[IPV6] ADDRCONF: Clean-up ipv6_dev_get_saddr().
[safe/jmp/linux-2.6] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16
17 /*
18  *      Changes:
19  *
20  *      Janos Farkas                    :       delete timer on ifdown
21  *      <chexum@bankinf.banki.hu>
22  *      Andi Kleen                      :       kill double kfree on module
23  *                                              unload.
24  *      Maciej W. Rozycki               :       FDDI support
25  *      sekiya@USAGI                    :       Don't send too many RS
26  *                                              packets.
27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
28  *                                              packets.
29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
30  *                                              address validation timer.
31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
32  *                                              support.
33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
34  *                                              address on a same interface.
35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
37  *                                              seq_file.
38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
39  *                                              selection; consider scope,
40  *                                              status etc.
41  */
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99
100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
108 {
109 }
110
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 {
113 }
114 #endif
115
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
120
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
123
124 static int ipv6_count_addresses(struct inet6_dev *idev);
125
126 /*
127  *      Configured unicast address hash table
128  */
129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
131
132 static void addrconf_verify(unsigned long);
133
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
136
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139
140 static int addrconf_ifdown(struct net_device *dev, int how);
141
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
149
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151                                 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153                               struct net_device *dev);
154
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156
157 struct ipv6_devconf ipv6_devconf __read_mostly = {
158         .forwarding             = 0,
159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
160         .mtu6                   = IPV6_MIN_MTU,
161         .accept_ra              = 1,
162         .accept_redirects       = 1,
163         .autoconf               = 1,
164         .force_mld_version      = 0,
165         .dad_transmits          = 1,
166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170         .use_tempaddr           = 0,
171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
173         .regen_max_retry        = REGEN_MAX_RETRY,
174         .max_desync_factor      = MAX_DESYNC_FACTOR,
175 #endif
176         .max_addresses          = IPV6_MAX_ADDRESSES,
177         .accept_ra_defrtr       = 1,
178         .accept_ra_pinfo        = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180         .accept_ra_rtr_pref     = 1,
181         .rtr_probe_interval     = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183         .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186         .proxy_ndp              = 0,
187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
188 };
189
190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
191         .forwarding             = 0,
192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
193         .mtu6                   = IPV6_MIN_MTU,
194         .accept_ra              = 1,
195         .accept_redirects       = 1,
196         .autoconf               = 1,
197         .dad_transmits          = 1,
198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
201 #ifdef CONFIG_IPV6_PRIVACY
202         .use_tempaddr           = 0,
203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
205         .regen_max_retry        = REGEN_MAX_RETRY,
206         .max_desync_factor      = MAX_DESYNC_FACTOR,
207 #endif
208         .max_addresses          = IPV6_MAX_ADDRESSES,
209         .accept_ra_defrtr       = 1,
210         .accept_ra_pinfo        = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212         .accept_ra_rtr_pref     = 1,
213         .rtr_probe_interval     = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215         .accept_ra_rt_info_max_plen = 0,
216 #endif
217 #endif
218         .proxy_ndp              = 0,
219         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
220 };
221
222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
225
226 /* Check if a valid qdisc is available */
227 static inline int addrconf_qdisc_ok(struct net_device *dev)
228 {
229         return (dev->qdisc != &noop_qdisc);
230 }
231
232 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
233 {
234         if (del_timer(&ifp->timer))
235                 __in6_ifa_put(ifp);
236 }
237
238 enum addrconf_timer_t
239 {
240         AC_NONE,
241         AC_DAD,
242         AC_RS,
243 };
244
245 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
246                                enum addrconf_timer_t what,
247                                unsigned long when)
248 {
249         if (!del_timer(&ifp->timer))
250                 in6_ifa_hold(ifp);
251
252         switch (what) {
253         case AC_DAD:
254                 ifp->timer.function = addrconf_dad_timer;
255                 break;
256         case AC_RS:
257                 ifp->timer.function = addrconf_rs_timer;
258                 break;
259         default:;
260         }
261         ifp->timer.expires = jiffies + when;
262         add_timer(&ifp->timer);
263 }
264
265 static int snmp6_alloc_dev(struct inet6_dev *idev)
266 {
267         if (snmp_mib_init((void **)idev->stats.ipv6,
268                           sizeof(struct ipstats_mib)) < 0)
269                 goto err_ip;
270         if (snmp_mib_init((void **)idev->stats.icmpv6,
271                           sizeof(struct icmpv6_mib)) < 0)
272                 goto err_icmp;
273         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
274                           sizeof(struct icmpv6msg_mib)) < 0)
275                 goto err_icmpmsg;
276
277         return 0;
278
279 err_icmpmsg:
280         snmp_mib_free((void **)idev->stats.icmpv6);
281 err_icmp:
282         snmp_mib_free((void **)idev->stats.ipv6);
283 err_ip:
284         return -ENOMEM;
285 }
286
287 static void snmp6_free_dev(struct inet6_dev *idev)
288 {
289         snmp_mib_free((void **)idev->stats.icmpv6msg);
290         snmp_mib_free((void **)idev->stats.icmpv6);
291         snmp_mib_free((void **)idev->stats.ipv6);
292 }
293
294 /* Nobody refers to this device, we may destroy it. */
295
296 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
297 {
298         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
299         kfree(idev);
300 }
301
302 void in6_dev_finish_destroy(struct inet6_dev *idev)
303 {
304         struct net_device *dev = idev->dev;
305         BUG_TRAP(idev->addr_list==NULL);
306         BUG_TRAP(idev->mc_list==NULL);
307 #ifdef NET_REFCNT_DEBUG
308         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
309 #endif
310         dev_put(dev);
311         if (!idev->dead) {
312                 printk("Freeing alive inet6 device %p\n", idev);
313                 return;
314         }
315         snmp6_free_dev(idev);
316         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
317 }
318
319 EXPORT_SYMBOL(in6_dev_finish_destroy);
320
321 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
322 {
323         struct inet6_dev *ndev;
324         struct in6_addr maddr;
325
326         ASSERT_RTNL();
327
328         if (dev->mtu < IPV6_MIN_MTU)
329                 return NULL;
330
331         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
332
333         if (ndev == NULL)
334                 return NULL;
335
336         rwlock_init(&ndev->lock);
337         ndev->dev = dev;
338         memcpy(&ndev->cnf, dev->nd_net->ipv6.devconf_dflt, sizeof(ndev->cnf));
339         ndev->cnf.mtu6 = dev->mtu;
340         ndev->cnf.sysctl = NULL;
341         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
342         if (ndev->nd_parms == NULL) {
343                 kfree(ndev);
344                 return NULL;
345         }
346         /* We refer to the device */
347         dev_hold(dev);
348
349         if (snmp6_alloc_dev(ndev) < 0) {
350                 ADBG((KERN_WARNING
351                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
352                         __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(idev->dev->nd_net, addr, idev->dev)) {
565                 ADBG(("ipv6_add_addr: already assigned\n"));
566                 err = -EEXIST;
567                 goto out;
568         }
569
570         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
571
572         if (ifa == NULL) {
573                 ADBG(("ipv6_add_addr: malloc failed\n"));
574                 err = -ENOBUFS;
575                 goto out;
576         }
577
578         rt = addrconf_dst_alloc(idev, addr, 0);
579         if (IS_ERR(rt)) {
580                 err = PTR_ERR(rt);
581                 goto out;
582         }
583
584         ipv6_addr_copy(&ifa->addr, addr);
585
586         spin_lock_init(&ifa->lock);
587         init_timer(&ifa->timer);
588         ifa->timer.data = (unsigned long) ifa;
589         ifa->scope = scope;
590         ifa->prefix_len = pfxlen;
591         ifa->flags = flags | IFA_F_TENTATIVE;
592         ifa->cstamp = ifa->tstamp = jiffies;
593
594         ifa->rt = rt;
595
596         /*
597          * part one of RFC 4429, section 3.3
598          * We should not configure an address as
599          * optimistic if we do not yet know the link
600          * layer address of our nexhop router
601          */
602
603         if (rt->rt6i_nexthop == NULL)
604                 ifa->flags &= ~IFA_F_OPTIMISTIC;
605
606         ifa->idev = idev;
607         in6_dev_hold(idev);
608         /* For caller */
609         in6_ifa_hold(ifa);
610
611         /* Add to big hash table */
612         hash = ipv6_addr_hash(addr);
613
614         ifa->lst_next = inet6_addr_lst[hash];
615         inet6_addr_lst[hash] = ifa;
616         in6_ifa_hold(ifa);
617         write_unlock(&addrconf_hash_lock);
618
619         write_lock(&idev->lock);
620         /* Add to inet6_dev unicast addr list. */
621         ipv6_link_dev_addr(idev, ifa);
622
623 #ifdef CONFIG_IPV6_PRIVACY
624         if (ifa->flags&IFA_F_TEMPORARY) {
625                 ifa->tmp_next = idev->tempaddr_list;
626                 idev->tempaddr_list = ifa;
627                 in6_ifa_hold(ifa);
628         }
629 #endif
630
631         in6_ifa_hold(ifa);
632         write_unlock(&idev->lock);
633 out2:
634         rcu_read_unlock_bh();
635
636         if (likely(err == 0))
637                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
638         else {
639                 kfree(ifa);
640                 ifa = ERR_PTR(err);
641         }
642
643         return ifa;
644 out:
645         write_unlock(&addrconf_hash_lock);
646         goto out2;
647 }
648
649 /* This function wants to get referenced ifp and releases it before return */
650
651 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
652 {
653         struct inet6_ifaddr *ifa, **ifap;
654         struct inet6_dev *idev = ifp->idev;
655         int hash;
656         int deleted = 0, onlink = 0;
657         unsigned long expires = jiffies;
658
659         hash = ipv6_addr_hash(&ifp->addr);
660
661         ifp->dead = 1;
662
663         write_lock_bh(&addrconf_hash_lock);
664         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
665              ifap = &ifa->lst_next) {
666                 if (ifa == ifp) {
667                         *ifap = ifa->lst_next;
668                         __in6_ifa_put(ifp);
669                         ifa->lst_next = NULL;
670                         break;
671                 }
672         }
673         write_unlock_bh(&addrconf_hash_lock);
674
675         write_lock_bh(&idev->lock);
676 #ifdef CONFIG_IPV6_PRIVACY
677         if (ifp->flags&IFA_F_TEMPORARY) {
678                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
679                      ifap = &ifa->tmp_next) {
680                         if (ifa == ifp) {
681                                 *ifap = ifa->tmp_next;
682                                 if (ifp->ifpub) {
683                                         in6_ifa_put(ifp->ifpub);
684                                         ifp->ifpub = NULL;
685                                 }
686                                 __in6_ifa_put(ifp);
687                                 ifa->tmp_next = NULL;
688                                 break;
689                         }
690                 }
691         }
692 #endif
693
694         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
695                 if (ifa == ifp) {
696                         *ifap = ifa->if_next;
697                         __in6_ifa_put(ifp);
698                         ifa->if_next = NULL;
699                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
700                                 break;
701                         deleted = 1;
702                         continue;
703                 } else if (ifp->flags & IFA_F_PERMANENT) {
704                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
705                                               ifp->prefix_len)) {
706                                 if (ifa->flags & IFA_F_PERMANENT) {
707                                         onlink = 1;
708                                         if (deleted)
709                                                 break;
710                                 } else {
711                                         unsigned long lifetime;
712
713                                         if (!onlink)
714                                                 onlink = -1;
715
716                                         spin_lock(&ifa->lock);
717                                         lifetime = min_t(unsigned long,
718                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
719                                         if (time_before(expires,
720                                                         ifa->tstamp + lifetime * HZ))
721                                                 expires = ifa->tstamp + lifetime * HZ;
722                                         spin_unlock(&ifa->lock);
723                                 }
724                         }
725                 }
726                 ifap = &ifa->if_next;
727         }
728         write_unlock_bh(&idev->lock);
729
730         ipv6_ifa_notify(RTM_DELADDR, ifp);
731
732         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
733
734         addrconf_del_timer(ifp);
735
736         /*
737          * Purge or update corresponding prefix
738          *
739          * 1) we don't purge prefix here if address was not permanent.
740          *    prefix is managed by its own lifetime.
741          * 2) if there're no addresses, delete prefix.
742          * 3) if there're still other permanent address(es),
743          *    corresponding prefix is still permanent.
744          * 4) otherwise, update prefix lifetime to the
745          *    longest valid lifetime among the corresponding
746          *    addresses on the device.
747          *    Note: subsequent RA will update lifetime.
748          *
749          * --yoshfuji
750          */
751         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
752                 struct in6_addr prefix;
753                 struct rt6_info *rt;
754                 struct net *net = ifp->idev->dev->nd_net;
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 };
913
914 static inline int ipv6_saddr_preferred(int type)
915 {
916         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
917                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
918                 return 1;
919         return 0;
920 }
921
922 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
923                                struct ipv6_saddr_dst *dst,
924                                int i)
925 {
926         int ret;
927
928         if (i <= score->rule) {
929                 switch (i) {
930                 case IPV6_SADDR_RULE_SCOPE:
931                         ret = score->scopedist;
932                         break;
933                 case IPV6_SADDR_RULE_PREFIX:
934                         ret = score->matchlen;
935                         break;
936                 default:
937                         ret = !!test_bit(i, score->scorebits);
938                 }
939                 goto out;
940         }
941
942         switch (i) {
943         case IPV6_SADDR_RULE_INIT:
944                 /* Rule 0: remember if hiscore is not ready yet */
945                 ret = !!score->ifa;
946                 break;
947         case IPV6_SADDR_RULE_LOCAL:
948                 /* Rule 1: Prefer same address */
949                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
950                 break;
951         case IPV6_SADDR_RULE_SCOPE:
952                 /* Rule 2: Prefer appropriate scope
953                  *
954                  *      ret
955                  *       ^
956                  *    -1 |  d 15
957                  *    ---+--+-+---> scope
958                  *       |
959                  *       |             d is scope of the destination.
960                  *  B-d  |  \
961                  *       |   \      <- smaller scope is better if
962                  *  B-15 |    \        if scope is enough for destinaion.
963                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
964                  * d-C-1 | /
965                  *       |/         <- greater is better
966                  *   -C  /             if scope is not enough for destination.
967                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
968                  *
969                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
970                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
971                  * Assume B = 0 and we get C > 29.
972                  */
973                 ret = __ipv6_addr_src_scope(score->addr_type);
974                 if (ret >= dst->scope)
975                         ret = -ret;
976                 else
977                         ret -= 128;     /* 30 is enough */
978                 score->scopedist = ret;
979                 break;
980         case IPV6_SADDR_RULE_PREFERRED:
981                 /* Rule 3: Avoid deprecated and optimistic addresses */
982                 ret = ipv6_saddr_preferred(score->addr_type) ||
983                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
984                 break;
985 #ifdef CONFIG_IPV6_MIP6
986         case IPV6_SADDR_RULE_HOA:
987                 /* Rule 4: Prefer home address */
988                 ret = !!(score->ifa->flags & IFA_F_HOMEADDRESS);
989                 break;
990 #endif
991         case IPV6_SADDR_RULE_OIF:
992                 /* Rule 5: Prefer outgoing interface */
993                 ret = (!dst->ifindex ||
994                        dst->ifindex == score->ifa->idev->dev->ifindex);
995                 break;
996         case IPV6_SADDR_RULE_LABEL:
997                 /* Rule 6: Prefer matching label */
998                 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
999                                       score->ifa->idev->dev->ifindex) == dst->label;
1000                 break;
1001 #ifdef CONFIG_IPV6_PRIVACY
1002         case IPV6_SADDR_RULE_PRIVACY:
1003                 /* Rule 7: Prefer public address
1004                  * Note: prefer temprary address if use_tempaddr >= 2
1005                  */
1006                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ (score->ifa->idev->cnf.use_tempaddr >= 2);
1007                 break;
1008 #endif
1009         case IPV6_SADDR_RULE_ORCHID:
1010                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1011                  *          non-ORCHID vs non-ORCHID
1012                  */
1013                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1014                         ipv6_addr_orchid(dst->addr));
1015                 break;
1016         case IPV6_SADDR_RULE_PREFIX:
1017                 /* Rule 8: Use longest matching prefix */
1018                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1019                                                        dst->addr);
1020                 break;
1021         default:
1022                 ret = 0;
1023         }
1024
1025         if (ret)
1026                 __set_bit(i, score->scorebits);
1027         score->rule = i;
1028 out:
1029         return ret;
1030 }
1031
1032 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1033                        struct in6_addr *daddr, struct in6_addr *saddr)
1034 {
1035         struct ipv6_saddr_score scores[2],
1036                                 *score = &scores[0], *hiscore = &scores[1];
1037         struct net *net = dst_dev->nd_net;
1038         struct ipv6_saddr_dst dst;
1039         struct net_device *dev;
1040         int dst_type;
1041
1042         dst_type = __ipv6_addr_type(daddr);
1043         dst.addr = daddr;
1044         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1045         dst.scope = __ipv6_addr_src_scope(dst_type);
1046         dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1047
1048         hiscore->rule = -1;
1049         hiscore->ifa = NULL;
1050
1051         read_lock(&dev_base_lock);
1052         rcu_read_lock();
1053
1054         for_each_netdev(net, dev) {
1055                 struct inet6_dev *idev;
1056
1057                 /* Candidate Source Address (section 4)
1058                  *  - multicast and link-local destination address,
1059                  *    the set of candidate source address MUST only
1060                  *    include addresses assigned to interfaces
1061                  *    belonging to the same link as the outgoing
1062                  *    interface.
1063                  * (- For site-local destination addresses, the
1064                  *    set of candidate source addresses MUST only
1065                  *    include addresses assigned to interfaces
1066                  *    belonging to the same site as the outgoing
1067                  *    interface.)
1068                  */
1069                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1070                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1071                     dst.ifindex && dev->ifindex != dst.ifindex)
1072                         continue;
1073
1074                 idev = __in6_dev_get(dev);
1075                 if (!idev)
1076                         continue;
1077
1078                 read_lock_bh(&idev->lock);
1079                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1080                         int i;
1081
1082                         /*
1083                          * - Tentative Address (RFC2462 section 5.4)
1084                          *  - A tentative address is not considered
1085                          *    "assigned to an interface" in the traditional
1086                          *    sense, unless it is also flagged as optimistic.
1087                          * - Candidate Source Address (section 4)
1088                          *  - In any case, anycast addresses, multicast
1089                          *    addresses, and the unspecified address MUST
1090                          *    NOT be included in a candidate set.
1091                          */
1092                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1093                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1094                                 continue;
1095
1096                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1097
1098                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1099                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1100                                 LIMIT_NETDEBUG(KERN_DEBUG
1101                                                "ADDRCONF: unspecified / multicast address "
1102                                                "assigned as unicast address on %s",
1103                                                dev->name);
1104                                 continue;
1105                         }
1106
1107                         score->rule = -1;
1108                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1109
1110                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1111                                 int minihiscore, miniscore;
1112
1113                                 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1114                                 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1115
1116                                 if (minihiscore > miniscore) {
1117                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1118                                             score->scopedist > 0) {
1119                                                 /*
1120                                                  * special case:
1121                                                  * each remaining entry
1122                                                  * has too small (not enough)
1123                                                  * scope, because ifa entries
1124                                                  * are sorted by their scope
1125                                                  * values.
1126                                                  */
1127                                                 goto try_nextdev;
1128                                         }
1129                                         break;
1130                                 } else if (minihiscore < miniscore) {
1131                                         struct ipv6_saddr_score *tmp;
1132
1133                                         if (hiscore->ifa)
1134                                                 in6_ifa_put(hiscore->ifa);
1135
1136                                         in6_ifa_hold(score->ifa);
1137
1138                                         tmp = hiscore;
1139                                         hiscore = score;
1140                                         score = tmp;
1141
1142                                         /* restore our iterator */
1143                                         score->ifa = hiscore->ifa;
1144
1145                                         break;
1146                                 }
1147                         }
1148                 }
1149 try_nextdev:
1150                 read_unlock_bh(&idev->lock);
1151         }
1152         rcu_read_unlock();
1153         read_unlock(&dev_base_lock);
1154
1155         if (!hiscore->ifa)
1156                 return -EADDRNOTAVAIL;
1157
1158         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1159         in6_ifa_put(hiscore->ifa);
1160         return 0;
1161 }
1162
1163 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1164
1165 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1166                     unsigned char banned_flags)
1167 {
1168         struct inet6_dev *idev;
1169         int err = -EADDRNOTAVAIL;
1170
1171         rcu_read_lock();
1172         if ((idev = __in6_dev_get(dev)) != NULL) {
1173                 struct inet6_ifaddr *ifp;
1174
1175                 read_lock_bh(&idev->lock);
1176                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1177                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1178                                 ipv6_addr_copy(addr, &ifp->addr);
1179                                 err = 0;
1180                                 break;
1181                         }
1182                 }
1183                 read_unlock_bh(&idev->lock);
1184         }
1185         rcu_read_unlock();
1186         return err;
1187 }
1188
1189 static int ipv6_count_addresses(struct inet6_dev *idev)
1190 {
1191         int cnt = 0;
1192         struct inet6_ifaddr *ifp;
1193
1194         read_lock_bh(&idev->lock);
1195         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1196                 cnt++;
1197         read_unlock_bh(&idev->lock);
1198         return cnt;
1199 }
1200
1201 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1202                   struct net_device *dev, int strict)
1203 {
1204         struct inet6_ifaddr * ifp;
1205         u8 hash = ipv6_addr_hash(addr);
1206
1207         read_lock_bh(&addrconf_hash_lock);
1208         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1209                 if (ifp->idev->dev->nd_net != net)
1210                         continue;
1211                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1212                     !(ifp->flags&IFA_F_TENTATIVE)) {
1213                         if (dev == NULL || ifp->idev->dev == dev ||
1214                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1215                                 break;
1216                 }
1217         }
1218         read_unlock_bh(&addrconf_hash_lock);
1219         return ifp != NULL;
1220 }
1221 EXPORT_SYMBOL(ipv6_chk_addr);
1222
1223 static
1224 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1225                        struct net_device *dev)
1226 {
1227         struct inet6_ifaddr * ifp;
1228         u8 hash = ipv6_addr_hash(addr);
1229
1230         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1231                 if (ifp->idev->dev->nd_net != net)
1232                         continue;
1233                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1234                         if (dev == NULL || ifp->idev->dev == dev)
1235                                 break;
1236                 }
1237         }
1238         return ifp != NULL;
1239 }
1240
1241 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1242                                      struct net_device *dev, int strict)
1243 {
1244         struct inet6_ifaddr * ifp;
1245         u8 hash = ipv6_addr_hash(addr);
1246
1247         read_lock_bh(&addrconf_hash_lock);
1248         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1249                 if (ifp->idev->dev->nd_net != net)
1250                         continue;
1251                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1252                         if (dev == NULL || ifp->idev->dev == dev ||
1253                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1254                                 in6_ifa_hold(ifp);
1255                                 break;
1256                         }
1257                 }
1258         }
1259         read_unlock_bh(&addrconf_hash_lock);
1260
1261         return ifp;
1262 }
1263
1264 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1265 {
1266         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1267         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1268         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1269         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1270         int sk_ipv6only = ipv6_only_sock(sk);
1271         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1272         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1273         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1274
1275         if (!sk2_rcv_saddr && !sk_ipv6only)
1276                 return 1;
1277
1278         if (addr_type2 == IPV6_ADDR_ANY &&
1279             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1280                 return 1;
1281
1282         if (addr_type == IPV6_ADDR_ANY &&
1283             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1284                 return 1;
1285
1286         if (sk2_rcv_saddr6 &&
1287             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1288                 return 1;
1289
1290         if (addr_type == IPV6_ADDR_MAPPED &&
1291             !sk2_ipv6only &&
1292             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1293                 return 1;
1294
1295         return 0;
1296 }
1297
1298 /* Gets referenced address, destroys ifaddr */
1299
1300 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1301 {
1302         if (ifp->flags&IFA_F_PERMANENT) {
1303                 spin_lock_bh(&ifp->lock);
1304                 addrconf_del_timer(ifp);
1305                 ifp->flags |= IFA_F_TENTATIVE;
1306                 spin_unlock_bh(&ifp->lock);
1307                 in6_ifa_put(ifp);
1308 #ifdef CONFIG_IPV6_PRIVACY
1309         } else if (ifp->flags&IFA_F_TEMPORARY) {
1310                 struct inet6_ifaddr *ifpub;
1311                 spin_lock_bh(&ifp->lock);
1312                 ifpub = ifp->ifpub;
1313                 if (ifpub) {
1314                         in6_ifa_hold(ifpub);
1315                         spin_unlock_bh(&ifp->lock);
1316                         ipv6_create_tempaddr(ifpub, ifp);
1317                         in6_ifa_put(ifpub);
1318                 } else {
1319                         spin_unlock_bh(&ifp->lock);
1320                 }
1321                 ipv6_del_addr(ifp);
1322 #endif
1323         } else
1324                 ipv6_del_addr(ifp);
1325 }
1326
1327 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1328 {
1329         if (net_ratelimit())
1330                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1331         addrconf_dad_stop(ifp);
1332 }
1333
1334 /* Join to solicited addr multicast group. */
1335
1336 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1337 {
1338         struct in6_addr maddr;
1339
1340         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1341                 return;
1342
1343         addrconf_addr_solict_mult(addr, &maddr);
1344         ipv6_dev_mc_inc(dev, &maddr);
1345 }
1346
1347 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1348 {
1349         struct in6_addr maddr;
1350
1351         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1352                 return;
1353
1354         addrconf_addr_solict_mult(addr, &maddr);
1355         __ipv6_dev_mc_dec(idev, &maddr);
1356 }
1357
1358 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1359 {
1360         struct in6_addr addr;
1361         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1362         if (ipv6_addr_any(&addr))
1363                 return;
1364         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1365 }
1366
1367 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1368 {
1369         struct in6_addr addr;
1370         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1371         if (ipv6_addr_any(&addr))
1372                 return;
1373         __ipv6_dev_ac_dec(ifp->idev, &addr);
1374 }
1375
1376 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1377 {
1378         if (dev->addr_len != ETH_ALEN)
1379                 return -1;
1380         memcpy(eui, dev->dev_addr, 3);
1381         memcpy(eui + 5, dev->dev_addr + 3, 3);
1382
1383         /*
1384          * The zSeries OSA network cards can be shared among various
1385          * OS instances, but the OSA cards have only one MAC address.
1386          * This leads to duplicate address conflicts in conjunction
1387          * with IPv6 if more than one instance uses the same card.
1388          *
1389          * The driver for these cards can deliver a unique 16-bit
1390          * identifier for each instance sharing the same card.  It is
1391          * placed instead of 0xFFFE in the interface identifier.  The
1392          * "u" bit of the interface identifier is not inverted in this
1393          * case.  Hence the resulting interface identifier has local
1394          * scope according to RFC2373.
1395          */
1396         if (dev->dev_id) {
1397                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1398                 eui[4] = dev->dev_id & 0xFF;
1399         } else {
1400                 eui[3] = 0xFF;
1401                 eui[4] = 0xFE;
1402                 eui[0] ^= 2;
1403         }
1404         return 0;
1405 }
1406
1407 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1408 {
1409         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1410         if (dev->addr_len != ARCNET_ALEN)
1411                 return -1;
1412         memset(eui, 0, 7);
1413         eui[7] = *(u8*)dev->dev_addr;
1414         return 0;
1415 }
1416
1417 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1418 {
1419         if (dev->addr_len != INFINIBAND_ALEN)
1420                 return -1;
1421         memcpy(eui, dev->dev_addr + 12, 8);
1422         eui[0] |= 2;
1423         return 0;
1424 }
1425
1426 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1427 {
1428         switch (dev->type) {
1429         case ARPHRD_ETHER:
1430         case ARPHRD_FDDI:
1431         case ARPHRD_IEEE802_TR:
1432                 return addrconf_ifid_eui48(eui, dev);
1433         case ARPHRD_ARCNET:
1434                 return addrconf_ifid_arcnet(eui, dev);
1435         case ARPHRD_INFINIBAND:
1436                 return addrconf_ifid_infiniband(eui, dev);
1437         case ARPHRD_SIT:
1438                 if (dev->priv_flags & IFF_ISATAP)
1439                         return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1440         }
1441         return -1;
1442 }
1443
1444 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1445 {
1446         int err = -1;
1447         struct inet6_ifaddr *ifp;
1448
1449         read_lock_bh(&idev->lock);
1450         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1451                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1452                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1453                         err = 0;
1454                         break;
1455                 }
1456         }
1457         read_unlock_bh(&idev->lock);
1458         return err;
1459 }
1460
1461 #ifdef CONFIG_IPV6_PRIVACY
1462 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1463 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1464 {
1465 regen:
1466         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1467         idev->rndid[0] &= ~0x02;
1468
1469         /*
1470          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1471          * check if generated address is not inappropriate
1472          *
1473          *  - Reserved subnet anycast (RFC 2526)
1474          *      11111101 11....11 1xxxxxxx
1475          *  - ISATAP (RFC4214) 6.1
1476          *      00-00-5E-FE-xx-xx-xx-xx
1477          *  - value 0
1478          *  - XXX: already assigned to an address on the device
1479          */
1480         if (idev->rndid[0] == 0xfd &&
1481             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1482             (idev->rndid[7]&0x80))
1483                 goto regen;
1484         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1485                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1486                         goto regen;
1487                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1488                         goto regen;
1489         }
1490
1491         return 0;
1492 }
1493
1494 static void ipv6_regen_rndid(unsigned long data)
1495 {
1496         struct inet6_dev *idev = (struct inet6_dev *) data;
1497         unsigned long expires;
1498
1499         rcu_read_lock_bh();
1500         write_lock_bh(&idev->lock);
1501
1502         if (idev->dead)
1503                 goto out;
1504
1505         if (__ipv6_regen_rndid(idev) < 0)
1506                 goto out;
1507
1508         expires = jiffies +
1509                 idev->cnf.temp_prefered_lft * HZ -
1510                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1511         if (time_before(expires, jiffies)) {
1512                 printk(KERN_WARNING
1513                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1514                         idev->dev->name);
1515                 goto out;
1516         }
1517
1518         if (!mod_timer(&idev->regen_timer, expires))
1519                 in6_dev_hold(idev);
1520
1521 out:
1522         write_unlock_bh(&idev->lock);
1523         rcu_read_unlock_bh();
1524         in6_dev_put(idev);
1525 }
1526
1527 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1528         int ret = 0;
1529
1530         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1531                 ret = __ipv6_regen_rndid(idev);
1532         return ret;
1533 }
1534 #endif
1535
1536 /*
1537  *      Add prefix route.
1538  */
1539
1540 static void
1541 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1542                       unsigned long expires, u32 flags)
1543 {
1544         struct fib6_config cfg = {
1545                 .fc_table = RT6_TABLE_PREFIX,
1546                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1547                 .fc_ifindex = dev->ifindex,
1548                 .fc_expires = expires,
1549                 .fc_dst_len = plen,
1550                 .fc_flags = RTF_UP | flags,
1551                 .fc_nlinfo.nl_net = dev->nd_net,
1552         };
1553
1554         ipv6_addr_copy(&cfg.fc_dst, pfx);
1555
1556         /* Prevent useless cloning on PtP SIT.
1557            This thing is done here expecting that the whole
1558            class of non-broadcast devices need not cloning.
1559          */
1560 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1561         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1562                 cfg.fc_flags |= RTF_NONEXTHOP;
1563 #endif
1564
1565         ip6_route_add(&cfg);
1566 }
1567
1568 /* Create "default" multicast route to the interface */
1569
1570 static void addrconf_add_mroute(struct net_device *dev)
1571 {
1572         struct fib6_config cfg = {
1573                 .fc_table = RT6_TABLE_LOCAL,
1574                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1575                 .fc_ifindex = dev->ifindex,
1576                 .fc_dst_len = 8,
1577                 .fc_flags = RTF_UP,
1578                 .fc_nlinfo.nl_net = dev->nd_net,
1579         };
1580
1581         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1582
1583         ip6_route_add(&cfg);
1584 }
1585
1586 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1587 static void sit_route_add(struct net_device *dev)
1588 {
1589         struct fib6_config cfg = {
1590                 .fc_table = RT6_TABLE_MAIN,
1591                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1592                 .fc_ifindex = dev->ifindex,
1593                 .fc_dst_len = 96,
1594                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1595                 .fc_nlinfo.nl_net = dev->nd_net,
1596         };
1597
1598         /* prefix length - 96 bits "::d.d.d.d" */
1599         ip6_route_add(&cfg);
1600 }
1601 #endif
1602
1603 static void addrconf_add_lroute(struct net_device *dev)
1604 {
1605         struct in6_addr addr;
1606
1607         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1608         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1609 }
1610
1611 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1612 {
1613         struct inet6_dev *idev;
1614
1615         ASSERT_RTNL();
1616
1617         if ((idev = ipv6_find_idev(dev)) == NULL)
1618                 return NULL;
1619
1620         /* Add default multicast route */
1621         addrconf_add_mroute(dev);
1622
1623         /* Add link local route */
1624         addrconf_add_lroute(dev);
1625         return idev;
1626 }
1627
1628 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1629 {
1630         struct prefix_info *pinfo;
1631         __u32 valid_lft;
1632         __u32 prefered_lft;
1633         int addr_type;
1634         unsigned long rt_expires;
1635         struct inet6_dev *in6_dev;
1636
1637         pinfo = (struct prefix_info *) opt;
1638
1639         if (len < sizeof(struct prefix_info)) {
1640                 ADBG(("addrconf: prefix option too short\n"));
1641                 return;
1642         }
1643
1644         /*
1645          *      Validation checks ([ADDRCONF], page 19)
1646          */
1647
1648         addr_type = ipv6_addr_type(&pinfo->prefix);
1649
1650         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1651                 return;
1652
1653         valid_lft = ntohl(pinfo->valid);
1654         prefered_lft = ntohl(pinfo->prefered);
1655
1656         if (prefered_lft > valid_lft) {
1657                 if (net_ratelimit())
1658                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1659                 return;
1660         }
1661
1662         in6_dev = in6_dev_get(dev);
1663
1664         if (in6_dev == NULL) {
1665                 if (net_ratelimit())
1666                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1667                 return;
1668         }
1669
1670         /*
1671          *      Two things going on here:
1672          *      1) Add routes for on-link prefixes
1673          *      2) Configure prefixes with the auto flag set
1674          */
1675
1676         /* Avoid arithmetic overflow. Really, we could
1677            save rt_expires in seconds, likely valid_lft,
1678            but it would require division in fib gc, that it
1679            not good.
1680          */
1681         if (valid_lft >= 0x7FFFFFFF/HZ)
1682                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1683         else
1684                 rt_expires = valid_lft * HZ;
1685
1686         /*
1687          * We convert this (in jiffies) to clock_t later.
1688          * Avoid arithmetic overflow there as well.
1689          * Overflow can happen only if HZ < USER_HZ.
1690          */
1691         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1692                 rt_expires = 0x7FFFFFFF / USER_HZ;
1693
1694         if (pinfo->onlink) {
1695                 struct rt6_info *rt;
1696                 rt = rt6_lookup(dev->nd_net, &pinfo->prefix, NULL,
1697                                 dev->ifindex, 1);
1698
1699                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1700                         if (rt->rt6i_flags&RTF_EXPIRES) {
1701                                 if (valid_lft == 0) {
1702                                         ip6_del_rt(rt);
1703                                         rt = NULL;
1704                                 } else {
1705                                         rt->rt6i_expires = jiffies + rt_expires;
1706                                 }
1707                         }
1708                 } else if (valid_lft) {
1709                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1710                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1711                 }
1712                 if (rt)
1713                         dst_release(&rt->u.dst);
1714         }
1715
1716         /* Try to figure out our local address for this prefix */
1717
1718         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1719                 struct inet6_ifaddr * ifp;
1720                 struct in6_addr addr;
1721                 int create = 0, update_lft = 0;
1722
1723                 if (pinfo->prefix_len == 64) {
1724                         memcpy(&addr, &pinfo->prefix, 8);
1725                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1726                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1727                                 in6_dev_put(in6_dev);
1728                                 return;
1729                         }
1730                         goto ok;
1731                 }
1732                 if (net_ratelimit())
1733                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1734                                pinfo->prefix_len);
1735                 in6_dev_put(in6_dev);
1736                 return;
1737
1738 ok:
1739
1740                 ifp = ipv6_get_ifaddr(dev->nd_net, &addr, dev, 1);
1741
1742                 if (ifp == NULL && valid_lft) {
1743                         int max_addresses = in6_dev->cnf.max_addresses;
1744                         u32 addr_flags = 0;
1745
1746 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1747                         if (in6_dev->cnf.optimistic_dad &&
1748                             !ipv6_devconf.forwarding)
1749                                 addr_flags = IFA_F_OPTIMISTIC;
1750 #endif
1751
1752                         /* Do not allow to create too much of autoconfigured
1753                          * addresses; this would be too easy way to crash kernel.
1754                          */
1755                         if (!max_addresses ||
1756                             ipv6_count_addresses(in6_dev) < max_addresses)
1757                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1758                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1759                                                     addr_flags);
1760
1761                         if (!ifp || IS_ERR(ifp)) {
1762                                 in6_dev_put(in6_dev);
1763                                 return;
1764                         }
1765
1766                         update_lft = create = 1;
1767                         ifp->cstamp = jiffies;
1768                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1769                 }
1770
1771                 if (ifp) {
1772                         int flags;
1773                         unsigned long now;
1774 #ifdef CONFIG_IPV6_PRIVACY
1775                         struct inet6_ifaddr *ift;
1776 #endif
1777                         u32 stored_lft;
1778
1779                         /* update lifetime (RFC2462 5.5.3 e) */
1780                         spin_lock(&ifp->lock);
1781                         now = jiffies;
1782                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1783                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1784                         else
1785                                 stored_lft = 0;
1786                         if (!update_lft && stored_lft) {
1787                                 if (valid_lft > MIN_VALID_LIFETIME ||
1788                                     valid_lft > stored_lft)
1789                                         update_lft = 1;
1790                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1791                                         /* valid_lft <= stored_lft is always true */
1792                                         /* XXX: IPsec */
1793                                         update_lft = 0;
1794                                 } else {
1795                                         valid_lft = MIN_VALID_LIFETIME;
1796                                         if (valid_lft < prefered_lft)
1797                                                 prefered_lft = valid_lft;
1798                                         update_lft = 1;
1799                                 }
1800                         }
1801
1802                         if (update_lft) {
1803                                 ifp->valid_lft = valid_lft;
1804                                 ifp->prefered_lft = prefered_lft;
1805                                 ifp->tstamp = now;
1806                                 flags = ifp->flags;
1807                                 ifp->flags &= ~IFA_F_DEPRECATED;
1808                                 spin_unlock(&ifp->lock);
1809
1810                                 if (!(flags&IFA_F_TENTATIVE))
1811                                         ipv6_ifa_notify(0, ifp);
1812                         } else
1813                                 spin_unlock(&ifp->lock);
1814
1815 #ifdef CONFIG_IPV6_PRIVACY
1816                         read_lock_bh(&in6_dev->lock);
1817                         /* update all temporary addresses in the list */
1818                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1819                                 /*
1820                                  * When adjusting the lifetimes of an existing
1821                                  * temporary address, only lower the lifetimes.
1822                                  * Implementations must not increase the
1823                                  * lifetimes of an existing temporary address
1824                                  * when processing a Prefix Information Option.
1825                                  */
1826                                 spin_lock(&ift->lock);
1827                                 flags = ift->flags;
1828                                 if (ift->valid_lft > valid_lft &&
1829                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1830                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1831                                 if (ift->prefered_lft > prefered_lft &&
1832                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1833                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1834                                 spin_unlock(&ift->lock);
1835                                 if (!(flags&IFA_F_TENTATIVE))
1836                                         ipv6_ifa_notify(0, ift);
1837                         }
1838
1839                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1840                                 /*
1841                                  * When a new public address is created as described in [ADDRCONF],
1842                                  * also create a new temporary address.
1843                                  */
1844                                 read_unlock_bh(&in6_dev->lock);
1845                                 ipv6_create_tempaddr(ifp, NULL);
1846                         } else {
1847                                 read_unlock_bh(&in6_dev->lock);
1848                         }
1849 #endif
1850                         in6_ifa_put(ifp);
1851                         addrconf_verify(0);
1852                 }
1853         }
1854         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1855         in6_dev_put(in6_dev);
1856 }
1857
1858 /*
1859  *      Set destination address.
1860  *      Special case for SIT interfaces where we create a new "virtual"
1861  *      device.
1862  */
1863 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1864 {
1865         struct in6_ifreq ireq;
1866         struct net_device *dev;
1867         int err = -EINVAL;
1868
1869         rtnl_lock();
1870
1871         err = -EFAULT;
1872         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1873                 goto err_exit;
1874
1875         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1876
1877         err = -ENODEV;
1878         if (dev == NULL)
1879                 goto err_exit;
1880
1881 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1882         if (dev->type == ARPHRD_SIT) {
1883                 struct ifreq ifr;
1884                 mm_segment_t    oldfs;
1885                 struct ip_tunnel_parm p;
1886
1887                 err = -EADDRNOTAVAIL;
1888                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1889                         goto err_exit;
1890
1891                 memset(&p, 0, sizeof(p));
1892                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1893                 p.iph.saddr = 0;
1894                 p.iph.version = 4;
1895                 p.iph.ihl = 5;
1896                 p.iph.protocol = IPPROTO_IPV6;
1897                 p.iph.ttl = 64;
1898                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1899
1900                 oldfs = get_fs(); set_fs(KERNEL_DS);
1901                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1902                 set_fs(oldfs);
1903
1904                 if (err == 0) {
1905                         err = -ENOBUFS;
1906                         dev = __dev_get_by_name(net, p.name);
1907                         if (!dev)
1908                                 goto err_exit;
1909                         err = dev_open(dev);
1910                 }
1911         }
1912 #endif
1913
1914 err_exit:
1915         rtnl_unlock();
1916         return err;
1917 }
1918
1919 /*
1920  *      Manual configuration of address on an interface
1921  */
1922 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
1923                           int plen, __u8 ifa_flags, __u32 prefered_lft,
1924                           __u32 valid_lft)
1925 {
1926         struct inet6_ifaddr *ifp;
1927         struct inet6_dev *idev;
1928         struct net_device *dev;
1929         int scope;
1930         u32 flags = RTF_EXPIRES;
1931
1932         ASSERT_RTNL();
1933
1934         /* check the lifetime */
1935         if (!valid_lft || prefered_lft > valid_lft)
1936                 return -EINVAL;
1937
1938         dev = __dev_get_by_index(net, ifindex);
1939         if (!dev)
1940                 return -ENODEV;
1941
1942         if ((idev = addrconf_add_dev(dev)) == NULL)
1943                 return -ENOBUFS;
1944
1945         scope = ipv6_addr_scope(pfx);
1946
1947         if (valid_lft == INFINITY_LIFE_TIME) {
1948                 ifa_flags |= IFA_F_PERMANENT;
1949                 flags = 0;
1950         } else if (valid_lft >= 0x7FFFFFFF/HZ)
1951                 valid_lft = 0x7FFFFFFF/HZ;
1952
1953         if (prefered_lft == 0)
1954                 ifa_flags |= IFA_F_DEPRECATED;
1955         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1956                  (prefered_lft != INFINITY_LIFE_TIME))
1957                 prefered_lft = 0x7FFFFFFF/HZ;
1958
1959         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1960
1961         if (!IS_ERR(ifp)) {
1962                 spin_lock_bh(&ifp->lock);
1963                 ifp->valid_lft = valid_lft;
1964                 ifp->prefered_lft = prefered_lft;
1965                 ifp->tstamp = jiffies;
1966                 spin_unlock_bh(&ifp->lock);
1967
1968                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1969                                       jiffies_to_clock_t(valid_lft * HZ), flags);
1970                 /*
1971                  * Note that section 3.1 of RFC 4429 indicates
1972                  * that the Optimistic flag should not be set for
1973                  * manually configured addresses
1974                  */
1975                 addrconf_dad_start(ifp, 0);
1976                 in6_ifa_put(ifp);
1977                 addrconf_verify(0);
1978                 return 0;
1979         }
1980
1981         return PTR_ERR(ifp);
1982 }
1983
1984 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
1985                           int plen)
1986 {
1987         struct inet6_ifaddr *ifp;
1988         struct inet6_dev *idev;
1989         struct net_device *dev;
1990
1991         dev = __dev_get_by_index(net, ifindex);
1992         if (!dev)
1993                 return -ENODEV;
1994
1995         if ((idev = __in6_dev_get(dev)) == NULL)
1996                 return -ENXIO;
1997
1998         read_lock_bh(&idev->lock);
1999         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2000                 if (ifp->prefix_len == plen &&
2001                     ipv6_addr_equal(pfx, &ifp->addr)) {
2002                         in6_ifa_hold(ifp);
2003                         read_unlock_bh(&idev->lock);
2004
2005                         ipv6_del_addr(ifp);
2006
2007                         /* If the last address is deleted administratively,
2008                            disable IPv6 on this interface.
2009                          */
2010                         if (idev->addr_list == NULL)
2011                                 addrconf_ifdown(idev->dev, 1);
2012                         return 0;
2013                 }
2014         }
2015         read_unlock_bh(&idev->lock);
2016         return -EADDRNOTAVAIL;
2017 }
2018
2019
2020 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2021 {
2022         struct in6_ifreq ireq;
2023         int err;
2024
2025         if (!capable(CAP_NET_ADMIN))
2026                 return -EPERM;
2027
2028         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2029                 return -EFAULT;
2030
2031         rtnl_lock();
2032         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2033                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2034                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2035         rtnl_unlock();
2036         return err;
2037 }
2038
2039 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2040 {
2041         struct in6_ifreq ireq;
2042         int err;
2043
2044         if (!capable(CAP_NET_ADMIN))
2045                 return -EPERM;
2046
2047         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2048                 return -EFAULT;
2049
2050         rtnl_lock();
2051         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2052                              ireq.ifr6_prefixlen);
2053         rtnl_unlock();
2054         return err;
2055 }
2056
2057 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2058 static void sit_add_v4_addrs(struct inet6_dev *idev)
2059 {
2060         struct inet6_ifaddr * ifp;
2061         struct in6_addr addr;
2062         struct net_device *dev;
2063         struct net *net = idev->dev->nd_net;
2064         int scope;
2065
2066         ASSERT_RTNL();
2067
2068         memset(&addr, 0, sizeof(struct in6_addr));
2069         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2070
2071         if (idev->dev->flags&IFF_POINTOPOINT) {
2072                 addr.s6_addr32[0] = htonl(0xfe800000);
2073                 scope = IFA_LINK;
2074         } else {
2075                 scope = IPV6_ADDR_COMPATv4;
2076         }
2077
2078         if (addr.s6_addr32[3]) {
2079                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2080                 if (!IS_ERR(ifp)) {
2081                         spin_lock_bh(&ifp->lock);
2082                         ifp->flags &= ~IFA_F_TENTATIVE;
2083                         spin_unlock_bh(&ifp->lock);
2084                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2085                         in6_ifa_put(ifp);
2086                 }
2087                 return;
2088         }
2089
2090         for_each_netdev(net, dev) {
2091                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2092                 if (in_dev && (dev->flags & IFF_UP)) {
2093                         struct in_ifaddr * ifa;
2094
2095                         int flag = scope;
2096
2097                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2098                                 int plen;
2099
2100                                 addr.s6_addr32[3] = ifa->ifa_local;
2101
2102                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2103                                         continue;
2104                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2105                                         if (idev->dev->flags&IFF_POINTOPOINT)
2106                                                 continue;
2107                                         flag |= IFA_HOST;
2108                                 }
2109                                 if (idev->dev->flags&IFF_POINTOPOINT)
2110                                         plen = 64;
2111                                 else
2112                                         plen = 96;
2113
2114                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2115                                                     IFA_F_PERMANENT);
2116                                 if (!IS_ERR(ifp)) {
2117                                         spin_lock_bh(&ifp->lock);
2118                                         ifp->flags &= ~IFA_F_TENTATIVE;
2119                                         spin_unlock_bh(&ifp->lock);
2120                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2121                                         in6_ifa_put(ifp);
2122                                 }
2123                         }
2124                 }
2125         }
2126 }
2127 #endif
2128
2129 static void init_loopback(struct net_device *dev)
2130 {
2131         struct inet6_dev  *idev;
2132         struct inet6_ifaddr * ifp;
2133
2134         /* ::1 */
2135
2136         ASSERT_RTNL();
2137
2138         if ((idev = ipv6_find_idev(dev)) == NULL) {
2139                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2140                 return;
2141         }
2142
2143         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2144         if (!IS_ERR(ifp)) {
2145                 spin_lock_bh(&ifp->lock);
2146                 ifp->flags &= ~IFA_F_TENTATIVE;
2147                 spin_unlock_bh(&ifp->lock);
2148                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2149                 in6_ifa_put(ifp);
2150         }
2151 }
2152
2153 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2154 {
2155         struct inet6_ifaddr * ifp;
2156         u32 addr_flags = IFA_F_PERMANENT;
2157
2158 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2159         if (idev->cnf.optimistic_dad &&
2160             !ipv6_devconf.forwarding)
2161                 addr_flags |= IFA_F_OPTIMISTIC;
2162 #endif
2163
2164
2165         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2166         if (!IS_ERR(ifp)) {
2167                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2168                 addrconf_dad_start(ifp, 0);
2169                 in6_ifa_put(ifp);
2170         }
2171 }
2172
2173 static void addrconf_dev_config(struct net_device *dev)
2174 {
2175         struct in6_addr addr;
2176         struct inet6_dev    * idev;
2177
2178         ASSERT_RTNL();
2179
2180         if ((dev->type != ARPHRD_ETHER) &&
2181             (dev->type != ARPHRD_FDDI) &&
2182             (dev->type != ARPHRD_IEEE802_TR) &&
2183             (dev->type != ARPHRD_ARCNET) &&
2184             (dev->type != ARPHRD_INFINIBAND)) {
2185                 /* Alas, we support only Ethernet autoconfiguration. */
2186                 return;
2187         }
2188
2189         idev = addrconf_add_dev(dev);
2190         if (idev == NULL)
2191                 return;
2192
2193         memset(&addr, 0, sizeof(struct in6_addr));
2194         addr.s6_addr32[0] = htonl(0xFE800000);
2195
2196         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2197                 addrconf_add_linklocal(idev, &addr);
2198 }
2199
2200 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2201 static void addrconf_sit_config(struct net_device *dev)
2202 {
2203         struct inet6_dev *idev;
2204
2205         ASSERT_RTNL();
2206
2207         /*
2208          * Configure the tunnel with one of our IPv4
2209          * addresses... we should configure all of
2210          * our v4 addrs in the tunnel
2211          */
2212
2213         if ((idev = ipv6_find_idev(dev)) == NULL) {
2214                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2215                 return;
2216         }
2217
2218         if (dev->priv_flags & IFF_ISATAP) {
2219                 struct in6_addr addr;
2220
2221                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2222                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2223                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2224                         addrconf_add_linklocal(idev, &addr);
2225                 return;
2226         }
2227
2228         sit_add_v4_addrs(idev);
2229
2230         if (dev->flags&IFF_POINTOPOINT) {
2231                 addrconf_add_mroute(dev);
2232                 addrconf_add_lroute(dev);
2233         } else
2234                 sit_route_add(dev);
2235 }
2236 #endif
2237
2238 static inline int
2239 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2240 {
2241         struct in6_addr lladdr;
2242
2243         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2244                 addrconf_add_linklocal(idev, &lladdr);
2245                 return 0;
2246         }
2247         return -1;
2248 }
2249
2250 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2251 {
2252         struct net_device *link_dev;
2253         struct net *net = idev->dev->nd_net;
2254
2255         /* first try to inherit the link-local address from the link device */
2256         if (idev->dev->iflink &&
2257             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2258                 if (!ipv6_inherit_linklocal(idev, link_dev))
2259                         return;
2260         }
2261         /* then try to inherit it from any device */
2262         for_each_netdev(net, link_dev) {
2263                 if (!ipv6_inherit_linklocal(idev, link_dev))
2264                         return;
2265         }
2266         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2267 }
2268
2269 /*
2270  * Autoconfigure tunnel with a link-local address so routing protocols,
2271  * DHCPv6, MLD etc. can be run over the virtual link
2272  */
2273
2274 static void addrconf_ip6_tnl_config(struct net_device *dev)
2275 {
2276         struct inet6_dev *idev;
2277
2278         ASSERT_RTNL();
2279
2280         if ((idev = addrconf_add_dev(dev)) == NULL) {
2281                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2282                 return;
2283         }
2284         ip6_tnl_add_linklocal(idev);
2285 }
2286
2287 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2288                            void * data)
2289 {
2290         struct net_device *dev = (struct net_device *) data;
2291         struct inet6_dev *idev = __in6_dev_get(dev);
2292         int run_pending = 0;
2293         int err;
2294
2295         switch(event) {
2296         case NETDEV_REGISTER:
2297                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2298                         idev = ipv6_add_dev(dev);
2299                         if (!idev)
2300                                 return notifier_from_errno(-ENOMEM);
2301                 }
2302                 break;
2303         case NETDEV_UP:
2304         case NETDEV_CHANGE:
2305                 if (dev->flags & IFF_SLAVE)
2306                         break;
2307
2308                 if (event == NETDEV_UP) {
2309                         if (!addrconf_qdisc_ok(dev)) {
2310                                 /* device is not ready yet. */
2311                                 printk(KERN_INFO
2312                                         "ADDRCONF(NETDEV_UP): %s: "
2313                                         "link is not ready\n",
2314                                         dev->name);
2315                                 break;
2316                         }
2317
2318                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2319                                 idev = ipv6_add_dev(dev);
2320
2321                         if (idev)
2322                                 idev->if_flags |= IF_READY;
2323                 } else {
2324                         if (!addrconf_qdisc_ok(dev)) {
2325                                 /* device is still not ready. */
2326                                 break;
2327                         }
2328
2329                         if (idev) {
2330                                 if (idev->if_flags & IF_READY) {
2331                                         /* device is already configured. */
2332                                         break;
2333                                 }
2334                                 idev->if_flags |= IF_READY;
2335                         }
2336
2337                         printk(KERN_INFO
2338                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2339                                         "link becomes ready\n",
2340                                         dev->name);
2341
2342                         run_pending = 1;
2343                 }
2344
2345                 switch(dev->type) {
2346 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2347                 case ARPHRD_SIT:
2348                         addrconf_sit_config(dev);
2349                         break;
2350 #endif
2351                 case ARPHRD_TUNNEL6:
2352                         addrconf_ip6_tnl_config(dev);
2353                         break;
2354                 case ARPHRD_LOOPBACK:
2355                         init_loopback(dev);
2356                         break;
2357
2358                 default:
2359                         addrconf_dev_config(dev);
2360                         break;
2361                 }
2362                 if (idev) {
2363                         if (run_pending)
2364                                 addrconf_dad_run(idev);
2365
2366                         /* If the MTU changed during the interface down, when the
2367                            interface up, the changed MTU must be reflected in the
2368                            idev as well as routers.
2369                          */
2370                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2371                                 rt6_mtu_change(dev, dev->mtu);
2372                                 idev->cnf.mtu6 = dev->mtu;
2373                         }
2374                         idev->tstamp = jiffies;
2375                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2376                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2377                            stop IPv6 on this interface.
2378                          */
2379                         if (dev->mtu < IPV6_MIN_MTU)
2380                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2381                 }
2382                 break;
2383
2384         case NETDEV_CHANGEMTU:
2385                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2386                         rt6_mtu_change(dev, dev->mtu);
2387                         idev->cnf.mtu6 = dev->mtu;
2388                         break;
2389                 }
2390
2391                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2392                         idev = ipv6_add_dev(dev);
2393                         if (idev)
2394                                 break;
2395                 }
2396
2397                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2398
2399         case NETDEV_DOWN:
2400         case NETDEV_UNREGISTER:
2401                 /*
2402                  *      Remove all addresses from this interface.
2403                  */
2404                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2405                 break;
2406
2407         case NETDEV_CHANGENAME:
2408                 if (idev) {
2409                         snmp6_unregister_dev(idev);
2410                         addrconf_sysctl_unregister(idev);
2411                         addrconf_sysctl_register(idev);
2412                         err = snmp6_register_dev(idev);
2413                         if (err)
2414                                 return notifier_from_errno(err);
2415                 }
2416                 break;
2417         }
2418
2419         return NOTIFY_OK;
2420 }
2421
2422 /*
2423  *      addrconf module should be notified of a device going up
2424  */
2425 static struct notifier_block ipv6_dev_notf = {
2426         .notifier_call = addrconf_notify,
2427         .priority = 0
2428 };
2429
2430 static int addrconf_ifdown(struct net_device *dev, int how)
2431 {
2432         struct inet6_dev *idev;
2433         struct inet6_ifaddr *ifa, **bifa;
2434         struct net *net = dev->nd_net;
2435         int i;
2436
2437         ASSERT_RTNL();
2438
2439         if (dev == init_net.loopback_dev && how == 1)
2440                 how = 0;
2441
2442         rt6_ifdown(net, dev);
2443         neigh_ifdown(&nd_tbl, dev);
2444
2445         idev = __in6_dev_get(dev);
2446         if (idev == NULL)
2447                 return -ENODEV;
2448
2449         /* Step 1: remove reference to ipv6 device from parent device.
2450                    Do not dev_put!
2451          */
2452         if (how == 1) {
2453                 idev->dead = 1;
2454
2455                 /* protected by rtnl_lock */
2456                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2457
2458                 /* Step 1.5: remove snmp6 entry */
2459                 snmp6_unregister_dev(idev);
2460
2461         }
2462
2463         /* Step 2: clear hash table */
2464         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2465                 bifa = &inet6_addr_lst[i];
2466
2467                 write_lock_bh(&addrconf_hash_lock);
2468                 while ((ifa = *bifa) != NULL) {
2469                         if (ifa->idev == idev) {
2470                                 *bifa = ifa->lst_next;
2471                                 ifa->lst_next = NULL;
2472                                 addrconf_del_timer(ifa);
2473                                 in6_ifa_put(ifa);
2474                                 continue;
2475                         }
2476                         bifa = &ifa->lst_next;
2477                 }
2478                 write_unlock_bh(&addrconf_hash_lock);
2479         }
2480
2481         write_lock_bh(&idev->lock);
2482
2483         /* Step 3: clear flags for stateless addrconf */
2484         if (how != 1)
2485                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2486
2487         /* Step 4: clear address list */
2488 #ifdef CONFIG_IPV6_PRIVACY
2489         if (how == 1 && del_timer(&idev->regen_timer))
2490                 in6_dev_put(idev);
2491
2492         /* clear tempaddr list */
2493         while ((ifa = idev->tempaddr_list) != NULL) {
2494                 idev->tempaddr_list = ifa->tmp_next;
2495                 ifa->tmp_next = NULL;
2496                 ifa->dead = 1;
2497                 write_unlock_bh(&idev->lock);
2498                 spin_lock_bh(&ifa->lock);
2499
2500                 if (ifa->ifpub) {
2501                         in6_ifa_put(ifa->ifpub);
2502                         ifa->ifpub = NULL;
2503                 }
2504                 spin_unlock_bh(&ifa->lock);
2505                 in6_ifa_put(ifa);
2506                 write_lock_bh(&idev->lock);
2507         }
2508 #endif
2509         while ((ifa = idev->addr_list) != NULL) {
2510                 idev->addr_list = ifa->if_next;
2511                 ifa->if_next = NULL;
2512                 ifa->dead = 1;
2513                 addrconf_del_timer(ifa);
2514                 write_unlock_bh(&idev->lock);
2515
2516                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2517                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2518                 in6_ifa_put(ifa);
2519
2520                 write_lock_bh(&idev->lock);
2521         }
2522         write_unlock_bh(&idev->lock);
2523
2524         /* Step 5: Discard multicast list */
2525
2526         if (how == 1)
2527                 ipv6_mc_destroy_dev(idev);
2528         else
2529                 ipv6_mc_down(idev);
2530
2531         idev->tstamp = jiffies;
2532
2533         /* Shot the device (if unregistered) */
2534
2535         if (how == 1) {
2536                 addrconf_sysctl_unregister(idev);
2537                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2538                 neigh_ifdown(&nd_tbl, dev);
2539                 in6_dev_put(idev);
2540         }
2541         return 0;
2542 }
2543
2544 static void addrconf_rs_timer(unsigned long data)
2545 {
2546         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2547
2548         if (ifp->idev->cnf.forwarding)
2549                 goto out;
2550
2551         if (ifp->idev->if_flags & IF_RA_RCVD) {
2552                 /*
2553                  *      Announcement received after solicitation
2554                  *      was sent
2555                  */
2556                 goto out;
2557         }
2558
2559         spin_lock(&ifp->lock);
2560         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2561                 struct in6_addr all_routers;
2562
2563                 /* The wait after the last probe can be shorter */
2564                 addrconf_mod_timer(ifp, AC_RS,
2565                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2566                                    ifp->idev->cnf.rtr_solicit_delay :
2567                                    ifp->idev->cnf.rtr_solicit_interval);
2568                 spin_unlock(&ifp->lock);
2569
2570                 ipv6_addr_all_routers(&all_routers);
2571
2572                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2573         } else {
2574                 spin_unlock(&ifp->lock);
2575                 /*
2576                  * Note: we do not support deprecated "all on-link"
2577                  * assumption any longer.
2578                  */
2579                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2580                        ifp->idev->dev->name);
2581         }
2582
2583 out:
2584         in6_ifa_put(ifp);
2585 }
2586
2587 /*
2588  *      Duplicate Address Detection
2589  */
2590 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2591 {
2592         unsigned long rand_num;
2593         struct inet6_dev *idev = ifp->idev;
2594
2595         if (ifp->flags & IFA_F_OPTIMISTIC)
2596                 rand_num = 0;
2597         else
2598                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2599
2600         ifp->probes = idev->cnf.dad_transmits;
2601         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2602 }
2603
2604 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2605 {
2606         struct inet6_dev *idev = ifp->idev;
2607         struct net_device *dev = idev->dev;
2608
2609         addrconf_join_solict(dev, &ifp->addr);
2610
2611         net_srandom(ifp->addr.s6_addr32[3]);
2612
2613         read_lock_bh(&idev->lock);
2614         if (ifp->dead)
2615                 goto out;
2616         spin_lock_bh(&ifp->lock);
2617
2618         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2619             !(ifp->flags&IFA_F_TENTATIVE) ||
2620             ifp->flags & IFA_F_NODAD) {
2621                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2622                 spin_unlock_bh(&ifp->lock);
2623                 read_unlock_bh(&idev->lock);
2624
2625                 addrconf_dad_completed(ifp);
2626                 return;
2627         }
2628
2629         if (!(idev->if_flags & IF_READY)) {
2630                 spin_unlock_bh(&ifp->lock);
2631                 read_unlock_bh(&idev->lock);
2632                 /*
2633                  * If the defice is not ready:
2634                  * - keep it tentative if it is a permanent address.
2635                  * - otherwise, kill it.
2636                  */
2637                 in6_ifa_hold(ifp);
2638                 addrconf_dad_stop(ifp);
2639                 return;
2640         }
2641
2642         /*
2643          * Optimistic nodes can start receiving
2644          * Frames right away
2645          */
2646         if(ifp->flags & IFA_F_OPTIMISTIC)
2647                 ip6_ins_rt(ifp->rt);
2648
2649         addrconf_dad_kick(ifp);
2650         spin_unlock_bh(&ifp->lock);
2651 out:
2652         read_unlock_bh(&idev->lock);
2653 }
2654
2655 static void addrconf_dad_timer(unsigned long data)
2656 {
2657         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2658         struct inet6_dev *idev = ifp->idev;
2659         struct in6_addr unspec;
2660         struct in6_addr mcaddr;
2661
2662         read_lock_bh(&idev->lock);
2663         if (idev->dead) {
2664                 read_unlock_bh(&idev->lock);
2665                 goto out;
2666         }
2667         spin_lock_bh(&ifp->lock);
2668         if (ifp->probes == 0) {
2669                 /*
2670                  * DAD was successful
2671                  */
2672
2673                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2674                 spin_unlock_bh(&ifp->lock);
2675                 read_unlock_bh(&idev->lock);
2676
2677                 addrconf_dad_completed(ifp);
2678
2679                 goto out;
2680         }
2681
2682         ifp->probes--;
2683         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2684         spin_unlock_bh(&ifp->lock);
2685         read_unlock_bh(&idev->lock);
2686
2687         /* send a neighbour solicitation for our addr */
2688         memset(&unspec, 0, sizeof(unspec));
2689         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2690         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2691 out:
2692         in6_ifa_put(ifp);
2693 }
2694
2695 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2696 {
2697         struct net_device *     dev = ifp->idev->dev;
2698
2699         /*
2700          *      Configure the address for reception. Now it is valid.
2701          */
2702
2703         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2704
2705         /* If added prefix is link local and forwarding is off,
2706            start sending router solicitations.
2707          */
2708
2709         if (ifp->idev->cnf.forwarding == 0 &&
2710             ifp->idev->cnf.rtr_solicits > 0 &&
2711             (dev->flags&IFF_LOOPBACK) == 0 &&
2712             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2713                 struct in6_addr all_routers;
2714
2715                 ipv6_addr_all_routers(&all_routers);
2716
2717                 /*
2718                  *      If a host as already performed a random delay
2719                  *      [...] as part of DAD [...] there is no need
2720                  *      to delay again before sending the first RS
2721                  */
2722                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2723
2724                 spin_lock_bh(&ifp->lock);
2725                 ifp->probes = 1;
2726                 ifp->idev->if_flags |= IF_RS_SENT;
2727                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2728                 spin_unlock_bh(&ifp->lock);
2729         }
2730 }
2731
2732 static void addrconf_dad_run(struct inet6_dev *idev) {
2733         struct inet6_ifaddr *ifp;
2734
2735         read_lock_bh(&idev->lock);
2736         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2737                 spin_lock_bh(&ifp->lock);
2738                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2739                         spin_unlock_bh(&ifp->lock);
2740                         continue;
2741                 }
2742                 spin_unlock_bh(&ifp->lock);
2743                 addrconf_dad_kick(ifp);
2744         }
2745         read_unlock_bh(&idev->lock);
2746 }
2747
2748 #ifdef CONFIG_PROC_FS
2749 struct if6_iter_state {
2750         struct seq_net_private p;
2751         int bucket;
2752 };
2753
2754 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2755 {
2756         struct inet6_ifaddr *ifa = NULL;
2757         struct if6_iter_state *state = seq->private;
2758         struct net *net = state->p.net;
2759
2760         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2761                 ifa = inet6_addr_lst[state->bucket];
2762
2763                 while (ifa && ifa->idev->dev->nd_net != net)
2764                         ifa = ifa->lst_next;
2765                 if (ifa)
2766                         break;
2767         }
2768         return ifa;
2769 }
2770
2771 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2772 {
2773         struct if6_iter_state *state = seq->private;
2774         struct net *net = state->p.net;
2775
2776         ifa = ifa->lst_next;
2777 try_again:
2778         if (ifa) {
2779                 if (ifa->idev->dev->nd_net != net) {
2780                         ifa = ifa->lst_next;
2781                         goto try_again;
2782                 }
2783         }
2784
2785         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2786                 ifa = inet6_addr_lst[state->bucket];
2787                 goto try_again;
2788         }
2789
2790         return ifa;
2791 }
2792
2793 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2794 {
2795         struct inet6_ifaddr *ifa = if6_get_first(seq);
2796
2797         if (ifa)
2798                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2799                         --pos;
2800         return pos ? NULL : ifa;
2801 }
2802
2803 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2804         __acquires(addrconf_hash_lock)
2805 {
2806         read_lock_bh(&addrconf_hash_lock);
2807         return if6_get_idx(seq, *pos);
2808 }
2809
2810 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2811 {
2812         struct inet6_ifaddr *ifa;
2813
2814         ifa = if6_get_next(seq, v);
2815         ++*pos;
2816         return ifa;
2817 }
2818
2819 static void if6_seq_stop(struct seq_file *seq, void *v)
2820         __releases(addrconf_hash_lock)
2821 {
2822         read_unlock_bh(&addrconf_hash_lock);
2823 }
2824
2825 static int if6_seq_show(struct seq_file *seq, void *v)
2826 {
2827         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2828         seq_printf(seq,
2829                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2830                    NIP6(ifp->addr),
2831                    ifp->idev->dev->ifindex,
2832                    ifp->prefix_len,
2833                    ifp->scope,
2834                    ifp->flags,
2835                    ifp->idev->dev->name);
2836         return 0;
2837 }
2838
2839 static const struct seq_operations if6_seq_ops = {
2840         .start  = if6_seq_start,
2841         .next   = if6_seq_next,
2842         .show   = if6_seq_show,
2843         .stop   = if6_seq_stop,
2844 };
2845
2846 static int if6_seq_open(struct inode *inode, struct file *file)
2847 {
2848         return seq_open_net(inode, file, &if6_seq_ops,
2849                             sizeof(struct if6_iter_state));
2850 }
2851
2852 static const struct file_operations if6_fops = {
2853         .owner          = THIS_MODULE,
2854         .open           = if6_seq_open,
2855         .read           = seq_read,
2856         .llseek         = seq_lseek,
2857         .release        = seq_release_net,
2858 };
2859
2860 static int if6_proc_net_init(struct net *net)
2861 {
2862         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2863                 return -ENOMEM;
2864         return 0;
2865 }
2866
2867 static void if6_proc_net_exit(struct net *net)
2868 {
2869        proc_net_remove(net, "if_inet6");
2870 }
2871
2872 static struct pernet_operations if6_proc_net_ops = {
2873        .init = if6_proc_net_init,
2874        .exit = if6_proc_net_exit,
2875 };
2876
2877 int __init if6_proc_init(void)
2878 {
2879         return register_pernet_subsys(&if6_proc_net_ops);
2880 }
2881
2882 void if6_proc_exit(void)
2883 {
2884         unregister_pernet_subsys(&if6_proc_net_ops);
2885 }
2886 #endif  /* CONFIG_PROC_FS */
2887
2888 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2889 /* Check if address is a home address configured on any interface. */
2890 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2891 {
2892         int ret = 0;
2893         struct inet6_ifaddr * ifp;
2894         u8 hash = ipv6_addr_hash(addr);
2895         read_lock_bh(&addrconf_hash_lock);
2896         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2897                 if (ifp->idev->dev->nd_net != net)
2898                         continue;
2899                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2900                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2901                         ret = 1;
2902                         break;
2903                 }
2904         }
2905         read_unlock_bh(&addrconf_hash_lock);
2906         return ret;
2907 }
2908 #endif
2909
2910 /*
2911  *      Periodic address status verification
2912  */
2913
2914 static void addrconf_verify(unsigned long foo)
2915 {
2916         struct inet6_ifaddr *ifp;
2917         unsigned long now, next;
2918         int i;
2919
2920         spin_lock_bh(&addrconf_verify_lock);
2921         now = jiffies;
2922         next = now + ADDR_CHECK_FREQUENCY;
2923
2924         del_timer(&addr_chk_timer);
2925
2926         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2927
2928 restart:
2929                 read_lock(&addrconf_hash_lock);
2930                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2931                         unsigned long age;
2932 #ifdef CONFIG_IPV6_PRIVACY
2933                         unsigned long regen_advance;
2934 #endif
2935
2936                         if (ifp->flags & IFA_F_PERMANENT)
2937                                 continue;
2938
2939                         spin_lock(&ifp->lock);
2940                         age = (now - ifp->tstamp) / HZ;
2941
2942 #ifdef CONFIG_IPV6_PRIVACY
2943                         regen_advance = ifp->idev->cnf.regen_max_retry *
2944                                         ifp->idev->cnf.dad_transmits *
2945                                         ifp->idev->nd_parms->retrans_time / HZ;
2946 #endif
2947
2948                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2949                             age >= ifp->valid_lft) {
2950                                 spin_unlock(&ifp->lock);
2951                                 in6_ifa_hold(ifp);
2952                                 read_unlock(&addrconf_hash_lock);
2953                                 ipv6_del_addr(ifp);
2954                                 goto restart;
2955                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2956                                 spin_unlock(&ifp->lock);
2957                                 continue;
2958                         } else if (age >= ifp->prefered_lft) {
2959                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2960                                 int deprecate = 0;
2961
2962                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2963                                         deprecate = 1;
2964                                         ifp->flags |= IFA_F_DEPRECATED;
2965                                 }
2966
2967                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2968                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2969
2970                                 spin_unlock(&ifp->lock);
2971
2972                                 if (deprecate) {
2973                                         in6_ifa_hold(ifp);
2974                                         read_unlock(&addrconf_hash_lock);
2975
2976                                         ipv6_ifa_notify(0, ifp);
2977                                         in6_ifa_put(ifp);
2978                                         goto restart;
2979                                 }
2980 #ifdef CONFIG_IPV6_PRIVACY
2981                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2982                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2983                                 if (age >= ifp->prefered_lft - regen_advance) {
2984                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2985                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2986                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2987                                         if (!ifp->regen_count && ifpub) {
2988                                                 ifp->regen_count++;
2989                                                 in6_ifa_hold(ifp);
2990                                                 in6_ifa_hold(ifpub);
2991                                                 spin_unlock(&ifp->lock);
2992                                                 read_unlock(&addrconf_hash_lock);
2993                                                 spin_lock(&ifpub->lock);
2994                                                 ifpub->regen_count = 0;
2995                                                 spin_unlock(&ifpub->lock);
2996                                                 ipv6_create_tempaddr(ifpub, ifp);
2997                                                 in6_ifa_put(ifpub);
2998                                                 in6_ifa_put(ifp);
2999                                                 goto restart;
3000                                         }
3001                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3002                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3003                                 spin_unlock(&ifp->lock);
3004 #endif
3005                         } else {
3006                                 /* ifp->prefered_lft <= ifp->valid_lft */
3007                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3008                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3009                                 spin_unlock(&ifp->lock);
3010                         }
3011                 }
3012                 read_unlock(&addrconf_hash_lock);
3013         }
3014
3015         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3016         add_timer(&addr_chk_timer);
3017         spin_unlock_bh(&addrconf_verify_lock);
3018 }
3019
3020 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3021 {
3022         struct in6_addr *pfx = NULL;
3023
3024         if (addr)
3025                 pfx = nla_data(addr);
3026
3027         if (local) {
3028                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3029                         pfx = NULL;
3030                 else
3031                         pfx = nla_data(local);
3032         }
3033
3034         return pfx;
3035 }
3036
3037 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3038         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3039         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3040         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3041 };
3042
3043 static int
3044 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3045 {
3046         struct net *net = skb->sk->sk_net;
3047         struct ifaddrmsg *ifm;
3048         struct nlattr *tb[IFA_MAX+1];
3049         struct in6_addr *pfx;
3050         int err;
3051
3052         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3053         if (err < 0)
3054                 return err;
3055
3056         ifm = nlmsg_data(nlh);
3057         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3058         if (pfx == NULL)
3059                 return -EINVAL;
3060
3061         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3062 }
3063
3064 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3065                              u32 prefered_lft, u32 valid_lft)
3066 {
3067         u32 flags = RTF_EXPIRES;
3068
3069         if (!valid_lft || (prefered_lft > valid_lft))
3070                 return -EINVAL;
3071
3072         if (valid_lft == INFINITY_LIFE_TIME) {
3073                 ifa_flags |= IFA_F_PERMANENT;
3074                 flags = 0;
3075         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3076                 valid_lft = 0x7FFFFFFF/HZ;
3077
3078         if (prefered_lft == 0)
3079                 ifa_flags |= IFA_F_DEPRECATED;
3080         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3081                  (prefered_lft != INFINITY_LIFE_TIME))
3082                 prefered_lft = 0x7FFFFFFF/HZ;
3083
3084         spin_lock_bh(&ifp->lock);
3085         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3086         ifp->tstamp = jiffies;
3087         ifp->valid_lft = valid_lft;
3088         ifp->prefered_lft = prefered_lft;
3089
3090         spin_unlock_bh(&ifp->lock);
3091         if (!(ifp->flags&IFA_F_TENTATIVE))
3092                 ipv6_ifa_notify(0, ifp);
3093
3094         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3095                               jiffies_to_clock_t(valid_lft * HZ), flags);
3096         addrconf_verify(0);
3097
3098         return 0;
3099 }
3100
3101 static int
3102 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3103 {
3104         struct net *net = skb->sk->sk_net;
3105         struct ifaddrmsg *ifm;
3106         struct nlattr *tb[IFA_MAX+1];
3107         struct in6_addr *pfx;
3108         struct inet6_ifaddr *ifa;
3109         struct net_device *dev;
3110         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3111         u8 ifa_flags;
3112         int err;
3113
3114         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3115         if (err < 0)
3116                 return err;
3117
3118         ifm = nlmsg_data(nlh);
3119         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3120         if (pfx == NULL)
3121                 return -EINVAL;
3122
3123         if (tb[IFA_CACHEINFO]) {
3124                 struct ifa_cacheinfo *ci;
3125
3126                 ci = nla_data(tb[IFA_CACHEINFO]);
3127                 valid_lft = ci->ifa_valid;
3128                 preferred_lft = ci->ifa_prefered;
3129         } else {
3130                 preferred_lft = INFINITY_LIFE_TIME;
3131                 valid_lft = INFINITY_LIFE_TIME;
3132         }
3133
3134         dev =  __dev_get_by_index(net, ifm->ifa_index);
3135         if (dev == NULL)
3136                 return -ENODEV;
3137
3138         /* We ignore other flags so far. */
3139         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3140
3141         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3142         if (ifa == NULL) {
3143                 /*
3144                  * It would be best to check for !NLM_F_CREATE here but
3145                  * userspace alreay relies on not having to provide this.
3146                  */
3147                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3148                                       ifm->ifa_prefixlen, ifa_flags,
3149                                       preferred_lft, valid_lft);
3150         }
3151
3152         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3153             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3154                 err = -EEXIST;
3155         else
3156                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3157
3158         in6_ifa_put(ifa);
3159
3160         return err;
3161 }
3162
3163 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3164                           u8 scope, int ifindex)
3165 {
3166         struct ifaddrmsg *ifm;
3167
3168         ifm = nlmsg_data(nlh);
3169         ifm->ifa_family = AF_INET6;
3170         ifm->ifa_prefixlen = prefixlen;
3171         ifm->ifa_flags = flags;
3172         ifm->ifa_scope = scope;
3173         ifm->ifa_index = ifindex;
3174 }
3175
3176 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3177                          unsigned long tstamp, u32 preferred, u32 valid)
3178 {
3179         struct ifa_cacheinfo ci;
3180
3181         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3182                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3183         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3184                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3185         ci.ifa_prefered = preferred;
3186         ci.ifa_valid = valid;
3187
3188         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3189 }
3190
3191 static inline int rt_scope(int ifa_scope)
3192 {
3193         if (ifa_scope & IFA_HOST)
3194                 return RT_SCOPE_HOST;
3195         else if (ifa_scope & IFA_LINK)
3196                 return RT_SCOPE_LINK;
3197         else if (ifa_scope & IFA_SITE)
3198                 return RT_SCOPE_SITE;
3199         else
3200                 return RT_SCOPE_UNIVERSE;
3201 }
3202
3203 static inline int inet6_ifaddr_msgsize(void)
3204 {
3205         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3206                + nla_total_size(16) /* IFA_ADDRESS */
3207                + nla_total_size(sizeof(struct ifa_cacheinfo));
3208 }
3209
3210 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3211                              u32 pid, u32 seq, int event, unsigned int flags)
3212 {
3213         struct nlmsghdr  *nlh;
3214         u32 preferred, valid;
3215
3216         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3217         if (nlh == NULL)
3218                 return -EMSGSIZE;
3219
3220         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3221                       ifa->idev->dev->ifindex);
3222
3223         if (!(ifa->flags&IFA_F_PERMANENT)) {
3224                 preferred = ifa->prefered_lft;
3225                 valid = ifa->valid_lft;
3226                 if (preferred != INFINITY_LIFE_TIME) {
3227                         long tval = (jiffies - ifa->tstamp)/HZ;
3228                         preferred -= tval;
3229                         if (valid != INFINITY_LIFE_TIME)
3230                                 valid -= tval;
3231                 }
3232         } else {
3233                 preferred = INFINITY_LIFE_TIME;
3234                 valid = INFINITY_LIFE_TIME;
3235         }
3236
3237         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3238             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3239                 nlmsg_cancel(skb, nlh);
3240                 return -EMSGSIZE;
3241         }
3242
3243         return nlmsg_end(skb, nlh);
3244 }
3245
3246 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3247                                 u32 pid, u32 seq, int event, u16 flags)
3248 {
3249         struct nlmsghdr  *nlh;
3250         u8 scope = RT_SCOPE_UNIVERSE;
3251         int ifindex = ifmca->idev->dev->ifindex;
3252
3253         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3254                 scope = RT_SCOPE_SITE;
3255
3256         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3257         if (nlh == NULL)
3258                 return -EMSGSIZE;
3259
3260         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3261         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3262             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3263                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3264                 nlmsg_cancel(skb, nlh);
3265                 return -EMSGSIZE;
3266         }
3267
3268         return nlmsg_end(skb, nlh);
3269 }
3270
3271 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3272                                 u32 pid, u32 seq, int event, unsigned int flags)
3273 {
3274         struct nlmsghdr  *nlh;
3275         u8 scope = RT_SCOPE_UNIVERSE;
3276         int ifindex = ifaca->aca_idev->dev->ifindex;
3277
3278         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3279                 scope = RT_SCOPE_SITE;
3280
3281         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3282         if (nlh == NULL)
3283                 return -EMSGSIZE;
3284
3285         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3286         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3287             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3288                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3289                 nlmsg_cancel(skb, nlh);
3290                 return -EMSGSIZE;
3291         }
3292
3293         return nlmsg_end(skb, nlh);
3294 }
3295
3296 enum addr_type_t
3297 {
3298         UNICAST_ADDR,
3299         MULTICAST_ADDR,
3300         ANYCAST_ADDR,
3301 };
3302
3303 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3304                            enum addr_type_t type)
3305 {
3306         int idx, ip_idx;
3307         int s_idx, s_ip_idx;
3308         int err = 1;
3309         struct net_device *dev;
3310         struct inet6_dev *idev = NULL;
3311         struct inet6_ifaddr *ifa;
3312         struct ifmcaddr6 *ifmca;
3313         struct ifacaddr6 *ifaca;
3314         struct net *net = skb->sk->sk_net;
3315
3316         s_idx = cb->args[0];
3317         s_ip_idx = ip_idx = cb->args[1];
3318
3319         idx = 0;
3320         for_each_netdev(net, dev) {
3321                 if (idx < s_idx)
3322                         goto cont;
3323                 if (idx > s_idx)
3324                         s_ip_idx = 0;
3325                 ip_idx = 0;
3326                 if ((idev = in6_dev_get(dev)) == NULL)
3327                         goto cont;
3328                 read_lock_bh(&idev->lock);
3329                 switch (type) {
3330                 case UNICAST_ADDR:
3331                         /* unicast address incl. temp addr */
3332                         for (ifa = idev->addr_list; ifa;
3333                              ifa = ifa->if_next, ip_idx++) {
3334                                 if (ip_idx < s_ip_idx)
3335                                         continue;
3336                                 err = inet6_fill_ifaddr(skb, ifa,
3337                                                         NETLINK_CB(cb->skb).pid,
3338                                                         cb->nlh->nlmsg_seq,
3339                                                         RTM_NEWADDR,
3340                                                         NLM_F_MULTI);
3341                         }
3342                         break;
3343                 case MULTICAST_ADDR:
3344                         /* multicast address */
3345                         for (ifmca = idev->mc_list; ifmca;
3346                              ifmca = ifmca->next, ip_idx++) {
3347                                 if (ip_idx < s_ip_idx)
3348                                         continue;
3349                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3350                                                           NETLINK_CB(cb->skb).pid,
3351                                                           cb->nlh->nlmsg_seq,
3352                                                           RTM_GETMULTICAST,
3353                                                           NLM_F_MULTI);
3354                         }
3355                         break;
3356                 case ANYCAST_ADDR:
3357                         /* anycast address */
3358                         for (ifaca = idev->ac_list; ifaca;
3359                              ifaca = ifaca->aca_next, ip_idx++) {
3360                                 if (ip_idx < s_ip_idx)
3361                                         continue;
3362                                 err = inet6_fill_ifacaddr(skb, ifaca,
3363                                                           NETLINK_CB(cb->skb).pid,
3364                                                           cb->nlh->nlmsg_seq,
3365                                                           RTM_GETANYCAST,
3366                                                           NLM_F_MULTI);
3367                         }
3368                         break;
3369                 default:
3370                         break;
3371                 }
3372                 read_unlock_bh(&idev->lock);
3373                 in6_dev_put(idev);
3374
3375                 if (err <= 0)
3376                         break;
3377 cont:
3378                 idx++;
3379         }
3380         cb->args[0] = idx;
3381         cb->args[1] = ip_idx;
3382         return skb->len;
3383 }
3384
3385 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3386 {
3387         enum addr_type_t type = UNICAST_ADDR;
3388
3389         return inet6_dump_addr(skb, cb, type);
3390 }
3391
3392 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3393 {
3394         enum addr_type_t type = MULTICAST_ADDR;
3395
3396         return inet6_dump_addr(skb, cb, type);
3397 }
3398
3399
3400 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3401 {
3402         enum addr_type_t type = ANYCAST_ADDR;
3403
3404         return inet6_dump_addr(skb, cb, type);
3405 }
3406
3407 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3408                              void *arg)
3409 {
3410         struct net *net = in_skb->sk->sk_net;
3411         struct ifaddrmsg *ifm;
3412         struct nlattr *tb[IFA_MAX+1];
3413         struct in6_addr *addr = NULL;
3414         struct net_device *dev = NULL;
3415         struct inet6_ifaddr *ifa;
3416         struct sk_buff *skb;
3417         int err;
3418
3419         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3420         if (err < 0)
3421                 goto errout;
3422
3423         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3424         if (addr == NULL) {
3425                 err = -EINVAL;
3426                 goto errout;
3427         }
3428
3429         ifm = nlmsg_data(nlh);
3430         if (ifm->ifa_index)
3431                 dev = __dev_get_by_index(net, ifm->ifa_index);
3432
3433         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3434                 err = -EADDRNOTAVAIL;
3435                 goto errout;
3436         }
3437
3438         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3439                 err = -ENOBUFS;
3440                 goto errout_ifa;
3441         }
3442
3443         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3444                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3445         if (err < 0) {
3446                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3447                 WARN_ON(err == -EMSGSIZE);
3448                 kfree_skb(skb);
3449                 goto errout_ifa;
3450         }
3451         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3452 errout_ifa:
3453         in6_ifa_put(ifa);
3454 errout:
3455         return err;
3456 }
3457
3458 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3459 {
3460         struct sk_buff *skb;
3461         struct net *net = ifa->idev->dev->nd_net;
3462         int err = -ENOBUFS;
3463
3464         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3465         if (skb == NULL)
3466                 goto errout;
3467
3468         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3469         if (err < 0) {
3470                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3471                 WARN_ON(err == -EMSGSIZE);
3472                 kfree_skb(skb);
3473                 goto errout;
3474         }
3475         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3476 errout:
3477         if (err < 0)
3478                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3479 }
3480
3481 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3482                                 __s32 *array, int bytes)
3483 {
3484         BUG_ON(bytes < (DEVCONF_MAX * 4));
3485
3486         memset(array, 0, bytes);
3487         array[DEVCONF_FORWARDING] = cnf->forwarding;
3488         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3489         array[DEVCONF_MTU6] = cnf->mtu6;
3490         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3491         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3492         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3493         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3494         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3495         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3496         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3497         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3498 #ifdef CONFIG_IPV6_PRIVACY
3499         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3500         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3501         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3502         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3503         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3504 #endif
3505         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3506         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3507         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3508 #ifdef CONFIG_IPV6_ROUTER_PREF
3509         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3510         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3511 #ifdef CONFIG_IPV6_ROUTE_INFO
3512         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3513 #endif
3514 #endif
3515         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3516         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3517 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3518         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3519 #endif
3520 }
3521
3522 static inline size_t inet6_if_nlmsg_size(void)
3523 {
3524         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3525                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3526                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3527                + nla_total_size(4) /* IFLA_MTU */
3528                + nla_total_size(4) /* IFLA_LINK */
3529                + nla_total_size( /* IFLA_PROTINFO */
3530                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3531                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3532                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3533                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3534                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3535                  );
3536 }
3537
3538 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3539                                       int bytes)
3540 {
3541         int i;
3542         int pad = bytes - sizeof(u64) * items;
3543         BUG_ON(pad < 0);
3544
3545         /* Use put_unaligned() because stats may not be aligned for u64. */
3546         put_unaligned(items, &stats[0]);
3547         for (i = 1; i < items; i++)
3548                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3549
3550         memset(&stats[items], 0, pad);
3551 }
3552
3553 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3554                              int bytes)
3555 {
3556         switch(attrtype) {
3557         case IFLA_INET6_STATS:
3558                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3559                 break;
3560         case IFLA_INET6_ICMP6STATS:
3561                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3562                 break;
3563         }
3564 }
3565
3566 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3567                              u32 pid, u32 seq, int event, unsigned int flags)
3568 {
3569         struct net_device *dev = idev->dev;
3570         struct nlattr *nla;
3571         struct ifinfomsg *hdr;
3572         struct nlmsghdr *nlh;
3573         void *protoinfo;
3574         struct ifla_cacheinfo ci;
3575
3576         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3577         if (nlh == NULL)
3578                 return -EMSGSIZE;
3579
3580         hdr = nlmsg_data(nlh);
3581         hdr->ifi_family = AF_INET6;
3582         hdr->__ifi_pad = 0;
3583         hdr->ifi_type = dev->type;
3584         hdr->ifi_index = dev->ifindex;
3585         hdr->ifi_flags = dev_get_flags(dev);
3586         hdr->ifi_change = 0;
3587
3588         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3589
3590         if (dev->addr_len)
3591                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3592
3593         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3594         if (dev->ifindex != dev->iflink)
3595                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3596
3597         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3598         if (protoinfo == NULL)
3599                 goto nla_put_failure;
3600
3601         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3602
3603         ci.max_reasm_len = IPV6_MAXPLEN;
3604         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3605                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3606         ci.reachable_time = idev->nd_parms->reachable_time;
3607         ci.retrans_time = idev->nd_parms->retrans_time;
3608         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3609
3610         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3611         if (nla == NULL)
3612                 goto nla_put_failure;
3613         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3614
3615         /* XXX - MC not implemented */
3616
3617         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3618         if (nla == NULL)
3619                 goto nla_put_failure;
3620         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3621
3622         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3623         if (nla == NULL)
3624                 goto nla_put_failure;
3625         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3626
3627         nla_nest_end(skb, protoinfo);
3628         return nlmsg_end(skb, nlh);
3629
3630 nla_put_failure:
3631         nlmsg_cancel(skb, nlh);
3632         return -EMSGSIZE;
3633 }
3634
3635 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3636 {
3637         struct net *net = skb->sk->sk_net;
3638         int idx, err;
3639         int s_idx = cb->args[0];
3640         struct net_device *dev;
3641         struct inet6_dev *idev;
3642
3643         read_lock(&dev_base_lock);
3644         idx = 0;
3645         for_each_netdev(net, dev) {
3646                 if (idx < s_idx)
3647                         goto cont;
3648                 if ((idev = in6_dev_get(dev)) == NULL)
3649                         goto cont;
3650                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3651                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3652                 in6_dev_put(idev);
3653                 if (err <= 0)
3654                         break;
3655 cont:
3656                 idx++;
3657         }
3658         read_unlock(&dev_base_lock);
3659         cb->args[0] = idx;
3660
3661         return skb->len;
3662 }
3663
3664 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3665 {
3666         struct sk_buff *skb;
3667         struct net *net = idev->dev->nd_net;
3668         int err = -ENOBUFS;
3669
3670         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3671         if (skb == NULL)
3672                 goto errout;
3673
3674         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3675         if (err < 0) {
3676                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3677                 WARN_ON(err == -EMSGSIZE);
3678                 kfree_skb(skb);
3679                 goto errout;
3680         }
3681         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3682 errout:
3683         if (err < 0)
3684                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3685 }
3686
3687 static inline size_t inet6_prefix_nlmsg_size(void)
3688 {
3689         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3690                + nla_total_size(sizeof(struct in6_addr))
3691                + nla_total_size(sizeof(struct prefix_cacheinfo));
3692 }
3693
3694 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3695                              struct prefix_info *pinfo, u32 pid, u32 seq,
3696                              int event, unsigned int flags)
3697 {
3698         struct prefixmsg *pmsg;
3699         struct nlmsghdr *nlh;
3700         struct prefix_cacheinfo ci;
3701
3702         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3703         if (nlh == NULL)
3704                 return -EMSGSIZE;
3705
3706         pmsg = nlmsg_data(nlh);
3707         pmsg->prefix_family = AF_INET6;
3708         pmsg->prefix_pad1 = 0;
3709         pmsg->prefix_pad2 = 0;
3710         pmsg->prefix_ifindex = idev->dev->ifindex;
3711         pmsg->prefix_len = pinfo->prefix_len;
3712         pmsg->prefix_type = pinfo->type;
3713         pmsg->prefix_pad3 = 0;
3714         pmsg->prefix_flags = 0;
3715         if (pinfo->onlink)
3716                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3717         if (pinfo->autoconf)
3718                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3719
3720         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3721
3722         ci.preferred_time = ntohl(pinfo->prefered);
3723         ci.valid_time = ntohl(pinfo->valid);
3724         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3725
3726         return nlmsg_end(skb, nlh);
3727
3728 nla_put_failure:
3729         nlmsg_cancel(skb, nlh);
3730         return -EMSGSIZE;
3731 }
3732
3733 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3734                          struct prefix_info *pinfo)
3735 {
3736         struct sk_buff *skb;
3737         struct net *net = idev->dev->nd_net;
3738         int err = -ENOBUFS;
3739
3740         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3741         if (skb == NULL)
3742                 goto errout;
3743
3744         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3745         if (err < 0) {
3746                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3747                 WARN_ON(err == -EMSGSIZE);
3748                 kfree_skb(skb);
3749                 goto errout;
3750         }
3751         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3752 errout:
3753         if (err < 0)
3754                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3755 }
3756
3757 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3758 {
3759         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3760
3761         switch (event) {
3762         case RTM_NEWADDR:
3763                 /*
3764                  * If the address was optimistic
3765                  * we inserted the route at the start of
3766                  * our DAD process, so we don't need
3767                  * to do it again
3768                  */
3769                 if (!(ifp->rt->rt6i_node))
3770                         ip6_ins_rt(ifp->rt);
3771                 if (ifp->idev->cnf.forwarding)
3772                         addrconf_join_anycast(ifp);
3773                 break;
3774         case RTM_DELADDR:
3775                 if (ifp->idev->cnf.forwarding)
3776                         addrconf_leave_anycast(ifp);
3777                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3778                 dst_hold(&ifp->rt->u.dst);
3779                 if (ip6_del_rt(ifp->rt))
3780                         dst_free(&ifp->rt->u.dst);
3781                 break;
3782         }
3783 }
3784
3785 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3786 {
3787         rcu_read_lock_bh();
3788         if (likely(ifp->idev->dead == 0))
3789                 __ipv6_ifa_notify(event, ifp);
3790         rcu_read_unlock_bh();
3791 }
3792
3793 #ifdef CONFIG_SYSCTL
3794
3795 static
3796 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3797                            void __user *buffer, size_t *lenp, loff_t *ppos)
3798 {
3799         int *valp = ctl->data;
3800         int val = *valp;
3801         int ret;
3802
3803         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3804
3805         if (write)
3806                 addrconf_fixup_forwarding(ctl, valp, val);
3807         return ret;
3808 }
3809
3810 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3811                                             int __user *name, int nlen,
3812                                             void __user *oldval,
3813                                             size_t __user *oldlenp,
3814                                             void __user *newval, size_t newlen)
3815 {
3816         int *valp = table->data;
3817         int val = *valp;
3818         int new;
3819
3820         if (!newval || !newlen)
3821                 return 0;
3822         if (newlen != sizeof(int))
3823                 return -EINVAL;
3824         if (get_user(new, (int __user *)newval))
3825                 return -EFAULT;
3826         if (new == *valp)
3827                 return 0;
3828         if (oldval && oldlenp) {
3829                 size_t len;
3830                 if (get_user(len, oldlenp))
3831                         return -EFAULT;
3832                 if (len) {
3833                         if (len > table->maxlen)
3834                                 len = table->maxlen;
3835                         if (copy_to_user(oldval, valp, len))
3836                                 return -EFAULT;
3837                         if (put_user(len, oldlenp))
3838                                 return -EFAULT;
3839                 }
3840         }
3841
3842         *valp = new;
3843         addrconf_fixup_forwarding(table, valp, val);
3844         return 1;
3845 }
3846
3847 static struct addrconf_sysctl_table
3848 {
3849         struct ctl_table_header *sysctl_header;
3850         ctl_table addrconf_vars[__NET_IPV6_MAX];
3851         char *dev_name;
3852 } addrconf_sysctl __read_mostly = {
3853         .sysctl_header = NULL,
3854         .addrconf_vars = {
3855                 {
3856                         .ctl_name       =       NET_IPV6_FORWARDING,
3857                         .procname       =       "forwarding",
3858                         .data           =       &ipv6_devconf.forwarding,
3859                         .maxlen         =       sizeof(int),
3860                         .mode           =       0644,
3861                         .proc_handler   =       &addrconf_sysctl_forward,
3862                         .strategy       =       &addrconf_sysctl_forward_strategy,
3863                 },
3864                 {
3865                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3866                         .procname       =       "hop_limit",
3867                         .data           =       &ipv6_devconf.hop_limit,
3868                         .maxlen         =       sizeof(int),
3869                         .mode           =       0644,
3870                         .proc_handler   =       proc_dointvec,
3871                 },
3872                 {
3873                         .ctl_name       =       NET_IPV6_MTU,
3874                         .procname       =       "mtu",
3875                         .data           =       &ipv6_devconf.mtu6,
3876                         .maxlen         =       sizeof(int),
3877                         .mode           =       0644,
3878                         .proc_handler   =       &proc_dointvec,
3879                 },
3880                 {
3881                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3882                         .procname       =       "accept_ra",
3883                         .data           =       &ipv6_devconf.accept_ra,
3884                         .maxlen         =       sizeof(int),
3885                         .mode           =       0644,
3886                         .proc_handler   =       &proc_dointvec,
3887                 },
3888                 {
3889                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3890                         .procname       =       "accept_redirects",
3891                         .data           =       &ipv6_devconf.accept_redirects,
3892                         .maxlen         =       sizeof(int),
3893                         .mode           =       0644,
3894                         .proc_handler   =       &proc_dointvec,
3895                 },
3896                 {
3897                         .ctl_name       =       NET_IPV6_AUTOCONF,
3898                         .procname       =       "autoconf",
3899                         .data           =       &ipv6_devconf.autoconf,
3900                         .maxlen         =       sizeof(int),
3901                         .mode           =       0644,
3902                         .proc_handler   =       &proc_dointvec,
3903                 },
3904                 {
3905                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3906                         .procname       =       "dad_transmits",
3907                         .data           =       &ipv6_devconf.dad_transmits,
3908                         .maxlen         =       sizeof(int),
3909                         .mode           =       0644,
3910                         .proc_handler   =       &proc_dointvec,
3911                 },
3912                 {
3913                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3914                         .procname       =       "router_solicitations",
3915                         .data           =       &ipv6_devconf.rtr_solicits,
3916                         .maxlen         =       sizeof(int),
3917                         .mode           =       0644,
3918                         .proc_handler   =       &proc_dointvec,
3919                 },
3920                 {
3921                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3922                         .procname       =       "router_solicitation_interval",
3923                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3924                         .maxlen         =       sizeof(int),
3925                         .mode           =       0644,
3926                         .proc_handler   =       &proc_dointvec_jiffies,
3927                         .strategy       =       &sysctl_jiffies,
3928                 },
3929                 {
3930                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3931                         .procname       =       "router_solicitation_delay",
3932                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3933                         .maxlen         =       sizeof(int),
3934                         .mode           =       0644,
3935                         .proc_handler   =       &proc_dointvec_jiffies,
3936                         .strategy       =       &sysctl_jiffies,
3937                 },
3938                 {
3939                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3940                         .procname       =       "force_mld_version",
3941                         .data           =       &ipv6_devconf.force_mld_version,
3942                         .maxlen         =       sizeof(int),
3943                         .mode           =       0644,
3944                         .proc_handler   =       &proc_dointvec,
3945                 },
3946 #ifdef CONFIG_IPV6_PRIVACY
3947                 {
3948                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3949                         .procname       =       "use_tempaddr",
3950                         .data           =       &ipv6_devconf.use_tempaddr,
3951                         .maxlen         =       sizeof(int),
3952                         .mode           =       0644,
3953                         .proc_handler   =       &proc_dointvec,
3954                 },
3955                 {
3956                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3957                         .procname       =       "temp_valid_lft",
3958                         .data           =       &ipv6_devconf.temp_valid_lft,
3959                         .maxlen         =       sizeof(int),
3960                         .mode           =       0644,
3961                         .proc_handler   =       &proc_dointvec,
3962                 },
3963                 {
3964                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3965                         .procname       =       "temp_prefered_lft",
3966                         .data           =       &ipv6_devconf.temp_prefered_lft,
3967                         .maxlen         =       sizeof(int),
3968                         .mode           =       0644,
3969                         .proc_handler   =       &proc_dointvec,
3970                 },
3971                 {
3972                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3973                         .procname       =       "regen_max_retry",
3974                         .data           =       &ipv6_devconf.regen_max_retry,
3975                         .maxlen         =       sizeof(int),
3976                         .mode           =       0644,
3977                         .proc_handler   =       &proc_dointvec,
3978                 },
3979                 {
3980                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3981                         .procname       =       "max_desync_factor",
3982                         .data           =       &ipv6_devconf.max_desync_factor,
3983                         .maxlen         =       sizeof(int),
3984                         .mode           =       0644,
3985                         .proc_handler   =       &proc_dointvec,
3986                 },
3987 #endif
3988                 {
3989                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3990                         .procname       =       "max_addresses",
3991                         .data           =       &ipv6_devconf.max_addresses,
3992                         .maxlen         =       sizeof(int),
3993                         .mode           =       0644,
3994                         .proc_handler   =       &proc_dointvec,
3995                 },
3996                 {
3997                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3998                         .procname       =       "accept_ra_defrtr",
3999                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4000                         .maxlen         =       sizeof(int),
4001                         .mode           =       0644,
4002                         .proc_handler   =       &proc_dointvec,
4003                 },
4004                 {
4005                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4006                         .procname       =       "accept_ra_pinfo",
4007                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4008                         .maxlen         =       sizeof(int),
4009                         .mode           =       0644,
4010                         .proc_handler   =       &proc_dointvec,
4011                 },
4012 #ifdef CONFIG_IPV6_ROUTER_PREF
4013                 {
4014                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4015                         .procname       =       "accept_ra_rtr_pref",
4016                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4017                         .maxlen         =       sizeof(int),
4018                         .mode           =       0644,
4019                         .proc_handler   =       &proc_dointvec,
4020                 },
4021                 {
4022                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4023                         .procname       =       "router_probe_interval",
4024                         .data           =       &ipv6_devconf.rtr_probe_interval,
4025                         .maxlen         =       sizeof(int),
4026                         .mode           =       0644,
4027                         .proc_handler   =       &proc_dointvec_jiffies,
4028                         .strategy       =       &sysctl_jiffies,
4029                 },
4030 #ifdef CONFIG_IPV6_ROUTE_INFO
4031                 {
4032                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4033                         .procname       =       "accept_ra_rt_info_max_plen",
4034                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4035                         .maxlen         =       sizeof(int),
4036                         .mode           =       0644,
4037                         .proc_handler   =       &proc_dointvec,
4038                 },
4039 #endif
4040 #endif
4041                 {
4042                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4043                         .procname       =       "proxy_ndp",
4044                         .data           =       &ipv6_devconf.proxy_ndp,
4045                         .maxlen         =       sizeof(int),
4046                         .mode           =       0644,
4047                         .proc_handler   =       &proc_dointvec,
4048                 },
4049                 {
4050                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4051                         .procname       =       "accept_source_route",
4052                         .data           =       &ipv6_devconf.accept_source_route,
4053                         .maxlen         =       sizeof(int),
4054                         .mode           =       0644,
4055                         .proc_handler   =       &proc_dointvec,
4056                 },
4057 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4058                 {
4059                         .ctl_name       =       CTL_UNNUMBERED,
4060                         .procname       =       "optimistic_dad",
4061                         .data           =       &ipv6_devconf.optimistic_dad,
4062                         .maxlen         =       sizeof(int),
4063                         .mode           =       0644,
4064                         .proc_handler   =       &proc_dointvec,
4065
4066                 },
4067 #endif
4068                 {
4069                         .ctl_name       =       0,      /* sentinel */
4070                 }
4071         },
4072 };
4073
4074 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4075                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4076 {
4077         int i;
4078         struct addrconf_sysctl_table *t;
4079
4080 #define ADDRCONF_CTL_PATH_DEV   3
4081
4082         struct ctl_path addrconf_ctl_path[] = {
4083                 { .procname = "net", .ctl_name = CTL_NET, },
4084                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4085                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4086                 { /* to be set */ },
4087                 { },
4088         };
4089
4090
4091         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4092         if (t == NULL)
4093                 goto out;
4094
4095         for (i=0; t->addrconf_vars[i].data; i++) {
4096                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4097                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4098                 t->addrconf_vars[i].extra2 = net;
4099         }
4100
4101         /*
4102          * Make a copy of dev_name, because '.procname' is regarded as const
4103          * by sysctl and we wouldn't want anyone to change it under our feet
4104          * (see SIOCSIFNAME).
4105          */
4106         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4107         if (!t->dev_name)
4108                 goto free;
4109
4110         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4111         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4112
4113         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4114                         t->addrconf_vars);
4115         if (t->sysctl_header == NULL)
4116                 goto free_procname;
4117
4118         p->sysctl = t;
4119         return 0;
4120
4121 free_procname:
4122         kfree(t->dev_name);
4123 free:
4124         kfree(t);
4125 out:
4126         return -ENOBUFS;
4127 }
4128
4129 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4130 {
4131         struct addrconf_sysctl_table *t;
4132
4133         if (p->sysctl == NULL)
4134                 return;
4135
4136         t = p->sysctl;
4137         p->sysctl = NULL;
4138         unregister_sysctl_table(t->sysctl_header);
4139         kfree(t->dev_name);
4140         kfree(t);
4141 }
4142
4143 static void addrconf_sysctl_register(struct inet6_dev *idev)
4144 {
4145         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4146                               NET_IPV6_NEIGH, "ipv6",
4147                               &ndisc_ifinfo_sysctl_change,
4148                               NULL);
4149         __addrconf_sysctl_register(idev->dev->nd_net, idev->dev->name,
4150                         idev->dev->ifindex, idev, &idev->cnf);
4151 }
4152
4153 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4154 {
4155         __addrconf_sysctl_unregister(&idev->cnf);
4156         neigh_sysctl_unregister(idev->nd_parms);
4157 }
4158
4159
4160 #endif
4161
4162 static int addrconf_init_net(struct net *net)
4163 {
4164         int err;
4165         struct ipv6_devconf *all, *dflt;
4166
4167         err = -ENOMEM;
4168         all = &ipv6_devconf;
4169         dflt = &ipv6_devconf_dflt;
4170
4171         if (net != &init_net) {
4172                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4173                 if (all == NULL)
4174                         goto err_alloc_all;
4175
4176                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4177                 if (dflt == NULL)
4178                         goto err_alloc_dflt;
4179         }
4180
4181         net->ipv6.devconf_all = all;
4182         net->ipv6.devconf_dflt = dflt;
4183
4184 #ifdef CONFIG_SYSCTL
4185         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4186                         NULL, all);
4187         if (err < 0)
4188                 goto err_reg_all;
4189
4190         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4191                         NULL, dflt);
4192         if (err < 0)
4193                 goto err_reg_dflt;
4194 #endif
4195         return 0;
4196
4197 #ifdef CONFIG_SYSCTL
4198 err_reg_dflt:
4199         __addrconf_sysctl_unregister(all);
4200 err_reg_all:
4201         kfree(dflt);
4202 #endif
4203 err_alloc_dflt:
4204         kfree(all);
4205 err_alloc_all:
4206         return err;
4207 }
4208
4209 static void addrconf_exit_net(struct net *net)
4210 {
4211 #ifdef CONFIG_SYSCTL
4212         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4213         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4214 #endif
4215         if (net != &init_net) {
4216                 kfree(net->ipv6.devconf_dflt);
4217                 kfree(net->ipv6.devconf_all);
4218         }
4219 }
4220
4221 static struct pernet_operations addrconf_ops = {
4222         .init = addrconf_init_net,
4223         .exit = addrconf_exit_net,
4224 };
4225
4226 /*
4227  *      Device notifier
4228  */
4229
4230 int register_inet6addr_notifier(struct notifier_block *nb)
4231 {
4232         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4233 }
4234
4235 EXPORT_SYMBOL(register_inet6addr_notifier);
4236
4237 int unregister_inet6addr_notifier(struct notifier_block *nb)
4238 {
4239         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4240 }
4241
4242 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4243
4244
4245 static int addrconf_net_init(struct net *net)
4246 {
4247         return 0;
4248 }
4249
4250 static void addrconf_net_exit(struct net *net)
4251 {
4252         struct net_device *dev;
4253
4254         rtnl_lock();
4255         /* clean dev list */
4256         for_each_netdev(net, dev) {
4257                 if (__in6_dev_get(dev) == NULL)
4258                         continue;
4259                 addrconf_ifdown(dev, 1);
4260         }
4261         addrconf_ifdown(net->loopback_dev, 2);
4262         rtnl_unlock();
4263 }
4264
4265 static struct pernet_operations addrconf_net_ops = {
4266         .init = addrconf_net_init,
4267         .exit = addrconf_net_exit,
4268 };
4269
4270 /*
4271  *      Init / cleanup code
4272  */
4273
4274 int __init addrconf_init(void)
4275 {
4276         int err;
4277
4278         if ((err = ipv6_addr_label_init()) < 0) {
4279                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4280                         err);
4281                 return err;
4282         }
4283
4284         register_pernet_subsys(&addrconf_ops);
4285
4286         /* The addrconf netdev notifier requires that loopback_dev
4287          * has it's ipv6 private information allocated and setup
4288          * before it can bring up and give link-local addresses
4289          * to other devices which are up.
4290          *
4291          * Unfortunately, loopback_dev is not necessarily the first
4292          * entry in the global dev_base list of net devices.  In fact,
4293          * it is likely to be the very last entry on that list.
4294          * So this causes the notifier registry below to try and
4295          * give link-local addresses to all devices besides loopback_dev
4296          * first, then loopback_dev, which cases all the non-loopback_dev
4297          * devices to fail to get a link-local address.
4298          *
4299          * So, as a temporary fix, allocate the ipv6 structure for
4300          * loopback_dev first by hand.
4301          * Longer term, all of the dependencies ipv6 has upon the loopback
4302          * device and it being up should be removed.
4303          */
4304         rtnl_lock();
4305         if (!ipv6_add_dev(init_net.loopback_dev))
4306                 err = -ENOMEM;
4307         rtnl_unlock();
4308         if (err)
4309                 goto errlo;
4310
4311         err = register_pernet_device(&addrconf_net_ops);
4312         if (err)
4313                 return err;
4314
4315         register_netdevice_notifier(&ipv6_dev_notf);
4316
4317         addrconf_verify(0);
4318
4319         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4320         if (err < 0)
4321                 goto errout;
4322
4323         /* Only the first call to __rtnl_register can fail */
4324         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4325         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4326         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4327         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4328         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4329
4330         ipv6_addr_label_rtnl_register();
4331
4332         return 0;
4333 errout:
4334         unregister_netdevice_notifier(&ipv6_dev_notf);
4335 errlo:
4336         unregister_pernet_subsys(&addrconf_ops);
4337
4338         return err;
4339 }
4340
4341 void addrconf_cleanup(void)
4342 {
4343         struct inet6_ifaddr *ifa;
4344         int i;
4345
4346         unregister_netdevice_notifier(&ipv6_dev_notf);
4347         unregister_pernet_device(&addrconf_net_ops);
4348
4349         unregister_pernet_subsys(&addrconf_ops);
4350
4351         rtnl_lock();
4352
4353         /*
4354          *      Check hash table.
4355          */
4356         write_lock_bh(&addrconf_hash_lock);
4357         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4358                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4359                         struct inet6_ifaddr *bifa;
4360
4361                         bifa = ifa;
4362                         ifa = ifa->lst_next;
4363                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4364                         /* Do not free it; something is wrong.
4365                            Now we can investigate it with debugger.
4366                          */
4367                 }
4368         }
4369         write_unlock_bh(&addrconf_hash_lock);
4370
4371         del_timer(&addr_chk_timer);
4372         rtnl_unlock();
4373
4374         unregister_pernet_subsys(&addrconf_net_ops);
4375 }