen_max_retry",
3985                         .data           =       &ipv6_devconf.regen_max_retry,
3986                         .maxlen         =       sizeof(int),
3987                         .mode           =       0644,
3988                         .proc_handler   =       &proc_dointvec,
3989                 },
3990                 {
3991                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3992                         .procname       =       "max_desync_factor",
3993                         .data           =       &ipv6_devconf.max_desync_factor,
3994                         .maxlen         =       sizeof(int),
3995                         .mode           =       0644,
3996                         .proc_handler   =       &proc_dointvec,
3997                 },
3998 #endif
3999                 {
4000                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4001                         .procname       =       "max_addresses",
4002                         .data           =       &ipv6_devconf.max_addresses,
4003                         .maxlen         =       sizeof(int),
4004                         .mode           =       0644,
4005                         .proc_handler   =       &proc_dointvec,
4006                 },
4007                 {
4008                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4009                         .procname       =       "accept_ra_defrtr",
4010                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4011                         .maxlen         =       sizeof(int),
4012                         .mode           =       0644,
4013                         .proc_handler   =       &proc_dointvec,
4014                 },
4015                 {
4016                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4017                         .procname       =       "accept_ra_pinfo",
4018                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4019                         .maxlen         =       sizeof(int),
4020                         .mode           =       0644,
4021                         .proc_handler   =       &proc_dointvec,
4022                 },
4023 #ifdef CONFIG_IPV6_ROUTER_PREF
4024                 {
4025                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4026                         .procname       =       "accept_ra_rtr_pref",
4027                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4028                         .maxlen         =       sizeof(int),
4029                         .mode           =       0644,
4030                         .proc_handler   =       &proc_dointvec,
4031                 },
4032                 {
4033                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4034                         .procname       =       "router_probe_interval",
4035                         .data           =       &ipv6_devconf.rtr_probe_interval,
4036                         .maxlen         =       sizeof(int),
4037                         .mode           =       0644,
4038                         .proc_handler   =       &proc_dointvec_jiffies,
4039                         .strategy       =       &sysctl_jiffies,
4040                 },
4041 #ifdef CONFIG_IPV6_ROUTE_INFO
4042                 {
4043                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4044                         .procname       =       "accept_ra_rt_info_max_plen",
4045                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4046                         .maxlen         =       sizeof(int),
4047                         .mode           =       0644,
4048                         .proc_handler   =       &proc_dointvec,
4049                 },
4050 #endif
4051 #endif
4052                 {
4053                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4054                         .procname       =       "proxy_ndp",
4055                         .data           =       &ipv6_devconf.proxy_ndp,
4056                         .maxlen         =       sizeof(int),
4057                         .mode           =       0644,
4058                         .proc_handler   =       &proc_dointvec,
4059                 },
4060                 {
4061                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4062                         .procname       =       "accept_source_route",
4063                         .data           =       &ipv6_devconf.accept_source_route,
4064                         .maxlen         =       sizeof(int),
4065                         .mode           =       0644,
4066                         .proc_handler   =       &proc_dointvec,
4067                 },
4068 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4069                 {
4070                         .ctl_name       =       CTL_UNNUMBERED,
4071                         .procname       =       "optimistic_dad",
4072                         .data           =       &ipv6_devconf.optimistic_dad,
4073                         .maxlen         =       sizeof(int),
4074                         .mode           =       0644,
4075                         .proc_handler   =       &proc_dointvec,
4076
4077                 },
4078 #endif
4079                 {
4080                         .ctl_name       =       0,      /* sentinel */
4081                 }
4082         },
4083 };
4084
4085 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4086                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4087 {
4088         int i;
4089         struct addrconf_sysctl_table *t;
4090
4091 #define ADDRCONF_CTL_PATH_DEV   3
4092
4093         struct ctl_path addrconf_ctl_path[] = {
4094                 { .procname = "net", .ctl_name = CTL_NET, },
4095                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4096                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4097                 { /* to be set */ },
4098                 { },
4099         };
4100
4101
4102         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4103         if (t == NULL)
4104                 goto out;
4105
4106         for (i=0; t->addrconf_vars[i].data; i++) {
4107                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4108                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4109                 t->addrconf_vars[i].extra2 = net;
4110         }
4111
4112         /*
4113          * Make a copy of dev_name, because '.procname' is regarded as const
4114          * by sysctl and we wouldn't want anyone to change it under our feet
4115          * (see SIOCSIFNAME).
4116          */
4117         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4118         if (!t->dev_name)
4119                 goto free;
4120
4121         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4122         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4123
4124         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4125                         t->addrconf_vars);
4126         if (t->sysctl_header == NULL)
4127                 goto free_procname;
4128
4129         p->sysctl = t;
4130         return 0;
4131
4132 free_procname:
4133         kfree(t->dev_name);
4134 free:
4135         kfree(t);
4136 out:
4137         return -ENOBUFS;
4138 }
4139
4140 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4141 {
4142         struct addrconf_sysctl_table *t;
4143
4144         if (p->sysctl == NULL)
4145                 return;
4146
4147         t = p->sysctl;
4148         p->sysctl = NULL;
4149         unregister_sysctl_table(t->sysctl_header);
4150         kfree(t->dev_name);
4151         kfree(t);
4152 }
4153
4154 static void addrconf_sysctl_register(struct inet6_dev *idev)
4155 {
4156         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4157                               NET_IPV6_NEIGH, "ipv6",
4158                               &ndisc_ifinfo_sysctl_change,
4159                               NULL);
4160         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4161                         idev->dev->ifindex, idev, &idev->cnf);
4162 }
4163
4164 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4165 {
4166         __addrconf_sysctl_unregister(&idev->cnf);
4167         neigh_sysctl_unregister(idev->nd_parms);
4168 }
4169
4170
4171 #endif
4172
4173 static int addrconf_init_net(struct net *net)
4174 {
4175         int err;
4176         struct ipv6_devconf *all, *dflt;
4177
4178         err = -ENOMEM;
4179         all = &ipv6_devconf;
4180         dflt = &ipv6_devconf_dflt;
4181
4182         if (net != &init_net) {
4183                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4184                 if (all == NULL)
4185                         goto err_alloc_all;
4186
4187                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4188                 if (dflt == NULL)
4189                         goto err_alloc_dflt;
4190         }
4191
4192         net->ipv6.devconf_all = all;
4193         net->ipv6.devconf_dflt = dflt;
4194
4195 #ifdef CONFIG_SYSCTL
4196         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4197                         NULL, all);
4198         if (err < 0)
4199                 goto err_reg_all;
4200
4201         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4202                         NULL, dflt);
4203         if (err < 0)
4204                 goto err_reg_dflt;
4205 #endif
4206         return 0;
4207
4208 #ifdef CONFIG_SYSCTL
4209 err_reg_dflt:
4210         __addrconf_sysctl_unregister(all);
4211 err_reg_all:
4212         kfree(dflt);
4213 #endif
4214 err_alloc_dflt:
4215         kfree(all);
4216 err_alloc_all:
4217         return err;
4218 }
4219
4220 static void addrconf_exit_net(struct net *net)
4221 {
4222 #ifdef CONFIG_SYSCTL
4223         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4224         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4225 #endif
4226         if (net != &init_net) {
4227                 kfree(net->ipv6.devconf_dflt);
4228                 kfree(net->ipv6.devconf_all);
4229         }
4230 }
4231
4232 static struct pernet_operations addrconf_ops = {
4233         .init = addrconf_init_net,
4234         .exit = addrconf_exit_net,
4235 };
4236
4237 /*
4238  *      Device notifier
4239  */
4240
4241 int register_inet6addr_notifier(struct notifier_block *nb)
4242 {
4243         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4244 }
4245
4246 EXPORT_SYMBOL(register_inet6addr_notifier);
4247
4248 int unregister_inet6addr_notifier(struct notifier_block *nb)
4249 {
4250         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4251 }
4252
4253 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4254
4255
4256 static int addrconf_net_init(struct net *net)
4257 {
4258         return 0;
4259 }
4260
4261 static void addrconf_net_exit(struct net *net)
4262 {
4263         struct net_device *dev;
4264
4265         rtnl_lock();
4266         /* clean dev list */
4267         for_each_netdev(net, dev) {
4268                 if (__in6_dev_get(dev) == NULL)
4269                         continue;
4270                 addrconf_ifdown(dev, 1);
4271         }
4272         addrconf_ifdown(net->loopback_dev, 2);
4273         rtnl_unlock();
4274 }
4275
4276 static struct pernet_operations addrconf_net_ops = {
4277         .init = addrconf_net_init,
4278         .exit = addrconf_net_exit,
4279 };
4280
4281 /*
4282  *      Init / cleanup code
4283  */
4284
4285 int __init addrconf_init(void)
4286 {
4287         int err;
4288
4289         if ((err = ipv6_addr_label_init()) < 0) {
4290                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4291                         err);
4292                 return err;
4293         }
4294
4295         register_pernet_subsys(&addrconf_ops);
4296
4297         /* The addrconf netdev notifier requires that loopback_dev
4298          * has it's ipv6 private information allocated and setup
4299          * before it can bring up and give link-local addresses
4300          * to other devices which are up.
4301          *
4302          * Unfortunately, loopback_dev is not necessarily the first
4303          * entry in the global dev_base list of net devices.  In fact,
4304          * it is likely to be the very last entry on that list.
4305          * So this causes the notifier registry below to try and
4306          * give link-local addresses to all devices besides loopback_dev
4307          * first, then loopback_dev, which cases all the non-loopback_dev
4308          * devices to fail to get a link-local address.
4309          *
4310          * So, as a temporary fix, allocate the ipv6 structure for
4311          * loopback_dev first by hand.
4312          * Longer term, all of the dependencies ipv6 has upon the loopback
4313          * device and it being up should be removed.
4314          */
4315         rtnl_lock();
4316         if (!ipv6_add_dev(init_net.loopback_dev))
4317                 err = -ENOMEM;
4318         rtnl_unlock();
4319         if (err)
4320                 goto errlo;
4321
4322         err = register_pernet_device(&addrconf_net_ops);
4323         if (err)
4324                 return err;
4325
4326         register_netdevice_notifier(&ipv6_dev_notf);
4327
4328         addrconf_verify(0);
4329
4330         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4331         if (err < 0)
4332                 goto errout;
4333
4334         /* Only the first call to __rtnl_register can fail */
4335         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4336         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4337         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4338         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4339         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4340
4341         ipv6_addr_label_rtnl_register();
4342
4343         return 0;
4344 errout:
4345         unregister_netdevice_notifier(&ipv6_dev_notf);
4346 errlo:
4347         unregister_pernet_subsys(&addrconf_ops);
4348
4349         return err;
4350 }
4351
4352 void addrconf_cleanup(void)
4353 {
4354         struct inet6_ifaddr *ifa;
4355         int i;
4356
4357         unregister_netdevice_notifier(&ipv6_dev_notf);
4358         unregister_pernet_device(&addrconf_net_ops);
4359
4360         unregister_pernet_subsys(&addrconf_ops);
4361
4362         rtnl_lock();
4363
4364         /*
4365          *      Check hash table.
4366          */
4367         write_lock_bh(&addrconf_hash_lock);
4368         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4369                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4370                         struct inet6_ifaddr *bifa;
4371
4372                         bifa = ifa;
4373                         ifa = ifa->lst_next;
4374                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4375                         /* Do not free it; something is wrong.
4376                            Now we can investigate it with debugger.
4377                          */
4378                 }
4379         }
4380         write_unlock_bh(&addrconf_hash_lock);
4381
4382         del_timer(&addr_chk_timer);
4383         rtnl_unlock();
4384
4385         unregister_pernet_subsys(&addrconf_net_ops);
4386 }