[IPV6] ADDRCONF: Uninline ipv6_addr_hash().
[linux-2.6.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16
17 /*
18  *      Changes:
19  *
20  *      Janos Farkas                    :       delete timer on ifdown
21  *      <chexum@bankinf.banki.hu>
22  *      Andi Kleen                      :       kill double kfree on module
23  *                                              unload.
24  *      Maciej W. Rozycki               :       FDDI support
25  *      sekiya@USAGI                    :       Don't send too many RS
26  *                                              packets.
27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
28  *                                              packets.
29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
30  *                                              address validation timer.
31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
32  *                                              support.
33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
34  *                                              address on a same interface.
35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
37  *                                              seq_file.
38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
39  *                                              selection; consider scope,
40  *                                              status etc.
41  */
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99
100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
108 {
109 }
110
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 {
113 }
114 #endif
115
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
120
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
123
124 static int ipv6_count_addresses(struct inet6_dev *idev);
125
126 /*
127  *      Configured unicast address hash table
128  */
129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
131
132 static void addrconf_verify(unsigned long);
133
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
136
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139
140 static int addrconf_ifdown(struct net_device *dev, int how);
141
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
149
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151                                 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153                               struct net_device *dev);
154
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156
157 struct ipv6_devconf ipv6_devconf __read_mostly = {
158         .forwarding             = 0,
159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
160         .mtu6                   = IPV6_MIN_MTU,
161         .accept_ra              = 1,
162         .accept_redirects       = 1,
163         .autoconf               = 1,
164         .force_mld_version      = 0,
165         .dad_transmits          = 1,
166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170         .use_tempaddr           = 0,
171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
173         .regen_max_retry        = REGEN_MAX_RETRY,
174         .max_desync_factor      = MAX_DESYNC_FACTOR,
175 #endif
176         .max_addresses          = IPV6_MAX_ADDRESSES,
177         .accept_ra_defrtr       = 1,
178         .accept_ra_pinfo        = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180         .accept_ra_rtr_pref     = 1,
181         .rtr_probe_interval     = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183         .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186         .proxy_ndp              = 0,
187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
188 };
189
190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
191         .forwarding             = 0,
192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
193         .mtu6                   = IPV6_MIN_MTU,
194         .accept_ra              = 1,
195         .accept_redirects       = 1,
196         .autoconf               = 1,
197         .dad_transmits          = 1,
198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
201 #ifdef CONFIG_IPV6_PRIVACY
202         .use_tempaddr           = 0,
203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
205         .regen_max_retry        = REGEN_MAX_RETRY,
206         .max_desync_factor      = MAX_DESYNC_FACTOR,
207 #endif
208         .max_addresses          = IPV6_MAX_ADDRESSES,
209         .accept_ra_defrtr       = 1,
210         .accept_ra_pinfo        = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212         .accept_ra_rtr_pref     = 1,
213         .rtr_probe_interval     = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215         .accept_ra_rt_info_max_plen = 0,
216 #endif
217 #endif
218         .proxy_ndp              = 0,
219         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
220 };
221
222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
225
226 /* Check if a valid qdisc is available */
227 static inline int addrconf_qdisc_ok(struct net_device *dev)
228 {
229         return (dev->qdisc != &noop_qdisc);
230 }
231
232 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
233 {
234         if (del_timer(&ifp->timer))
235                 __in6_ifa_put(ifp);
236 }
237
238 enum addrconf_timer_t
239 {
240         AC_NONE,
241         AC_DAD,
242         AC_RS,
243 };
244
245 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
246                                enum addrconf_timer_t what,
247                                unsigned long when)
248 {
249         if (!del_timer(&ifp->timer))
250                 in6_ifa_hold(ifp);
251
252         switch (what) {
253         case AC_DAD:
254                 ifp->timer.function = addrconf_dad_timer;
255                 break;
256         case AC_RS:
257                 ifp->timer.function = addrconf_rs_timer;
258                 break;
259         default:;
260         }
261         ifp->timer.expires = jiffies + when;
262         add_timer(&ifp->timer);
263 }
264
265 static int snmp6_alloc_dev(struct inet6_dev *idev)
266 {
267         if (snmp_mib_init((void **)idev->stats.ipv6,
268                           sizeof(struct ipstats_mib)) < 0)
269                 goto err_ip;
270         if (snmp_mib_init((void **)idev->stats.icmpv6,
271                           sizeof(struct icmpv6_mib)) < 0)
272                 goto err_icmp;
273         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
274                           sizeof(struct icmpv6msg_mib)) < 0)
275                 goto err_icmpmsg;
276
277         return 0;
278
279 err_icmpmsg:
280         snmp_mib_free((void **)idev->stats.icmpv6);
281 err_icmp:
282         snmp_mib_free((void **)idev->stats.ipv6);
283 err_ip:
284         return -ENOMEM;
285 }
286
287 static void snmp6_free_dev(struct inet6_dev *idev)
288 {
289         snmp_mib_free((void **)idev->stats.icmpv6msg);
290         snmp_mib_free((void **)idev->stats.icmpv6);
291         snmp_mib_free((void **)idev->stats.ipv6);
292 }
293
294 /* Nobody refers to this device, we may destroy it. */
295
296 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
297 {
298         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
299         kfree(idev);
300 }
301
302 void in6_dev_finish_destroy(struct inet6_dev *idev)
303 {
304         struct net_device *dev = idev->dev;
305         BUG_TRAP(idev->addr_list==NULL);
306         BUG_TRAP(idev->mc_list==NULL);
307 #ifdef NET_REFCNT_DEBUG
308         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
309 #endif
310         dev_put(dev);
311         if (!idev->dead) {
312                 printk("Freeing alive inet6 device %p\n", idev);
313                 return;
314         }
315         snmp6_free_dev(idev);
316         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
317 }
318
319 EXPORT_SYMBOL(in6_dev_finish_destroy);
320
321 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
322 {
323         struct inet6_dev *ndev;
324         struct in6_addr maddr;
325
326         ASSERT_RTNL();
327
328         if (dev->mtu < IPV6_MIN_MTU)
329                 return NULL;
330
331         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
332
333         if (ndev == NULL)
334                 return NULL;
335
336         rwlock_init(&ndev->lock);
337         ndev->dev = dev;
338         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
339         ndev->cnf.mtu6 = dev->mtu;
340         ndev->cnf.sysctl = NULL;
341         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
342         if (ndev->nd_parms == NULL) {
343                 kfree(ndev);
344                 return NULL;
345         }
346         /* We refer to the device */
347         dev_hold(dev);
348
349         if (snmp6_alloc_dev(ndev) < 0) {
350                 ADBG((KERN_WARNING
351                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
352                         __func__, dev->name));
353                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
354                 ndev->dead = 1;
355                 in6_dev_finish_destroy(ndev);
356                 return NULL;
357         }
358
359         if (snmp6_register_dev(ndev) < 0) {
360                 ADBG((KERN_WARNING
361                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
362                         __func__, dev->name));
363                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
364                 ndev->dead = 1;
365                 in6_dev_finish_destroy(ndev);
366                 return NULL;
367         }
368
369         /* One reference from device.  We must do this before
370          * we invoke __ipv6_regen_rndid().
371          */
372         in6_dev_hold(ndev);
373
374 #ifdef CONFIG_IPV6_PRIVACY
375         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
376         if ((dev->flags&IFF_LOOPBACK) ||
377             dev->type == ARPHRD_TUNNEL ||
378 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
379             dev->type == ARPHRD_SIT ||
380 #endif
381             dev->type == ARPHRD_NONE) {
382                 printk(KERN_INFO
383                        "%s: Disabled Privacy Extensions\n",
384                        dev->name);
385                 ndev->cnf.use_tempaddr = -1;
386
387                 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
388                         printk(KERN_INFO
389                                "%s: Disabled Multicast RS\n",
390                                dev->name);
391                         ndev->cnf.rtr_solicits = 0;
392                 }
393         } else {
394                 in6_dev_hold(ndev);
395                 ipv6_regen_rndid((unsigned long) ndev);
396         }
397 #endif
398
399         if (netif_running(dev) && addrconf_qdisc_ok(dev))
400                 ndev->if_flags |= IF_READY;
401
402         ipv6_mc_init_dev(ndev);
403         ndev->tstamp = jiffies;
404         addrconf_sysctl_register(ndev);
405         /* protected by rtnl_lock */
406         rcu_assign_pointer(dev->ip6_ptr, ndev);
407
408         /* Join all-node multicast group */
409         ipv6_addr_all_nodes(&maddr);
410         ipv6_dev_mc_inc(dev, &maddr);
411
412         return ndev;
413 }
414
415 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 /*
545  *      Hash function taken from net_alias.c
546  */
547 static u8 ipv6_addr_hash(const struct in6_addr *addr)
548 {
549         __u32 word;
550
551         /*
552          * We perform the hash function over the last 64 bits of the address
553          * This will include the IEEE address token on links that support it.
554          */
555
556         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
557         word ^= (word >> 16);
558         word ^= (word >> 8);
559
560         return ((word ^ (word >> 4)) & 0x0f);
561 }
562
563 /* On success it returns ifp with increased reference count */
564
565 static struct inet6_ifaddr *
566 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
567               int scope, u32 flags)
568 {
569         struct inet6_ifaddr *ifa = NULL;
570         struct rt6_info *rt;
571         int hash;
572         int err = 0;
573
574         rcu_read_lock_bh();
575         if (idev->dead) {
576                 err = -ENODEV;                  /*XXX*/
577                 goto out2;
578         }
579
580         write_lock(&addrconf_hash_lock);
581
582         /* Ignore adding duplicate addresses on an interface */
583         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
584                 ADBG(("ipv6_add_addr: already assigned\n"));
585                 err = -EEXIST;
586                 goto out;
587         }
588
589         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
590
591         if (ifa == NULL) {
592                 ADBG(("ipv6_add_addr: malloc failed\n"));
593                 err = -ENOBUFS;
594                 goto out;
595         }
596
597         rt = addrconf_dst_alloc(idev, addr, 0);
598         if (IS_ERR(rt)) {
599                 err = PTR_ERR(rt);
600                 goto out;
601         }
602
603         ipv6_addr_copy(&ifa->addr, addr);
604
605         spin_lock_init(&ifa->lock);
606         init_timer(&ifa->timer);
607         ifa->timer.data = (unsigned long) ifa;
608         ifa->scope = scope;
609         ifa->prefix_len = pfxlen;
610         ifa->flags = flags | IFA_F_TENTATIVE;
611         ifa->cstamp = ifa->tstamp = jiffies;
612
613         ifa->rt = rt;
614
615         /*
616          * part one of RFC 4429, section 3.3
617          * We should not configure an address as
618          * optimistic if we do not yet know the link
619          * layer address of our nexhop router
620          */
621
622         if (rt->rt6i_nexthop == NULL)
623                 ifa->flags &= ~IFA_F_OPTIMISTIC;
624
625         ifa->idev = idev;
626         in6_dev_hold(idev);
627         /* For caller */
628         in6_ifa_hold(ifa);
629
630         /* Add to big hash table */
631         hash = ipv6_addr_hash(addr);
632
633         ifa->lst_next = inet6_addr_lst[hash];
634         inet6_addr_lst[hash] = ifa;
635         in6_ifa_hold(ifa);
636         write_unlock(&addrconf_hash_lock);
637
638         write_lock(&idev->lock);
639         /* Add to inet6_dev unicast addr list. */
640         ipv6_link_dev_addr(idev, ifa);
641
642 #ifdef CONFIG_IPV6_PRIVACY
643         if (ifa->flags&IFA_F_TEMPORARY) {
644                 ifa->tmp_next = idev->tempaddr_list;
645                 idev->tempaddr_list = ifa;
646                 in6_ifa_hold(ifa);
647         }
648 #endif
649
650         in6_ifa_hold(ifa);
651         write_unlock(&idev->lock);
652 out2:
653         rcu_read_unlock_bh();
654
655         if (likely(err == 0))
656                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
657         else {
658                 kfree(ifa);
659                 ifa = ERR_PTR(err);
660         }
661
662         return ifa;
663 out:
664         write_unlock(&addrconf_hash_lock);
665         goto out2;
666 }
667
668 /* This function wants to get referenced ifp and releases it before return */
669
670 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
671 {
672         struct inet6_ifaddr *ifa, **ifap;
673         struct inet6_dev *idev = ifp->idev;
674         int hash;
675         int deleted = 0, onlink = 0;
676         unsigned long expires = jiffies;
677
678         hash = ipv6_addr_hash(&ifp->addr);
679
680         ifp->dead = 1;
681
682         write_lock_bh(&addrconf_hash_lock);
683         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
684              ifap = &ifa->lst_next) {
685                 if (ifa == ifp) {
686                         *ifap = ifa->lst_next;
687                         __in6_ifa_put(ifp);
688                         ifa->lst_next = NULL;
689                         break;
690                 }
691         }
692         write_unlock_bh(&addrconf_hash_lock);
693
694         write_lock_bh(&idev->lock);
695 #ifdef CONFIG_IPV6_PRIVACY
696         if (ifp->flags&IFA_F_TEMPORARY) {
697                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
698                      ifap = &ifa->tmp_next) {
699                         if (ifa == ifp) {
700                                 *ifap = ifa->tmp_next;
701                                 if (ifp->ifpub) {
702                                         in6_ifa_put(ifp->ifpub);
703                                         ifp->ifpub = NULL;
704                                 }
705                                 __in6_ifa_put(ifp);
706                                 ifa->tmp_next = NULL;
707                                 break;
708                         }
709                 }
710         }
711 #endif
712
713         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
714                 if (ifa == ifp) {
715                         *ifap = ifa->if_next;
716                         __in6_ifa_put(ifp);
717                         ifa->if_next = NULL;
718                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
719                                 break;
720                         deleted = 1;
721                         continue;
722                 } else if (ifp->flags & IFA_F_PERMANENT) {
723                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
724                                               ifp->prefix_len)) {
725                                 if (ifa->flags & IFA_F_PERMANENT) {
726                                         onlink = 1;
727                                         if (deleted)
728                                                 break;
729                                 } else {
730                                         unsigned long lifetime;
731
732                                         if (!onlink)
733                                                 onlink = -1;
734
735                                         spin_lock(&ifa->lock);
736                                         lifetime = min_t(unsigned long,
737                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
738                                         if (time_before(expires,
739                                                         ifa->tstamp + lifetime * HZ))
740                                                 expires = ifa->tstamp + lifetime * HZ;
741                                         spin_unlock(&ifa->lock);
742                                 }
743                         }
744                 }
745                 ifap = &ifa->if_next;
746         }
747         write_unlock_bh(&idev->lock);
748
749         ipv6_ifa_notify(RTM_DELADDR, ifp);
750
751         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
752
753         addrconf_del_timer(ifp);
754
755         /*
756          * Purge or update corresponding prefix
757          *
758          * 1) we don't purge prefix here if address was not permanent.
759          *    prefix is managed by its own lifetime.
760          * 2) if there're no addresses, delete prefix.
761          * 3) if there're still other permanent address(es),
762          *    corresponding prefix is still permanent.
763          * 4) otherwise, update prefix lifetime to the
764          *    longest valid lifetime among the corresponding
765          *    addresses on the device.
766          *    Note: subsequent RA will update lifetime.
767          *
768          * --yoshfuji
769          */
770         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
771                 struct in6_addr prefix;
772                 struct rt6_info *rt;
773                 struct net *net = dev_net(ifp->idev->dev);
774                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
775                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
776
777                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
778                         if (onlink == 0) {
779                                 ip6_del_rt(rt);
780                                 rt = NULL;
781                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
782                                 rt->rt6i_expires = expires;
783                                 rt->rt6i_flags |= RTF_EXPIRES;
784                         }
785                 }
786                 dst_release(&rt->u.dst);
787         }
788
789         in6_ifa_put(ifp);
790 }
791
792 #ifdef CONFIG_IPV6_PRIVACY
793 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
794 {
795         struct inet6_dev *idev = ifp->idev;
796         struct in6_addr addr, *tmpaddr;
797         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
798         unsigned long regen_advance;
799         int tmp_plen;
800         int ret = 0;
801         int max_addresses;
802         u32 addr_flags;
803
804         write_lock(&idev->lock);
805         if (ift) {
806                 spin_lock_bh(&ift->lock);
807                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
808                 spin_unlock_bh(&ift->lock);
809                 tmpaddr = &addr;
810         } else {
811                 tmpaddr = NULL;
812         }
813 retry:
814         in6_dev_hold(idev);
815         if (idev->cnf.use_tempaddr <= 0) {
816                 write_unlock(&idev->lock);
817                 printk(KERN_INFO
818                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
819                 in6_dev_put(idev);
820                 ret = -1;
821                 goto out;
822         }
823         spin_lock_bh(&ifp->lock);
824         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
825                 idev->cnf.use_tempaddr = -1;    /*XXX*/
826                 spin_unlock_bh(&ifp->lock);
827                 write_unlock(&idev->lock);
828                 printk(KERN_WARNING
829                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
830                 in6_dev_put(idev);
831                 ret = -1;
832                 goto out;
833         }
834         in6_ifa_hold(ifp);
835         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
836         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
837                 spin_unlock_bh(&ifp->lock);
838                 write_unlock(&idev->lock);
839                 printk(KERN_WARNING
840                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
841                 in6_ifa_put(ifp);
842                 in6_dev_put(idev);
843                 ret = -1;
844                 goto out;
845         }
846         memcpy(&addr.s6_addr[8], idev->rndid, 8);
847         tmp_valid_lft = min_t(__u32,
848                               ifp->valid_lft,
849                               idev->cnf.temp_valid_lft);
850         tmp_prefered_lft = min_t(__u32,
851                                  ifp->prefered_lft,
852                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
853         tmp_plen = ifp->prefix_len;
854         max_addresses = idev->cnf.max_addresses;
855         tmp_cstamp = ifp->cstamp;
856         tmp_tstamp = ifp->tstamp;
857         spin_unlock_bh(&ifp->lock);
858
859         regen_advance = idev->cnf.regen_max_retry *
860                         idev->cnf.dad_transmits *
861                         idev->nd_parms->retrans_time / HZ;
862         write_unlock(&idev->lock);
863
864         /* A temporary address is created only if this calculated Preferred
865          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
866          * an implementation must not create a temporary address with a zero
867          * Preferred Lifetime.
868          */
869         if (tmp_prefered_lft <= regen_advance) {
870                 in6_ifa_put(ifp);
871                 in6_dev_put(idev);
872                 ret = -1;
873                 goto out;
874         }
875
876         addr_flags = IFA_F_TEMPORARY;
877         /* set in addrconf_prefix_rcv() */
878         if (ifp->flags & IFA_F_OPTIMISTIC)
879                 addr_flags |= IFA_F_OPTIMISTIC;
880
881         ift = !max_addresses ||
882               ipv6_count_addresses(idev) < max_addresses ?
883                 ipv6_add_addr(idev, &addr, tmp_plen,
884                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
885                               addr_flags) : NULL;
886         if (!ift || IS_ERR(ift)) {
887                 in6_ifa_put(ifp);
888                 in6_dev_put(idev);
889                 printk(KERN_INFO
890                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
891                 tmpaddr = &addr;
892                 write_lock(&idev->lock);
893                 goto retry;
894         }
895
896         spin_lock_bh(&ift->lock);
897         ift->ifpub = ifp;
898         ift->valid_lft = tmp_valid_lft;
899         ift->prefered_lft = tmp_prefered_lft;
900         ift->cstamp = tmp_cstamp;
901         ift->tstamp = tmp_tstamp;
902         spin_unlock_bh(&ift->lock);
903
904         addrconf_dad_start(ift, 0);
905         in6_ifa_put(ift);
906         in6_dev_put(idev);
907 out:
908         return ret;
909 }
910 #endif
911
912 /*
913  *      Choose an appropriate source address (RFC3484)
914  */
915 enum {
916         IPV6_SADDR_RULE_INIT = 0,
917         IPV6_SADDR_RULE_LOCAL,
918         IPV6_SADDR_RULE_SCOPE,
919         IPV6_SADDR_RULE_PREFERRED,
920 #ifdef CONFIG_IPV6_MIP6
921         IPV6_SADDR_RULE_HOA,
922 #endif
923         IPV6_SADDR_RULE_OIF,
924         IPV6_SADDR_RULE_LABEL,
925 #ifdef CONFIG_IPV6_PRIVACY
926         IPV6_SADDR_RULE_PRIVACY,
927 #endif
928         IPV6_SADDR_RULE_ORCHID,
929         IPV6_SADDR_RULE_PREFIX,
930         IPV6_SADDR_RULE_MAX
931 };
932
933 struct ipv6_saddr_score {
934         int                     rule;
935         int                     addr_type;
936         struct inet6_ifaddr     *ifa;
937         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
938         int                     scopedist;
939         int                     matchlen;
940 };
941
942 struct ipv6_saddr_dst {
943         struct in6_addr *addr;
944         int ifindex;
945         int scope;
946         int label;
947         unsigned int prefs;
948 };
949
950 static inline int ipv6_saddr_preferred(int type)
951 {
952         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
953                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
954                 return 1;
955         return 0;
956 }
957
958 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
959                                struct ipv6_saddr_dst *dst,
960                                int i)
961 {
962         int ret;
963
964         if (i <= score->rule) {
965                 switch (i) {
966                 case IPV6_SADDR_RULE_SCOPE:
967                         ret = score->scopedist;
968                         break;
969                 case IPV6_SADDR_RULE_PREFIX:
970                         ret = score->matchlen;
971                         break;
972                 default:
973                         ret = !!test_bit(i, score->scorebits);
974                 }
975                 goto out;
976         }
977
978         switch (i) {
979         case IPV6_SADDR_RULE_INIT:
980                 /* Rule 0: remember if hiscore is not ready yet */
981                 ret = !!score->ifa;
982                 break;
983         case IPV6_SADDR_RULE_LOCAL:
984                 /* Rule 1: Prefer same address */
985                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
986                 break;
987         case IPV6_SADDR_RULE_SCOPE:
988                 /* Rule 2: Prefer appropriate scope
989                  *
990                  *      ret
991                  *       ^
992                  *    -1 |  d 15
993                  *    ---+--+-+---> scope
994                  *       |
995                  *       |             d is scope of the destination.
996                  *  B-d  |  \
997                  *       |   \      <- smaller scope is better if
998                  *  B-15 |    \        if scope is enough for destinaion.
999                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1000                  * d-C-1 | /
1001                  *       |/         <- greater is better
1002                  *   -C  /             if scope is not enough for destination.
1003                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1004                  *
1005                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1006                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1007                  * Assume B = 0 and we get C > 29.
1008                  */
1009                 ret = __ipv6_addr_src_scope(score->addr_type);
1010                 if (ret >= dst->scope)
1011                         ret = -ret;
1012                 else
1013                         ret -= 128;     /* 30 is enough */
1014                 score->scopedist = ret;
1015                 break;
1016         case IPV6_SADDR_RULE_PREFERRED:
1017                 /* Rule 3: Avoid deprecated and optimistic addresses */
1018                 ret = ipv6_saddr_preferred(score->addr_type) ||
1019                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1020                 break;
1021 #ifdef CONFIG_IPV6_MIP6
1022         case IPV6_SADDR_RULE_HOA:
1023             {
1024                 /* Rule 4: Prefer home address */
1025                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1026                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1027                 break;
1028             }
1029 #endif
1030         case IPV6_SADDR_RULE_OIF:
1031                 /* Rule 5: Prefer outgoing interface */
1032                 ret = (!dst->ifindex ||
1033                        dst->ifindex == score->ifa->idev->dev->ifindex);
1034                 break;
1035         case IPV6_SADDR_RULE_LABEL:
1036                 /* Rule 6: Prefer matching label */
1037                 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
1038                                       score->ifa->idev->dev->ifindex) == dst->label;
1039                 break;
1040 #ifdef CONFIG_IPV6_PRIVACY
1041         case IPV6_SADDR_RULE_PRIVACY:
1042             {
1043                 /* Rule 7: Prefer public address
1044                  * Note: prefer temprary address if use_tempaddr >= 2
1045                  */
1046                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1047                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1048                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1049                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1050                 break;
1051             }
1052 #endif
1053         case IPV6_SADDR_RULE_ORCHID:
1054                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1055                  *          non-ORCHID vs non-ORCHID
1056                  */
1057                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1058                         ipv6_addr_orchid(dst->addr));
1059                 break;
1060         case IPV6_SADDR_RULE_PREFIX:
1061                 /* Rule 8: Use longest matching prefix */
1062                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1063                                                        dst->addr);
1064                 break;
1065         default:
1066                 ret = 0;
1067         }
1068
1069         if (ret)
1070                 __set_bit(i, score->scorebits);
1071         score->rule = i;
1072 out:
1073         return ret;
1074 }
1075
1076 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1077                        struct in6_addr *daddr, unsigned int prefs,
1078                        struct in6_addr *saddr)
1079 {
1080         struct ipv6_saddr_score scores[2],
1081                                 *score = &scores[0], *hiscore = &scores[1];
1082         struct net *net = dev_net(dst_dev);
1083         struct ipv6_saddr_dst dst;
1084         struct net_device *dev;
1085         int dst_type;
1086
1087         dst_type = __ipv6_addr_type(daddr);
1088         dst.addr = daddr;
1089         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1090         dst.scope = __ipv6_addr_src_scope(dst_type);
1091         dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1092         dst.prefs = prefs;
1093
1094         hiscore->rule = -1;
1095         hiscore->ifa = NULL;
1096
1097         read_lock(&dev_base_lock);
1098         rcu_read_lock();
1099
1100         for_each_netdev(net, dev) {
1101                 struct inet6_dev *idev;
1102
1103                 /* Candidate Source Address (section 4)
1104                  *  - multicast and link-local destination address,
1105                  *    the set of candidate source address MUST only
1106                  *    include addresses assigned to interfaces
1107                  *    belonging to the same link as the outgoing
1108                  *    interface.
1109                  * (- For site-local destination addresses, the
1110                  *    set of candidate source addresses MUST only
1111                  *    include addresses assigned to interfaces
1112                  *    belonging to the same site as the outgoing
1113                  *    interface.)
1114                  */
1115                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1116                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1117                     dst.ifindex && dev->ifindex != dst.ifindex)
1118                         continue;
1119
1120                 idev = __in6_dev_get(dev);
1121                 if (!idev)
1122                         continue;
1123
1124                 read_lock_bh(&idev->lock);
1125                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1126                         int i;
1127
1128                         /*
1129                          * - Tentative Address (RFC2462 section 5.4)
1130                          *  - A tentative address is not considered
1131                          *    "assigned to an interface" in the traditional
1132                          *    sense, unless it is also flagged as optimistic.
1133                          * - Candidate Source Address (section 4)
1134                          *  - In any case, anycast addresses, multicast
1135                          *    addresses, and the unspecified address MUST
1136                          *    NOT be included in a candidate set.
1137                          */
1138                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1139                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1140                                 continue;
1141
1142                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1143
1144                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1145                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1146                                 LIMIT_NETDEBUG(KERN_DEBUG
1147                                                "ADDRCONF: unspecified / multicast address "
1148                                                "assigned as unicast address on %s",
1149                                                dev->name);
1150                                 continue;
1151                         }
1152
1153                         score->rule = -1;
1154                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1155
1156                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1157                                 int minihiscore, miniscore;
1158
1159                                 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1160                                 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1161
1162                                 if (minihiscore > miniscore) {
1163                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1164                                             score->scopedist > 0) {
1165                                                 /*
1166                                                  * special case:
1167                                                  * each remaining entry
1168                                                  * has too small (not enough)
1169                                                  * scope, because ifa entries
1170                                                  * are sorted by their scope
1171                                                  * values.
1172                                                  */
1173                                                 goto try_nextdev;
1174                                         }
1175                                         break;
1176                                 } else if (minihiscore < miniscore) {
1177                                         struct ipv6_saddr_score *tmp;
1178
1179                                         if (hiscore->ifa)
1180                                                 in6_ifa_put(hiscore->ifa);
1181
1182                                         in6_ifa_hold(score->ifa);
1183
1184                                         tmp = hiscore;
1185                                         hiscore = score;
1186                                         score = tmp;
1187
1188                                         /* restore our iterator */
1189                                         score->ifa = hiscore->ifa;
1190
1191                                         break;
1192                                 }
1193                         }
1194                 }
1195 try_nextdev:
1196                 read_unlock_bh(&idev->lock);
1197         }
1198         rcu_read_unlock();
1199         read_unlock(&dev_base_lock);
1200
1201         if (!hiscore->ifa)
1202                 return -EADDRNOTAVAIL;
1203
1204         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1205         in6_ifa_put(hiscore->ifa);
1206         return 0;
1207 }
1208
1209 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1210
1211 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1212                     unsigned char banned_flags)
1213 {
1214         struct inet6_dev *idev;
1215         int err = -EADDRNOTAVAIL;
1216
1217         rcu_read_lock();
1218         if ((idev = __in6_dev_get(dev)) != NULL) {
1219                 struct inet6_ifaddr *ifp;
1220
1221                 read_lock_bh(&idev->lock);
1222                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1223                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1224                                 ipv6_addr_copy(addr, &ifp->addr);
1225                                 err = 0;
1226                                 break;
1227                         }
1228                 }
1229                 read_unlock_bh(&idev->lock);
1230         }
1231         rcu_read_unlock();
1232         return err;
1233 }
1234
1235 static int ipv6_count_addresses(struct inet6_dev *idev)
1236 {
1237         int cnt = 0;
1238         struct inet6_ifaddr *ifp;
1239
1240         read_lock_bh(&idev->lock);
1241         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1242                 cnt++;
1243         read_unlock_bh(&idev->lock);
1244         return cnt;
1245 }
1246
1247 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1248                   struct net_device *dev, int strict)
1249 {
1250         struct inet6_ifaddr * ifp;
1251         u8 hash = ipv6_addr_hash(addr);
1252
1253         read_lock_bh(&addrconf_hash_lock);
1254         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1255                 if (!net_eq(dev_net(ifp->idev->dev), net))
1256                         continue;
1257                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1258                     !(ifp->flags&IFA_F_TENTATIVE)) {
1259                         if (dev == NULL || ifp->idev->dev == dev ||
1260                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1261                                 break;
1262                 }
1263         }
1264         read_unlock_bh(&addrconf_hash_lock);
1265         return ifp != NULL;
1266 }
1267 EXPORT_SYMBOL(ipv6_chk_addr);
1268
1269 static
1270 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1271                        struct net_device *dev)
1272 {
1273         struct inet6_ifaddr * ifp;
1274         u8 hash = ipv6_addr_hash(addr);
1275
1276         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1277                 if (!net_eq(dev_net(ifp->idev->dev), net))
1278                         continue;
1279                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1280                         if (dev == NULL || ifp->idev->dev == dev)
1281                                 break;
1282                 }
1283         }
1284         return ifp != NULL;
1285 }
1286
1287 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1288 {
1289         struct inet6_dev *idev;
1290         struct inet6_ifaddr *ifa;
1291         int     onlink;
1292
1293         onlink = 0;
1294         rcu_read_lock();
1295         idev = __in6_dev_get(dev);
1296         if (idev) {
1297                 read_lock_bh(&idev->lock);
1298                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1299                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1300                                                    ifa->prefix_len);
1301                         if (onlink)
1302                                 break;
1303                 }
1304                 read_unlock_bh(&idev->lock);
1305         }
1306         rcu_read_unlock();
1307         return onlink;
1308 }
1309
1310 EXPORT_SYMBOL(ipv6_chk_prefix);
1311
1312 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1313                                      struct net_device *dev, int strict)
1314 {
1315         struct inet6_ifaddr * ifp;
1316         u8 hash = ipv6_addr_hash(addr);
1317
1318         read_lock_bh(&addrconf_hash_lock);
1319         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1320                 if (!net_eq(dev_net(ifp->idev->dev), net))
1321                         continue;
1322                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1323                         if (dev == NULL || ifp->idev->dev == dev ||
1324                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1325                                 in6_ifa_hold(ifp);
1326                                 break;
1327                         }
1328                 }
1329         }
1330         read_unlock_bh(&addrconf_hash_lock);
1331
1332         return ifp;
1333 }
1334
1335 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1336 {
1337         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1338         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1339         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1340         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1341         int sk_ipv6only = ipv6_only_sock(sk);
1342         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1343         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1344         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1345
1346         if (!sk2_rcv_saddr && !sk_ipv6only)
1347                 return 1;
1348
1349         if (addr_type2 == IPV6_ADDR_ANY &&
1350             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1351                 return 1;
1352
1353         if (addr_type == IPV6_ADDR_ANY &&
1354             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1355                 return 1;
1356
1357         if (sk2_rcv_saddr6 &&
1358             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1359                 return 1;
1360
1361         if (addr_type == IPV6_ADDR_MAPPED &&
1362             !sk2_ipv6only &&
1363             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1364                 return 1;
1365
1366         return 0;
1367 }
1368
1369 /* Gets referenced address, destroys ifaddr */
1370
1371 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1372 {
1373         if (ifp->flags&IFA_F_PERMANENT) {
1374                 spin_lock_bh(&ifp->lock);
1375                 addrconf_del_timer(ifp);
1376                 ifp->flags |= IFA_F_TENTATIVE;
1377                 spin_unlock_bh(&ifp->lock);
1378                 in6_ifa_put(ifp);
1379 #ifdef CONFIG_IPV6_PRIVACY
1380         } else if (ifp->flags&IFA_F_TEMPORARY) {
1381                 struct inet6_ifaddr *ifpub;
1382                 spin_lock_bh(&ifp->lock);
1383                 ifpub = ifp->ifpub;
1384                 if (ifpub) {
1385                         in6_ifa_hold(ifpub);
1386                         spin_unlock_bh(&ifp->lock);
1387                         ipv6_create_tempaddr(ifpub, ifp);
1388                         in6_ifa_put(ifpub);
1389                 } else {
1390                         spin_unlock_bh(&ifp->lock);
1391                 }
1392                 ipv6_del_addr(ifp);
1393 #endif
1394         } else
1395                 ipv6_del_addr(ifp);
1396 }
1397
1398 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1399 {
1400         if (net_ratelimit())
1401                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1402         addrconf_dad_stop(ifp);
1403 }
1404
1405 /* Join to solicited addr multicast group. */
1406
1407 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1408 {
1409         struct in6_addr maddr;
1410
1411         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1412                 return;
1413
1414         addrconf_addr_solict_mult(addr, &maddr);
1415         ipv6_dev_mc_inc(dev, &maddr);
1416 }
1417
1418 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1419 {
1420         struct in6_addr maddr;
1421
1422         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1423                 return;
1424
1425         addrconf_addr_solict_mult(addr, &maddr);
1426         __ipv6_dev_mc_dec(idev, &maddr);
1427 }
1428
1429 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1430 {
1431         struct in6_addr addr;
1432         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1433         if (ipv6_addr_any(&addr))
1434                 return;
1435         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1436 }
1437
1438 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1439 {
1440         struct in6_addr addr;
1441         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1442         if (ipv6_addr_any(&addr))
1443                 return;
1444         __ipv6_dev_ac_dec(ifp->idev, &addr);
1445 }
1446
1447 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1448 {
1449         if (dev->addr_len != ETH_ALEN)
1450                 return -1;
1451         memcpy(eui, dev->dev_addr, 3);
1452         memcpy(eui + 5, dev->dev_addr + 3, 3);
1453
1454         /*
1455          * The zSeries OSA network cards can be shared among various
1456          * OS instances, but the OSA cards have only one MAC address.
1457          * This leads to duplicate address conflicts in conjunction
1458          * with IPv6 if more than one instance uses the same card.
1459          *
1460          * The driver for these cards can deliver a unique 16-bit
1461          * identifier for each instance sharing the same card.  It is
1462          * placed instead of 0xFFFE in the interface identifier.  The
1463          * "u" bit of the interface identifier is not inverted in this
1464          * case.  Hence the resulting interface identifier has local
1465          * scope according to RFC2373.
1466          */
1467         if (dev->dev_id) {
1468                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1469                 eui[4] = dev->dev_id & 0xFF;
1470         } else {
1471                 eui[3] = 0xFF;
1472                 eui[4] = 0xFE;
1473                 eui[0] ^= 2;
1474         }
1475         return 0;
1476 }
1477
1478 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1479 {
1480         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1481         if (dev->addr_len != ARCNET_ALEN)
1482                 return -1;
1483         memset(eui, 0, 7);
1484         eui[7] = *(u8*)dev->dev_addr;
1485         return 0;
1486 }
1487
1488 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1489 {
1490         if (dev->addr_len != INFINIBAND_ALEN)
1491                 return -1;
1492         memcpy(eui, dev->dev_addr + 12, 8);
1493         eui[0] |= 2;
1494         return 0;
1495 }
1496
1497 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1498 {
1499         switch (dev->type) {
1500         case ARPHRD_ETHER:
1501         case ARPHRD_FDDI:
1502         case ARPHRD_IEEE802_TR:
1503                 return addrconf_ifid_eui48(eui, dev);
1504         case ARPHRD_ARCNET:
1505                 return addrconf_ifid_arcnet(eui, dev);
1506         case ARPHRD_INFINIBAND:
1507                 return addrconf_ifid_infiniband(eui, dev);
1508         case ARPHRD_SIT:
1509                 if (dev->priv_flags & IFF_ISATAP)
1510                         return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1511         }
1512         return -1;
1513 }
1514
1515 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1516 {
1517         int err = -1;
1518         struct inet6_ifaddr *ifp;
1519
1520         read_lock_bh(&idev->lock);
1521         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1522                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1523                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1524                         err = 0;
1525                         break;
1526                 }
1527         }
1528         read_unlock_bh(&idev->lock);
1529         return err;
1530 }
1531
1532 #ifdef CONFIG_IPV6_PRIVACY
1533 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1534 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1535 {
1536 regen:
1537         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1538         idev->rndid[0] &= ~0x02;
1539
1540         /*
1541          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1542          * check if generated address is not inappropriate
1543          *
1544          *  - Reserved subnet anycast (RFC 2526)
1545          *      11111101 11....11 1xxxxxxx
1546          *  - ISATAP (RFC4214) 6.1
1547          *      00-00-5E-FE-xx-xx-xx-xx
1548          *  - value 0
1549          *  - XXX: already assigned to an address on the device
1550          */
1551         if (idev->rndid[0] == 0xfd &&
1552             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1553             (idev->rndid[7]&0x80))
1554                 goto regen;
1555         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1556                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1557                         goto regen;
1558                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1559                         goto regen;
1560         }
1561
1562         return 0;
1563 }
1564
1565 static void ipv6_regen_rndid(unsigned long data)
1566 {
1567         struct inet6_dev *idev = (struct inet6_dev *) data;
1568         unsigned long expires;
1569
1570         rcu_read_lock_bh();
1571         write_lock_bh(&idev->lock);
1572
1573         if (idev->dead)
1574                 goto out;
1575
1576         if (__ipv6_regen_rndid(idev) < 0)
1577                 goto out;
1578
1579         expires = jiffies +
1580                 idev->cnf.temp_prefered_lft * HZ -
1581                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1582         if (time_before(expires, jiffies)) {
1583                 printk(KERN_WARNING
1584                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1585                         idev->dev->name);
1586                 goto out;
1587         }
1588
1589         if (!mod_timer(&idev->regen_timer, expires))
1590                 in6_dev_hold(idev);
1591
1592 out:
1593         write_unlock_bh(&idev->lock);
1594         rcu_read_unlock_bh();
1595         in6_dev_put(idev);
1596 }
1597
1598 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1599         int ret = 0;
1600
1601         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1602                 ret = __ipv6_regen_rndid(idev);
1603         return ret;
1604 }
1605 #endif
1606
1607 /*
1608  *      Add prefix route.
1609  */
1610
1611 static void
1612 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1613                       unsigned long expires, u32 flags)
1614 {
1615         struct fib6_config cfg = {
1616                 .fc_table = RT6_TABLE_PREFIX,
1617                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1618                 .fc_ifindex = dev->ifindex,
1619                 .fc_expires = expires,
1620                 .fc_dst_len = plen,
1621                 .fc_flags = RTF_UP | flags,
1622                 .fc_nlinfo.nl_net = dev_net(dev),
1623         };
1624
1625         ipv6_addr_copy(&cfg.fc_dst, pfx);
1626
1627         /* Prevent useless cloning on PtP SIT.
1628            This thing is done here expecting that the whole
1629            class of non-broadcast devices need not cloning.
1630          */
1631 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1632         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1633                 cfg.fc_flags |= RTF_NONEXTHOP;
1634 #endif
1635
1636         ip6_route_add(&cfg);
1637 }
1638
1639 /* Create "default" multicast route to the interface */
1640
1641 static void addrconf_add_mroute(struct net_device *dev)
1642 {
1643         struct fib6_config cfg = {
1644                 .fc_table = RT6_TABLE_LOCAL,
1645                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1646                 .fc_ifindex = dev->ifindex,
1647                 .fc_dst_len = 8,
1648                 .fc_flags = RTF_UP,
1649                 .fc_nlinfo.nl_net = dev_net(dev),
1650         };
1651
1652         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1653
1654         ip6_route_add(&cfg);
1655 }
1656
1657 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1658 static void sit_route_add(struct net_device *dev)
1659 {
1660         struct fib6_config cfg = {
1661                 .fc_table = RT6_TABLE_MAIN,
1662                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1663                 .fc_ifindex = dev->ifindex,
1664                 .fc_dst_len = 96,
1665                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1666                 .fc_nlinfo.nl_net = dev_net(dev),
1667         };
1668
1669         /* prefix length - 96 bits "::d.d.d.d" */
1670         ip6_route_add(&cfg);
1671 }
1672 #endif
1673
1674 static void addrconf_add_lroute(struct net_device *dev)
1675 {
1676         struct in6_addr addr;
1677
1678         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1679         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1680 }
1681
1682 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1683 {
1684         struct inet6_dev *idev;
1685
1686         ASSERT_RTNL();
1687
1688         if ((idev = ipv6_find_idev(dev)) == NULL)
1689                 return NULL;
1690
1691         /* Add default multicast route */
1692         addrconf_add_mroute(dev);
1693
1694         /* Add link local route */
1695         addrconf_add_lroute(dev);
1696         return idev;
1697 }
1698
1699 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1700 {
1701         struct prefix_info *pinfo;
1702         __u32 valid_lft;
1703         __u32 prefered_lft;
1704         int addr_type;
1705         unsigned long rt_expires;
1706         struct inet6_dev *in6_dev;
1707
1708         pinfo = (struct prefix_info *) opt;
1709
1710         if (len < sizeof(struct prefix_info)) {
1711                 ADBG(("addrconf: prefix option too short\n"));
1712                 return;
1713         }
1714
1715         /*
1716          *      Validation checks ([ADDRCONF], page 19)
1717          */
1718
1719         addr_type = ipv6_addr_type(&pinfo->prefix);
1720
1721         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1722                 return;
1723
1724         valid_lft = ntohl(pinfo->valid);
1725         prefered_lft = ntohl(pinfo->prefered);
1726
1727         if (prefered_lft > valid_lft) {
1728                 if (net_ratelimit())
1729                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1730                 return;
1731         }
1732
1733         in6_dev = in6_dev_get(dev);
1734
1735         if (in6_dev == NULL) {
1736                 if (net_ratelimit())
1737                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1738                 return;
1739         }
1740
1741         /*
1742          *      Two things going on here:
1743          *      1) Add routes for on-link prefixes
1744          *      2) Configure prefixes with the auto flag set
1745          */
1746
1747         /* Avoid arithmetic overflow. Really, we could
1748            save rt_expires in seconds, likely valid_lft,
1749            but it would require division in fib gc, that it
1750            not good.
1751          */
1752         if (valid_lft >= 0x7FFFFFFF/HZ)
1753                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1754         else
1755                 rt_expires = valid_lft * HZ;
1756
1757         /*
1758          * We convert this (in jiffies) to clock_t later.
1759          * Avoid arithmetic overflow there as well.
1760          * Overflow can happen only if HZ < USER_HZ.
1761          */
1762         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1763                 rt_expires = 0x7FFFFFFF / USER_HZ;
1764
1765         if (pinfo->onlink) {
1766                 struct rt6_info *rt;
1767                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1768                                 dev->ifindex, 1);
1769
1770                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1771                         if (rt->rt6i_flags&RTF_EXPIRES) {
1772                                 if (valid_lft == 0) {
1773                                         ip6_del_rt(rt);
1774                                         rt = NULL;
1775                                 } else {
1776                                         rt->rt6i_expires = jiffies + rt_expires;
1777                                 }
1778                         }
1779                 } else if (valid_lft) {
1780                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1781                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1782                 }
1783                 if (rt)
1784                         dst_release(&rt->u.dst);
1785         }
1786
1787         /* Try to figure out our local address for this prefix */
1788
1789         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1790                 struct inet6_ifaddr * ifp;
1791                 struct in6_addr addr;
1792                 int create = 0, update_lft = 0;
1793
1794                 if (pinfo->prefix_len == 64) {
1795                         memcpy(&addr, &pinfo->prefix, 8);
1796                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1797                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1798                                 in6_dev_put(in6_dev);
1799                                 return;
1800                         }
1801                         goto ok;
1802                 }
1803                 if (net_ratelimit())
1804                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1805                                pinfo->prefix_len);
1806                 in6_dev_put(in6_dev);
1807                 return;
1808
1809 ok:
1810
1811                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1812
1813                 if (ifp == NULL && valid_lft) {
1814                         int max_addresses = in6_dev->cnf.max_addresses;
1815                         u32 addr_flags = 0;
1816
1817 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1818                         if (in6_dev->cnf.optimistic_dad &&
1819                             !ipv6_devconf.forwarding)
1820                                 addr_flags = IFA_F_OPTIMISTIC;
1821 #endif
1822
1823                         /* Do not allow to create too much of autoconfigured
1824                          * addresses; this would be too easy way to crash kernel.
1825                          */
1826                         if (!max_addresses ||
1827                             ipv6_count_addresses(in6_dev) < max_addresses)
1828                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1829                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1830                                                     addr_flags);
1831
1832                         if (!ifp || IS_ERR(ifp)) {
1833                                 in6_dev_put(in6_dev);
1834                                 return;
1835                         }
1836
1837                         update_lft = create = 1;
1838                         ifp->cstamp = jiffies;
1839                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1840                 }
1841
1842                 if (ifp) {
1843                         int flags;
1844                         unsigned long now;
1845 #ifdef CONFIG_IPV6_PRIVACY
1846                         struct inet6_ifaddr *ift;
1847 #endif
1848                         u32 stored_lft;
1849
1850                         /* update lifetime (RFC2462 5.5.3 e) */
1851                         spin_lock(&ifp->lock);
1852                         now = jiffies;
1853                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1854                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1855                         else
1856                                 stored_lft = 0;
1857                         if (!update_lft && stored_lft) {
1858                                 if (valid_lft > MIN_VALID_LIFETIME ||
1859                                     valid_lft > stored_lft)
1860                                         update_lft = 1;
1861                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1862                                         /* valid_lft <= stored_lft is always true */
1863                                         /* XXX: IPsec */
1864                                         update_lft = 0;
1865                                 } else {
1866                                         valid_lft = MIN_VALID_LIFETIME;
1867                                         if (valid_lft < prefered_lft)
1868                                                 prefered_lft = valid_lft;
1869                                         update_lft = 1;
1870                                 }
1871                         }
1872
1873                         if (update_lft) {
1874                                 ifp->valid_lft = valid_lft;
1875                                 ifp->prefered_lft = prefered_lft;
1876                                 ifp->tstamp = now;
1877                                 flags = ifp->flags;
1878                                 ifp->flags &= ~IFA_F_DEPRECATED;
1879                                 spin_unlock(&ifp->lock);
1880
1881                                 if (!(flags&IFA_F_TENTATIVE))
1882                                         ipv6_ifa_notify(0, ifp);
1883                         } else
1884                                 spin_unlock(&ifp->lock);
1885
1886 #ifdef CONFIG_IPV6_PRIVACY
1887                         read_lock_bh(&in6_dev->lock);
1888                         /* update all temporary addresses in the list */
1889                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1890                                 /*
1891                                  * When adjusting the lifetimes of an existing
1892                                  * temporary address, only lower the lifetimes.
1893                                  * Implementations must not increase the
1894                                  * lifetimes of an existing temporary address
1895                                  * when processing a Prefix Information Option.
1896                                  */
1897                                 if (ifp != ift->ifpub)
1898                                         continue;
1899
1900                                 spin_lock(&ift->lock);
1901                                 flags = ift->flags;
1902                                 if (ift->valid_lft > valid_lft &&
1903                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1904                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1905                                 if (ift->prefered_lft > prefered_lft &&
1906                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1907                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1908                                 spin_unlock(&ift->lock);
1909                                 if (!(flags&IFA_F_TENTATIVE))
1910                                         ipv6_ifa_notify(0, ift);
1911                         }
1912
1913                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1914                                 /*
1915                                  * When a new public address is created as described in [ADDRCONF],
1916                                  * also create a new temporary address.
1917                                  */
1918                                 read_unlock_bh(&in6_dev->lock);
1919                                 ipv6_create_tempaddr(ifp, NULL);
1920                         } else {
1921                                 read_unlock_bh(&in6_dev->lock);
1922                         }
1923 #endif
1924                         in6_ifa_put(ifp);
1925                         addrconf_verify(0);
1926                 }
1927         }
1928         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1929         in6_dev_put(in6_dev);
1930 }
1931
1932 /*
1933  *      Set destination address.
1934  *      Special case for SIT interfaces where we create a new "virtual"
1935  *      device.
1936  */
1937 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1938 {
1939         struct in6_ifreq ireq;
1940         struct net_device *dev;
1941         int err = -EINVAL;
1942
1943         rtnl_lock();
1944
1945         err = -EFAULT;
1946         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1947                 goto err_exit;
1948
1949         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1950
1951         err = -ENODEV;
1952         if (dev == NULL)
1953                 goto err_exit;
1954
1955 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1956         if (dev->type == ARPHRD_SIT) {
1957                 struct ifreq ifr;
1958                 mm_segment_t    oldfs;
1959                 struct ip_tunnel_parm p;
1960
1961                 err = -EADDRNOTAVAIL;
1962                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1963                         goto err_exit;
1964
1965                 memset(&p, 0, sizeof(p));
1966                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1967                 p.iph.saddr = 0;
1968                 p.iph.version = 4;
1969                 p.iph.ihl = 5;
1970                 p.iph.protocol = IPPROTO_IPV6;
1971                 p.iph.ttl = 64;
1972                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1973
1974                 oldfs = get_fs(); set_fs(KERNEL_DS);
1975                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1976                 set_fs(oldfs);
1977
1978                 if (err == 0) {
1979                         err = -ENOBUFS;
1980                         dev = __dev_get_by_name(net, p.name);
1981                         if (!dev)
1982                                 goto err_exit;
1983                         err = dev_open(dev);
1984                 }
1985         }
1986 #endif
1987
1988 err_exit:
1989         rtnl_unlock();
1990         return err;
1991 }
1992
1993 /*
1994  *      Manual configuration of address on an interface
1995  */
1996 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
1997                           int plen, __u8 ifa_flags, __u32 prefered_lft,
1998                           __u32 valid_lft)
1999 {
2000         struct inet6_ifaddr *ifp;
2001         struct inet6_dev *idev;
2002         struct net_device *dev;
2003         int scope;
2004         u32 flags = RTF_EXPIRES;
2005
2006         ASSERT_RTNL();
2007
2008         /* check the lifetime */
2009         if (!valid_lft || prefered_lft > valid_lft)
2010                 return -EINVAL;
2011
2012         dev = __dev_get_by_index(net, ifindex);
2013         if (!dev)
2014                 return -ENODEV;
2015
2016         if ((idev = addrconf_add_dev(dev)) == NULL)
2017                 return -ENOBUFS;
2018
2019         scope = ipv6_addr_scope(pfx);
2020
2021         if (valid_lft == INFINITY_LIFE_TIME) {
2022                 ifa_flags |= IFA_F_PERMANENT;
2023                 flags = 0;
2024         } else if (valid_lft >= 0x7FFFFFFF/HZ)
2025                 valid_lft = 0x7FFFFFFF/HZ;
2026
2027         if (prefered_lft == 0)
2028                 ifa_flags |= IFA_F_DEPRECATED;
2029         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2030                  (prefered_lft != INFINITY_LIFE_TIME))
2031                 prefered_lft = 0x7FFFFFFF/HZ;
2032
2033         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2034
2035         if (!IS_ERR(ifp)) {
2036                 spin_lock_bh(&ifp->lock);
2037                 ifp->valid_lft = valid_lft;
2038                 ifp->prefered_lft = prefered_lft;
2039                 ifp->tstamp = jiffies;
2040                 spin_unlock_bh(&ifp->lock);
2041
2042                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2043                                       jiffies_to_clock_t(valid_lft * HZ), flags);
2044                 /*
2045                  * Note that section 3.1 of RFC 4429 indicates
2046                  * that the Optimistic flag should not be set for
2047                  * manually configured addresses
2048                  */
2049                 addrconf_dad_start(ifp, 0);
2050                 in6_ifa_put(ifp);
2051                 addrconf_verify(0);
2052                 return 0;
2053         }
2054
2055         return PTR_ERR(ifp);
2056 }
2057
2058 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2059                           int plen)
2060 {
2061         struct inet6_ifaddr *ifp;
2062         struct inet6_dev *idev;
2063         struct net_device *dev;
2064
2065         dev = __dev_get_by_index(net, ifindex);
2066         if (!dev)
2067                 return -ENODEV;
2068
2069         if ((idev = __in6_dev_get(dev)) == NULL)
2070                 return -ENXIO;
2071
2072         read_lock_bh(&idev->lock);
2073         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2074                 if (ifp->prefix_len == plen &&
2075                     ipv6_addr_equal(pfx, &ifp->addr)) {
2076                         in6_ifa_hold(ifp);
2077                         read_unlock_bh(&idev->lock);
2078
2079                         ipv6_del_addr(ifp);
2080
2081                         /* If the last address is deleted administratively,
2082                            disable IPv6 on this interface.
2083                          */
2084                         if (idev->addr_list == NULL)
2085                                 addrconf_ifdown(idev->dev, 1);
2086                         return 0;
2087                 }
2088         }
2089         read_unlock_bh(&idev->lock);
2090         return -EADDRNOTAVAIL;
2091 }
2092
2093
2094 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2095 {
2096         struct in6_ifreq ireq;
2097         int err;
2098
2099         if (!capable(CAP_NET_ADMIN))
2100                 return -EPERM;
2101
2102         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2103                 return -EFAULT;
2104
2105         rtnl_lock();
2106         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2107                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2108                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2109         rtnl_unlock();
2110         return err;
2111 }
2112
2113 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2114 {
2115         struct in6_ifreq ireq;
2116         int err;
2117
2118         if (!capable(CAP_NET_ADMIN))
2119                 return -EPERM;
2120
2121         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2122                 return -EFAULT;
2123
2124         rtnl_lock();
2125         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2126                              ireq.ifr6_prefixlen);
2127         rtnl_unlock();
2128         return err;
2129 }
2130
2131 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2132 static void sit_add_v4_addrs(struct inet6_dev *idev)
2133 {
2134         struct inet6_ifaddr * ifp;
2135         struct in6_addr addr;
2136         struct net_device *dev;
2137         struct net *net = dev_net(idev->dev);
2138         int scope;
2139
2140         ASSERT_RTNL();
2141
2142         memset(&addr, 0, sizeof(struct in6_addr));
2143         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2144
2145         if (idev->dev->flags&IFF_POINTOPOINT) {
2146                 addr.s6_addr32[0] = htonl(0xfe800000);
2147                 scope = IFA_LINK;
2148         } else {
2149                 scope = IPV6_ADDR_COMPATv4;
2150         }
2151
2152         if (addr.s6_addr32[3]) {
2153                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2154                 if (!IS_ERR(ifp)) {
2155                         spin_lock_bh(&ifp->lock);
2156                         ifp->flags &= ~IFA_F_TENTATIVE;
2157                         spin_unlock_bh(&ifp->lock);
2158                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2159                         in6_ifa_put(ifp);
2160                 }
2161                 return;
2162         }
2163
2164         for_each_netdev(net, dev) {
2165                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2166                 if (in_dev && (dev->flags & IFF_UP)) {
2167                         struct in_ifaddr * ifa;
2168
2169                         int flag = scope;
2170
2171                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2172                                 int plen;
2173
2174                                 addr.s6_addr32[3] = ifa->ifa_local;
2175
2176                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2177                                         continue;
2178                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2179                                         if (idev->dev->flags&IFF_POINTOPOINT)
2180                                                 continue;
2181                                         flag |= IFA_HOST;
2182                                 }
2183                                 if (idev->dev->flags&IFF_POINTOPOINT)
2184                                         plen = 64;
2185                                 else
2186                                         plen = 96;
2187
2188                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2189                                                     IFA_F_PERMANENT);
2190                                 if (!IS_ERR(ifp)) {
2191                                         spin_lock_bh(&ifp->lock);
2192                                         ifp->flags &= ~IFA_F_TENTATIVE;
2193                                         spin_unlock_bh(&ifp->lock);
2194                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2195                                         in6_ifa_put(ifp);
2196                                 }
2197                         }
2198                 }
2199         }
2200 }
2201 #endif
2202
2203 static void init_loopback(struct net_device *dev)
2204 {
2205         struct inet6_dev  *idev;
2206         struct inet6_ifaddr * ifp;
2207
2208         /* ::1 */
2209
2210         ASSERT_RTNL();
2211
2212         if ((idev = ipv6_find_idev(dev)) == NULL) {
2213                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2214                 return;
2215         }
2216
2217         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2218         if (!IS_ERR(ifp)) {
2219                 spin_lock_bh(&ifp->lock);
2220                 ifp->flags &= ~IFA_F_TENTATIVE;
2221                 spin_unlock_bh(&ifp->lock);
2222                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2223                 in6_ifa_put(ifp);
2224         }
2225 }
2226
2227 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2228 {
2229         struct inet6_ifaddr * ifp;
2230         u32 addr_flags = IFA_F_PERMANENT;
2231
2232 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2233         if (idev->cnf.optimistic_dad &&
2234             !ipv6_devconf.forwarding)
2235                 addr_flags |= IFA_F_OPTIMISTIC;
2236 #endif
2237
2238
2239         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2240         if (!IS_ERR(ifp)) {
2241                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2242                 addrconf_dad_start(ifp, 0);
2243                 in6_ifa_put(ifp);
2244         }
2245 }
2246
2247 static void addrconf_dev_config(struct net_device *dev)
2248 {
2249         struct in6_addr addr;
2250         struct inet6_dev    * idev;
2251
2252         ASSERT_RTNL();
2253
2254         if ((dev->type != ARPHRD_ETHER) &&
2255             (dev->type != ARPHRD_FDDI) &&
2256             (dev->type != ARPHRD_IEEE802_TR) &&
2257             (dev->type != ARPHRD_ARCNET) &&
2258             (dev->type != ARPHRD_INFINIBAND)) {
2259                 /* Alas, we support only Ethernet autoconfiguration. */
2260                 return;
2261         }
2262
2263         idev = addrconf_add_dev(dev);
2264         if (idev == NULL)
2265                 return;
2266
2267         memset(&addr, 0, sizeof(struct in6_addr));
2268         addr.s6_addr32[0] = htonl(0xFE800000);
2269
2270         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2271                 addrconf_add_linklocal(idev, &addr);
2272 }
2273
2274 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2275 static void addrconf_sit_config(struct net_device *dev)
2276 {
2277         struct inet6_dev *idev;
2278
2279         ASSERT_RTNL();
2280
2281         /*
2282          * Configure the tunnel with one of our IPv4
2283          * addresses... we should configure all of
2284          * our v4 addrs in the tunnel
2285          */
2286
2287         if ((idev = ipv6_find_idev(dev)) == NULL) {
2288                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2289                 return;
2290         }
2291
2292         if (dev->priv_flags & IFF_ISATAP) {
2293                 struct in6_addr addr;
2294
2295                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2296                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2297                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2298                         addrconf_add_linklocal(idev, &addr);
2299                 return;
2300         }
2301
2302         sit_add_v4_addrs(idev);
2303
2304         if (dev->flags&IFF_POINTOPOINT) {
2305                 addrconf_add_mroute(dev);
2306                 addrconf_add_lroute(dev);
2307         } else
2308                 sit_route_add(dev);
2309 }
2310 #endif
2311
2312 static inline int
2313 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2314 {
2315         struct in6_addr lladdr;
2316
2317         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2318                 addrconf_add_linklocal(idev, &lladdr);
2319                 return 0;
2320         }
2321         return -1;
2322 }
2323
2324 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2325 {
2326         struct net_device *link_dev;
2327         struct net *net = dev_net(idev->dev);
2328
2329         /* first try to inherit the link-local address from the link device */
2330         if (idev->dev->iflink &&
2331             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2332                 if (!ipv6_inherit_linklocal(idev, link_dev))
2333                         return;
2334         }
2335         /* then try to inherit it from any device */
2336         for_each_netdev(net, link_dev) {
2337                 if (!ipv6_inherit_linklocal(idev, link_dev))
2338                         return;
2339         }
2340         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2341 }
2342
2343 /*
2344  * Autoconfigure tunnel with a link-local address so routing protocols,
2345  * DHCPv6, MLD etc. can be run over the virtual link
2346  */
2347
2348 static void addrconf_ip6_tnl_config(struct net_device *dev)
2349 {
2350         struct inet6_dev *idev;
2351
2352         ASSERT_RTNL();
2353
2354         if ((idev = addrconf_add_dev(dev)) == NULL) {
2355                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2356                 return;
2357         }
2358         ip6_tnl_add_linklocal(idev);
2359 }
2360
2361 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2362                            void * data)
2363 {
2364         struct net_device *dev = (struct net_device *) data;
2365         struct inet6_dev *idev = __in6_dev_get(dev);
2366         int run_pending = 0;
2367         int err;
2368
2369         switch(event) {
2370         case NETDEV_REGISTER:
2371                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2372                         idev = ipv6_add_dev(dev);
2373                         if (!idev)
2374                                 return notifier_from_errno(-ENOMEM);
2375                 }
2376                 break;
2377         case NETDEV_UP:
2378         case NETDEV_CHANGE:
2379                 if (dev->flags & IFF_SLAVE)
2380                         break;
2381
2382                 if (event == NETDEV_UP) {
2383                         if (!addrconf_qdisc_ok(dev)) {
2384                                 /* device is not ready yet. */
2385                                 printk(KERN_INFO
2386                                         "ADDRCONF(NETDEV_UP): %s: "
2387                                         "link is not ready\n",
2388                                         dev->name);
2389                                 break;
2390                         }
2391
2392                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2393                                 idev = ipv6_add_dev(dev);
2394
2395                         if (idev)
2396                                 idev->if_flags |= IF_READY;
2397                 } else {
2398                         if (!addrconf_qdisc_ok(dev)) {
2399                                 /* device is still not ready. */
2400                                 break;
2401                         }
2402
2403                         if (idev) {
2404                                 if (idev->if_flags & IF_READY) {
2405                                         /* device is already configured. */
2406                                         break;
2407                                 }
2408                                 idev->if_flags |= IF_READY;
2409                         }
2410
2411                         printk(KERN_INFO
2412                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2413                                         "link becomes ready\n",
2414                                         dev->name);
2415
2416                         run_pending = 1;
2417                 }
2418
2419                 switch(dev->type) {
2420 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2421                 case ARPHRD_SIT:
2422                         addrconf_sit_config(dev);
2423                         break;
2424 #endif
2425                 case ARPHRD_TUNNEL6:
2426                         addrconf_ip6_tnl_config(dev);
2427                         break;
2428                 case ARPHRD_LOOPBACK:
2429                         init_loopback(dev);
2430                         break;
2431
2432                 default:
2433                         addrconf_dev_config(dev);
2434                         break;
2435                 }
2436                 if (idev) {
2437                         if (run_pending)
2438                                 addrconf_dad_run(idev);
2439
2440                         /* If the MTU changed during the interface down, when the
2441                            interface up, the changed MTU must be reflected in the
2442                            idev as well as routers.
2443                          */
2444                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2445                                 rt6_mtu_change(dev, dev->mtu);
2446                                 idev->cnf.mtu6 = dev->mtu;
2447                         }
2448                         idev->tstamp = jiffies;
2449                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2450                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2451                            stop IPv6 on this interface.
2452                          */
2453                         if (dev->mtu < IPV6_MIN_MTU)
2454                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2455                 }
2456                 break;
2457
2458         case NETDEV_CHANGEMTU:
2459                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2460                         rt6_mtu_change(dev, dev->mtu);
2461                         idev->cnf.mtu6 = dev->mtu;
2462                         break;
2463                 }
2464
2465                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2466                         idev = ipv6_add_dev(dev);
2467                         if (idev)
2468                                 break;
2469                 }
2470
2471                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2472
2473         case NETDEV_DOWN:
2474         case NETDEV_UNREGISTER:
2475                 /*
2476                  *      Remove all addresses from this interface.
2477                  */
2478                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2479                 break;
2480
2481         case NETDEV_CHANGENAME:
2482                 if (idev) {
2483                         snmp6_unregister_dev(idev);
2484                         addrconf_sysctl_unregister(idev);
2485                         addrconf_sysctl_register(idev);
2486                         err = snmp6_register_dev(idev);
2487                         if (err)
2488                                 return notifier_from_errno(err);
2489                 }
2490                 break;
2491         }
2492
2493         return NOTIFY_OK;
2494 }
2495
2496 /*
2497  *      addrconf module should be notified of a device going up
2498  */
2499 static struct notifier_block ipv6_dev_notf = {
2500         .notifier_call = addrconf_notify,
2501         .priority = 0
2502 };
2503
2504 static int addrconf_ifdown(struct net_device *dev, int how)
2505 {
2506         struct inet6_dev *idev;
2507         struct inet6_ifaddr *ifa, **bifa;
2508         struct net *net = dev_net(dev);
2509         int i;
2510
2511         ASSERT_RTNL();
2512
2513         if (dev == init_net.loopback_dev && how == 1)
2514                 how = 0;
2515
2516         rt6_ifdown(net, dev);
2517         neigh_ifdown(&nd_tbl, dev);
2518
2519         idev = __in6_dev_get(dev);
2520         if (idev == NULL)
2521                 return -ENODEV;
2522
2523         /* Step 1: remove reference to ipv6 device from parent device.
2524                    Do not dev_put!
2525          */
2526         if (how == 1) {
2527                 idev->dead = 1;
2528
2529                 /* protected by rtnl_lock */
2530                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2531
2532                 /* Step 1.5: remove snmp6 entry */
2533                 snmp6_unregister_dev(idev);
2534
2535         }
2536
2537         /* Step 2: clear hash table */
2538         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2539                 bifa = &inet6_addr_lst[i];
2540
2541                 write_lock_bh(&addrconf_hash_lock);
2542                 while ((ifa = *bifa) != NULL) {
2543                         if (ifa->idev == idev) {
2544                                 *bifa = ifa->lst_next;
2545                                 ifa->lst_next = NULL;
2546                                 addrconf_del_timer(ifa);
2547                                 in6_ifa_put(ifa);
2548                                 continue;
2549                         }
2550                         bifa = &ifa->lst_next;
2551                 }
2552                 write_unlock_bh(&addrconf_hash_lock);
2553         }
2554
2555         write_lock_bh(&idev->lock);
2556
2557         /* Step 3: clear flags for stateless addrconf */
2558         if (how != 1)
2559                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2560
2561         /* Step 4: clear address list */
2562 #ifdef CONFIG_IPV6_PRIVACY
2563         if (how == 1 && del_timer(&idev->regen_timer))
2564                 in6_dev_put(idev);
2565
2566         /* clear tempaddr list */
2567         while ((ifa = idev->tempaddr_list) != NULL) {
2568                 idev->tempaddr_list = ifa->tmp_next;
2569                 ifa->tmp_next = NULL;
2570                 ifa->dead = 1;
2571                 write_unlock_bh(&idev->lock);
2572                 spin_lock_bh(&ifa->lock);
2573
2574                 if (ifa->ifpub) {
2575                         in6_ifa_put(ifa->ifpub);
2576                         ifa->ifpub = NULL;
2577                 }
2578                 spin_unlock_bh(&ifa->lock);
2579                 in6_ifa_put(ifa);
2580                 write_lock_bh(&idev->lock);
2581         }
2582 #endif
2583         while ((ifa = idev->addr_list) != NULL) {
2584                 idev->addr_list = ifa->if_next;
2585                 ifa->if_next = NULL;
2586                 ifa->dead = 1;
2587                 addrconf_del_timer(ifa);
2588                 write_unlock_bh(&idev->lock);
2589
2590                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2591                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2592                 in6_ifa_put(ifa);
2593
2594                 write_lock_bh(&idev->lock);
2595         }
2596         write_unlock_bh(&idev->lock);
2597
2598         /* Step 5: Discard multicast list */
2599
2600         if (how == 1)
2601                 ipv6_mc_destroy_dev(idev);
2602         else
2603                 ipv6_mc_down(idev);
2604
2605         idev->tstamp = jiffies;
2606
2607         /* Shot the device (if unregistered) */
2608
2609         if (how == 1) {
2610                 addrconf_sysctl_unregister(idev);
2611                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2612                 neigh_ifdown(&nd_tbl, dev);
2613                 in6_dev_put(idev);
2614         }
2615         return 0;
2616 }
2617
2618 static void addrconf_rs_timer(unsigned long data)
2619 {
2620         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2621
2622         if (ifp->idev->cnf.forwarding)
2623                 goto out;
2624
2625         if (ifp->idev->if_flags & IF_RA_RCVD) {
2626                 /*
2627                  *      Announcement received after solicitation
2628                  *      was sent
2629                  */
2630                 goto out;
2631         }
2632
2633         spin_lock(&ifp->lock);
2634         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2635                 struct in6_addr all_routers;
2636
2637                 /* The wait after the last probe can be shorter */
2638                 addrconf_mod_timer(ifp, AC_RS,
2639                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2640                                    ifp->idev->cnf.rtr_solicit_delay :
2641                                    ifp->idev->cnf.rtr_solicit_interval);
2642                 spin_unlock(&ifp->lock);
2643
2644                 ipv6_addr_all_routers(&all_routers);
2645
2646                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2647         } else {
2648                 spin_unlock(&ifp->lock);
2649                 /*
2650                  * Note: we do not support deprecated "all on-link"
2651                  * assumption any longer.
2652                  */
2653                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2654                        ifp->idev->dev->name);
2655         }
2656
2657 out:
2658         in6_ifa_put(ifp);
2659 }
2660
2661 /*
2662  *      Duplicate Address Detection
2663  */
2664 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2665 {
2666         unsigned long rand_num;
2667         struct inet6_dev *idev = ifp->idev;
2668
2669         if (ifp->flags & IFA_F_OPTIMISTIC)
2670                 rand_num = 0;
2671         else
2672                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2673
2674         ifp->probes = idev->cnf.dad_transmits;
2675         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2676 }
2677
2678 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2679 {
2680         struct inet6_dev *idev = ifp->idev;
2681         struct net_device *dev = idev->dev;
2682
2683         addrconf_join_solict(dev, &ifp->addr);
2684
2685         net_srandom(ifp->addr.s6_addr32[3]);
2686
2687         read_lock_bh(&idev->lock);
2688         if (ifp->dead)
2689                 goto out;
2690         spin_lock_bh(&ifp->lock);
2691
2692         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2693             !(ifp->flags&IFA_F_TENTATIVE) ||
2694             ifp->flags & IFA_F_NODAD) {
2695                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2696                 spin_unlock_bh(&ifp->lock);
2697                 read_unlock_bh(&idev->lock);
2698
2699                 addrconf_dad_completed(ifp);
2700                 return;
2701         }
2702
2703         if (!(idev->if_flags & IF_READY)) {
2704                 spin_unlock_bh(&ifp->lock);
2705                 read_unlock_bh(&idev->lock);
2706                 /*
2707                  * If the defice is not ready:
2708                  * - keep it tentative if it is a permanent address.
2709                  * - otherwise, kill it.
2710                  */
2711                 in6_ifa_hold(ifp);
2712                 addrconf_dad_stop(ifp);
2713                 return;
2714         }
2715
2716         /*
2717          * Optimistic nodes can start receiving
2718          * Frames right away
2719          */
2720         if(ifp->flags & IFA_F_OPTIMISTIC)
2721                 ip6_ins_rt(ifp->rt);
2722
2723         addrconf_dad_kick(ifp);
2724         spin_unlock_bh(&ifp->lock);
2725 out:
2726         read_unlock_bh(&idev->lock);
2727 }
2728
2729 static void addrconf_dad_timer(unsigned long data)
2730 {
2731         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2732         struct inet6_dev *idev = ifp->idev;
2733         struct in6_addr unspec;
2734         struct in6_addr mcaddr;
2735
2736         read_lock_bh(&idev->lock);
2737         if (idev->dead) {
2738                 read_unlock_bh(&idev->lock);
2739                 goto out;
2740         }
2741         spin_lock_bh(&ifp->lock);
2742         if (ifp->probes == 0) {
2743                 /*
2744                  * DAD was successful
2745                  */
2746
2747                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2748                 spin_unlock_bh(&ifp->lock);
2749                 read_unlock_bh(&idev->lock);
2750
2751                 addrconf_dad_completed(ifp);
2752
2753                 goto out;
2754         }
2755
2756         ifp->probes--;
2757         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2758         spin_unlock_bh(&ifp->lock);
2759         read_unlock_bh(&idev->lock);
2760
2761         /* send a neighbour solicitation for our addr */
2762         memset(&unspec, 0, sizeof(unspec));
2763         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2764         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2765 out:
2766         in6_ifa_put(ifp);
2767 }
2768
2769 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2770 {
2771         struct net_device *     dev = ifp->idev->dev;
2772
2773         /*
2774          *      Configure the address for reception. Now it is valid.
2775          */
2776
2777         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2778
2779         /* If added prefix is link local and forwarding is off,
2780            start sending router solicitations.
2781          */
2782
2783         if (ifp->idev->cnf.forwarding == 0 &&
2784             ifp->idev->cnf.rtr_solicits > 0 &&
2785             (dev->flags&IFF_LOOPBACK) == 0 &&
2786             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2787                 struct in6_addr all_routers;
2788
2789                 ipv6_addr_all_routers(&all_routers);
2790
2791                 /*
2792                  *      If a host as already performed a random delay
2793                  *      [...] as part of DAD [...] there is no need
2794                  *      to delay again before sending the first RS
2795                  */
2796                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2797
2798                 spin_lock_bh(&ifp->lock);
2799                 ifp->probes = 1;
2800                 ifp->idev->if_flags |= IF_RS_SENT;
2801                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2802                 spin_unlock_bh(&ifp->lock);
2803         }
2804 }
2805
2806 static void addrconf_dad_run(struct inet6_dev *idev) {
2807         struct inet6_ifaddr *ifp;
2808
2809         read_lock_bh(&idev->lock);
2810         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2811                 spin_lock_bh(&ifp->lock);
2812                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2813                         spin_unlock_bh(&ifp->lock);
2814                         continue;
2815                 }
2816                 spin_unlock_bh(&ifp->lock);
2817                 addrconf_dad_kick(ifp);
2818         }
2819         read_unlock_bh(&idev->lock);
2820 }
2821
2822 #ifdef CONFIG_PROC_FS
2823 struct if6_iter_state {
2824         struct seq_net_private p;
2825         int bucket;
2826 };
2827
2828 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2829 {
2830         struct inet6_ifaddr *ifa = NULL;
2831         struct if6_iter_state *state = seq->private;
2832         struct net *net = seq_file_net(seq);
2833
2834         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2835                 ifa = inet6_addr_lst[state->bucket];
2836
2837                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2838                         ifa = ifa->lst_next;
2839                 if (ifa)
2840                         break;
2841         }
2842         return ifa;
2843 }
2844
2845 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2846 {
2847         struct if6_iter_state *state = seq->private;
2848         struct net *net = seq_file_net(seq);
2849
2850         ifa = ifa->lst_next;
2851 try_again:
2852         if (ifa) {
2853                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2854                         ifa = ifa->lst_next;
2855                         goto try_again;
2856                 }
2857         }
2858
2859         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2860                 ifa = inet6_addr_lst[state->bucket];
2861                 goto try_again;
2862         }
2863
2864         return ifa;
2865 }
2866
2867 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2868 {
2869         struct inet6_ifaddr *ifa = if6_get_first(seq);
2870
2871         if (ifa)
2872                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2873                         --pos;
2874         return pos ? NULL : ifa;
2875 }
2876
2877 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2878         __acquires(addrconf_hash_lock)
2879 {
2880         read_lock_bh(&addrconf_hash_lock);
2881         return if6_get_idx(seq, *pos);
2882 }
2883
2884 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2885 {
2886         struct inet6_ifaddr *ifa;
2887
2888         ifa = if6_get_next(seq, v);
2889         ++*pos;
2890         return ifa;
2891 }
2892
2893 static void if6_seq_stop(struct seq_file *seq, void *v)
2894         __releases(addrconf_hash_lock)
2895 {
2896         read_unlock_bh(&addrconf_hash_lock);
2897 }
2898
2899 static int if6_seq_show(struct seq_file *seq, void *v)
2900 {
2901         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2902         seq_printf(seq,
2903                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2904                    NIP6(ifp->addr),
2905                    ifp->idev->dev->ifindex,
2906                    ifp->prefix_len,
2907                    ifp->scope,
2908                    ifp->flags,
2909                    ifp->idev->dev->name);
2910         return 0;
2911 }
2912
2913 static const struct seq_operations if6_seq_ops = {
2914         .start  = if6_seq_start,
2915         .next   = if6_seq_next,
2916         .show   = if6_seq_show,
2917         .stop   = if6_seq_stop,
2918 };
2919
2920 static int if6_seq_open(struct inode *inode, struct file *file)
2921 {
2922         return seq_open_net(inode, file, &if6_seq_ops,
2923                             sizeof(struct if6_iter_state));
2924 }
2925
2926 static const struct file_operations if6_fops = {
2927         .owner          = THIS_MODULE,
2928         .open           = if6_seq_open,
2929         .read           = seq_read,
2930         .llseek         = seq_lseek,
2931         .release        = seq_release_net,
2932 };
2933
2934 static int if6_proc_net_init(struct net *net)
2935 {
2936         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2937                 return -ENOMEM;
2938         return 0;
2939 }
2940
2941 static void if6_proc_net_exit(struct net *net)
2942 {
2943        proc_net_remove(net, "if_inet6");
2944 }
2945
2946 static struct pernet_operations if6_proc_net_ops = {
2947        .init = if6_proc_net_init,
2948        .exit = if6_proc_net_exit,
2949 };
2950
2951 int __init if6_proc_init(void)
2952 {
2953         return register_pernet_subsys(&if6_proc_net_ops);
2954 }
2955
2956 void if6_proc_exit(void)
2957 {
2958         unregister_pernet_subsys(&if6_proc_net_ops);
2959 }
2960 #endif  /* CONFIG_PROC_FS */
2961
2962 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2963 /* Check if address is a home address configured on any interface. */
2964 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2965 {
2966         int ret = 0;
2967         struct inet6_ifaddr * ifp;
2968         u8 hash = ipv6_addr_hash(addr);
2969         read_lock_bh(&addrconf_hash_lock);
2970         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2971                 if (!net_eq(dev_net(ifp->idev->dev), net))
2972                         continue;
2973                 if (ipv6_addr_equal(&ifp->addr, addr) &&
2974                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2975                         ret = 1;
2976                         break;
2977                 }
2978         }
2979         read_unlock_bh(&addrconf_hash_lock);
2980         return ret;
2981 }
2982 #endif
2983
2984 /*
2985  *      Periodic address status verification
2986  */
2987
2988 static void addrconf_verify(unsigned long foo)
2989 {
2990         struct inet6_ifaddr *ifp;
2991         unsigned long now, next;
2992         int i;
2993
2994         spin_lock_bh(&addrconf_verify_lock);
2995         now = jiffies;
2996         next = now + ADDR_CHECK_FREQUENCY;
2997
2998         del_timer(&addr_chk_timer);
2999
3000         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3001
3002 restart:
3003                 read_lock(&addrconf_hash_lock);
3004                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3005                         unsigned long age;
3006 #ifdef CONFIG_IPV6_PRIVACY
3007                         unsigned long regen_advance;
3008 #endif
3009
3010                         if (ifp->flags & IFA_F_PERMANENT)
3011                                 continue;
3012
3013                         spin_lock(&ifp->lock);
3014                         age = (now - ifp->tstamp) / HZ;
3015
3016 #ifdef CONFIG_IPV6_PRIVACY
3017                         regen_advance = ifp->idev->cnf.regen_max_retry *
3018                                         ifp->idev->cnf.dad_transmits *
3019                                         ifp->idev->nd_parms->retrans_time / HZ;
3020 #endif
3021
3022                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3023                             age >= ifp->valid_lft) {
3024                                 spin_unlock(&ifp->lock);
3025                                 in6_ifa_hold(ifp);
3026                                 read_unlock(&addrconf_hash_lock);
3027                                 ipv6_del_addr(ifp);
3028                                 goto restart;
3029                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3030                                 spin_unlock(&ifp->lock);
3031                                 continue;
3032                         } else if (age >= ifp->prefered_lft) {
3033                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3034                                 int deprecate = 0;
3035
3036                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3037                                         deprecate = 1;
3038                                         ifp->flags |= IFA_F_DEPRECATED;
3039                                 }
3040
3041                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3042                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3043
3044                                 spin_unlock(&ifp->lock);
3045
3046                                 if (deprecate) {
3047                                         in6_ifa_hold(ifp);
3048                                         read_unlock(&addrconf_hash_lock);
3049
3050                                         ipv6_ifa_notify(0, ifp);
3051                                         in6_ifa_put(ifp);
3052                                         goto restart;
3053                                 }
3054 #ifdef CONFIG_IPV6_PRIVACY
3055                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3056                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3057                                 if (age >= ifp->prefered_lft - regen_advance) {
3058                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3059                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3060                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3061                                         if (!ifp->regen_count && ifpub) {
3062                                                 ifp->regen_count++;
3063                                                 in6_ifa_hold(ifp);
3064                                                 in6_ifa_hold(ifpub);
3065                                                 spin_unlock(&ifp->lock);
3066                                                 read_unlock(&addrconf_hash_lock);
3067                                                 spin_lock(&ifpub->lock);
3068                                                 ifpub->regen_count = 0;
3069                                                 spin_unlock(&ifpub->lock);
3070                                                 ipv6_create_tempaddr(ifpub, ifp);
3071                                                 in6_ifa_put(ifpub);
3072                                                 in6_ifa_put(ifp);
3073                                                 goto restart;
3074                                         }
3075                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3076                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3077                                 spin_unlock(&ifp->lock);
3078 #endif
3079                         } else {
3080                                 /* ifp->prefered_lft <= ifp->valid_lft */
3081                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3082                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3083                                 spin_unlock(&ifp->lock);
3084                         }
3085                 }
3086                 read_unlock(&addrconf_hash_lock);
3087         }
3088
3089         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3090         add_timer(&addr_chk_timer);
3091         spin_unlock_bh(&addrconf_verify_lock);
3092 }
3093
3094 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3095 {
3096         struct in6_addr *pfx = NULL;
3097
3098         if (addr)
3099                 pfx = nla_data(addr);
3100
3101         if (local) {
3102                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3103                         pfx = NULL;
3104                 else
3105                         pfx = nla_data(local);
3106         }
3107
3108         return pfx;
3109 }
3110
3111 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3112         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3113         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3114         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3115 };
3116
3117 static int
3118 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3119 {
3120         struct net *net = sock_net(skb->sk);
3121         struct ifaddrmsg *ifm;
3122         struct nlattr *tb[IFA_MAX+1];
3123         struct in6_addr *pfx;
3124         int err;
3125
3126         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3127         if (err < 0)
3128                 return err;
3129
3130         ifm = nlmsg_data(nlh);
3131         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3132         if (pfx == NULL)
3133                 return -EINVAL;
3134
3135         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3136 }
3137
3138 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3139                              u32 prefered_lft, u32 valid_lft)
3140 {
3141         u32 flags = RTF_EXPIRES;
3142
3143         if (!valid_lft || (prefered_lft > valid_lft))
3144                 return -EINVAL;
3145
3146         if (valid_lft == INFINITY_LIFE_TIME) {
3147                 ifa_flags |= IFA_F_PERMANENT;
3148                 flags = 0;
3149         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3150                 valid_lft = 0x7FFFFFFF/HZ;
3151
3152         if (prefered_lft == 0)
3153                 ifa_flags |= IFA_F_DEPRECATED;
3154         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3155                  (prefered_lft != INFINITY_LIFE_TIME))
3156                 prefered_lft = 0x7FFFFFFF/HZ;
3157
3158         spin_lock_bh(&ifp->lock);
3159         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3160         ifp->tstamp = jiffies;
3161         ifp->valid_lft = valid_lft;
3162         ifp->prefered_lft = prefered_lft;
3163
3164         spin_unlock_bh(&ifp->lock);
3165         if (!(ifp->flags&IFA_F_TENTATIVE))
3166                 ipv6_ifa_notify(0, ifp);
3167
3168         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3169                               jiffies_to_clock_t(valid_lft * HZ), flags);
3170         addrconf_verify(0);
3171
3172         return 0;
3173 }
3174
3175 static int
3176 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3177 {
3178         struct net *net = sock_net(skb->sk);
3179         struct ifaddrmsg *ifm;
3180         struct nlattr *tb[IFA_MAX+1];
3181         struct in6_addr *pfx;
3182         struct inet6_ifaddr *ifa;
3183         struct net_device *dev;
3184         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3185         u8 ifa_flags;
3186         int err;
3187
3188         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3189         if (err < 0)
3190                 return err;
3191
3192         ifm = nlmsg_data(nlh);
3193         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3194         if (pfx == NULL)
3195                 return -EINVAL;
3196
3197         if (tb[IFA_CACHEINFO]) {
3198                 struct ifa_cacheinfo *ci;
3199
3200                 ci = nla_data(tb[IFA_CACHEINFO]);
3201                 valid_lft = ci->ifa_valid;
3202                 preferred_lft = ci->ifa_prefered;
3203         } else {
3204                 preferred_lft = INFINITY_LIFE_TIME;
3205                 valid_lft = INFINITY_LIFE_TIME;
3206         }
3207
3208         dev =  __dev_get_by_index(net, ifm->ifa_index);
3209         if (dev == NULL)
3210                 return -ENODEV;
3211
3212         /* We ignore other flags so far. */
3213         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3214
3215         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3216         if (ifa == NULL) {
3217                 /*
3218                  * It would be best to check for !NLM_F_CREATE here but
3219                  * userspace alreay relies on not having to provide this.
3220                  */
3221                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3222                                       ifm->ifa_prefixlen, ifa_flags,
3223                                       preferred_lft, valid_lft);
3224         }
3225
3226         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3227             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3228                 err = -EEXIST;
3229         else
3230                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3231
3232         in6_ifa_put(ifa);
3233
3234         return err;
3235 }
3236
3237 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3238                           u8 scope, int ifindex)
3239 {
3240         struct ifaddrmsg *ifm;
3241
3242         ifm = nlmsg_data(nlh);
3243         ifm->ifa_family = AF_INET6;
3244         ifm->ifa_prefixlen = prefixlen;
3245         ifm->ifa_flags = flags;
3246         ifm->ifa_scope = scope;
3247         ifm->ifa_index = ifindex;
3248 }
3249
3250 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3251                          unsigned long tstamp, u32 preferred, u32 valid)
3252 {
3253         struct ifa_cacheinfo ci;
3254
3255         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3256                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3257         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3258                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3259         ci.ifa_prefered = preferred;
3260         ci.ifa_valid = valid;
3261
3262         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3263 }
3264
3265 static inline int rt_scope(int ifa_scope)
3266 {
3267         if (ifa_scope & IFA_HOST)
3268                 return RT_SCOPE_HOST;
3269         else if (ifa_scope & IFA_LINK)
3270                 return RT_SCOPE_LINK;
3271         else if (ifa_scope & IFA_SITE)
3272                 return RT_SCOPE_SITE;
3273         else
3274                 return RT_SCOPE_UNIVERSE;
3275 }
3276
3277 static inline int inet6_ifaddr_msgsize(void)
3278 {
3279         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3280                + nla_total_size(16) /* IFA_ADDRESS */
3281                + nla_total_size(sizeof(struct ifa_cacheinfo));
3282 }
3283
3284 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3285                              u32 pid, u32 seq, int event, unsigned int flags)
3286 {
3287         struct nlmsghdr  *nlh;
3288         u32 preferred, valid;
3289
3290         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3291         if (nlh == NULL)
3292                 return -EMSGSIZE;
3293
3294         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3295                       ifa->idev->dev->ifindex);
3296
3297         if (!(ifa->flags&IFA_F_PERMANENT)) {
3298                 preferred = ifa->prefered_lft;
3299                 valid = ifa->valid_lft;
3300                 if (preferred != INFINITY_LIFE_TIME) {
3301                         long tval = (jiffies - ifa->tstamp)/HZ;
3302                         preferred -= tval;
3303                         if (valid != INFINITY_LIFE_TIME)
3304                                 valid -= tval;
3305                 }
3306         } else {
3307                 preferred = INFINITY_LIFE_TIME;
3308                 valid = INFINITY_LIFE_TIME;
3309         }
3310
3311         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3312             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3313                 nlmsg_cancel(skb, nlh);
3314                 return -EMSGSIZE;
3315         }
3316
3317         return nlmsg_end(skb, nlh);
3318 }
3319
3320 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3321                                 u32 pid, u32 seq, int event, u16 flags)
3322 {
3323         struct nlmsghdr  *nlh;
3324         u8 scope = RT_SCOPE_UNIVERSE;
3325         int ifindex = ifmca->idev->dev->ifindex;
3326
3327         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3328                 scope = RT_SCOPE_SITE;
3329
3330         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3331         if (nlh == NULL)
3332                 return -EMSGSIZE;
3333
3334         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3335         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3336             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3337                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3338                 nlmsg_cancel(skb, nlh);
3339                 return -EMSGSIZE;
3340         }
3341
3342         return nlmsg_end(skb, nlh);
3343 }
3344
3345 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3346                                 u32 pid, u32 seq, int event, unsigned int flags)
3347 {
3348         struct nlmsghdr  *nlh;
3349         u8 scope = RT_SCOPE_UNIVERSE;
3350         int ifindex = ifaca->aca_idev->dev->ifindex;
3351
3352         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3353                 scope = RT_SCOPE_SITE;
3354
3355         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3356         if (nlh == NULL)
3357                 return -EMSGSIZE;
3358
3359         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3360         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3361             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3362                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3363                 nlmsg_cancel(skb, nlh);
3364                 return -EMSGSIZE;
3365         }
3366
3367         return nlmsg_end(skb, nlh);
3368 }
3369
3370 enum addr_type_t
3371 {
3372         UNICAST_ADDR,
3373         MULTICAST_ADDR,
3374         ANYCAST_ADDR,
3375 };
3376
3377 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3378                            enum addr_type_t type)
3379 {
3380         int idx, ip_idx;
3381         int s_idx, s_ip_idx;
3382         int err = 1;
3383         struct net_device *dev;
3384         struct inet6_dev *idev = NULL;
3385         struct inet6_ifaddr *ifa;
3386         struct ifmcaddr6 *ifmca;
3387         struct ifacaddr6 *ifaca;
3388         struct net *net = sock_net(skb->sk);
3389
3390         s_idx = cb->args[0];
3391         s_ip_idx = ip_idx = cb->args[1];
3392
3393         idx = 0;
3394         for_each_netdev(net, dev) {
3395                 if (idx < s_idx)
3396                         goto cont;
3397                 if (idx > s_idx)
3398                         s_ip_idx = 0;
3399                 ip_idx = 0;
3400                 if ((idev = in6_dev_get(dev)) == NULL)
3401                         goto cont;
3402                 read_lock_bh(&idev->lock);
3403                 switch (type) {
3404                 case UNICAST_ADDR:
3405                         /* unicast address incl. temp addr */
3406                         for (ifa = idev->addr_list; ifa;
3407                              ifa = ifa->if_next, ip_idx++) {
3408                                 if (ip_idx < s_ip_idx)
3409                                         continue;
3410                                 err = inet6_fill_ifaddr(skb, ifa,
3411                                                         NETLINK_CB(cb->skb).pid,
3412                                                         cb->nlh->nlmsg_seq,
3413                                                         RTM_NEWADDR,
3414                                                         NLM_F_MULTI);
3415                         }
3416                         break;
3417                 case MULTICAST_ADDR:
3418                         /* multicast address */
3419                         for (ifmca = idev->mc_list; ifmca;
3420                              ifmca = ifmca->next, ip_idx++) {
3421                                 if (ip_idx < s_ip_idx)
3422                                         continue;
3423                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3424                                                           NETLINK_CB(cb->skb).pid,
3425                                                           cb->nlh->nlmsg_seq,
3426                                                           RTM_GETMULTICAST,
3427                                                           NLM_F_MULTI);
3428                         }
3429                         break;
3430                 case ANYCAST_ADDR:
3431                         /* anycast address */
3432                         for (ifaca = idev->ac_list; ifaca;
3433                              ifaca = ifaca->aca_next, ip_idx++) {
3434                                 if (ip_idx < s_ip_idx)
3435                                         continue;
3436                                 err = inet6_fill_ifacaddr(skb, ifaca,
3437                                                           NETLINK_CB(cb->skb).pid,
3438                                                           cb->nlh->nlmsg_seq,
3439                                                           RTM_GETANYCAST,
3440                                                           NLM_F_MULTI);
3441                         }
3442                         break;
3443                 default:
3444                         break;
3445                 }
3446                 read_unlock_bh(&idev->lock);
3447                 in6_dev_put(idev);
3448
3449                 if (err <= 0)
3450                         break;
3451 cont:
3452                 idx++;
3453         }
3454         cb->args[0] = idx;
3455         cb->args[1] = ip_idx;
3456         return skb->len;
3457 }
3458
3459 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3460 {
3461         enum addr_type_t type = UNICAST_ADDR;
3462
3463         return inet6_dump_addr(skb, cb, type);
3464 }
3465
3466 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3467 {
3468         enum addr_type_t type = MULTICAST_ADDR;
3469
3470         return inet6_dump_addr(skb, cb, type);
3471 }
3472
3473
3474 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3475 {
3476         enum addr_type_t type = ANYCAST_ADDR;
3477
3478         return inet6_dump_addr(skb, cb, type);
3479 }
3480
3481 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3482                              void *arg)
3483 {
3484         struct net *net = sock_net(in_skb->sk);
3485         struct ifaddrmsg *ifm;
3486         struct nlattr *tb[IFA_MAX+1];
3487         struct in6_addr *addr = NULL;
3488         struct net_device *dev = NULL;
3489         struct inet6_ifaddr *ifa;
3490         struct sk_buff *skb;
3491         int err;
3492
3493         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3494         if (err < 0)
3495                 goto errout;
3496
3497         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3498         if (addr == NULL) {
3499                 err = -EINVAL;
3500                 goto errout;
3501         }
3502
3503         ifm = nlmsg_data(nlh);
3504         if (ifm->ifa_index)
3505                 dev = __dev_get_by_index(net, ifm->ifa_index);
3506
3507         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3508                 err = -EADDRNOTAVAIL;
3509                 goto errout;
3510         }
3511
3512         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3513                 err = -ENOBUFS;
3514                 goto errout_ifa;
3515         }
3516
3517         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3518                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3519         if (err < 0) {
3520                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3521                 WARN_ON(err == -EMSGSIZE);
3522                 kfree_skb(skb);
3523                 goto errout_ifa;
3524         }
3525         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3526 errout_ifa:
3527         in6_ifa_put(ifa);
3528 errout:
3529         return err;
3530 }
3531
3532 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3533 {
3534         struct sk_buff *skb;
3535         struct net *net = dev_net(ifa->idev->dev);
3536         int err = -ENOBUFS;
3537
3538         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3539         if (skb == NULL)
3540                 goto errout;
3541
3542         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3543         if (err < 0) {
3544                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3545                 WARN_ON(err == -EMSGSIZE);
3546                 kfree_skb(skb);
3547                 goto errout;
3548         }
3549         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3550 errout:
3551         if (err < 0)
3552                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3553 }
3554
3555 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3556                                 __s32 *array, int bytes)
3557 {
3558         BUG_ON(bytes < (DEVCONF_MAX * 4));
3559
3560         memset(array, 0, bytes);
3561         array[DEVCONF_FORWARDING] = cnf->forwarding;
3562         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3563         array[DEVCONF_MTU6] = cnf->mtu6;
3564         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3565         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3566         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3567         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3568         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3569         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3570         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3571         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3572 #ifdef CONFIG_IPV6_PRIVACY
3573         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3574         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3575         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3576         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3577         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3578 #endif
3579         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3580         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3581         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3582 #ifdef CONFIG_IPV6_ROUTER_PREF
3583         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3584         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3585 #ifdef CONFIG_IPV6_ROUTE_INFO
3586         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3587 #endif
3588 #endif
3589         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3590         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3591 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3592         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3593 #endif
3594 #ifdef CONFIG_IPV6_MROUTE
3595         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3596 #endif
3597 }
3598
3599 static inline size_t inet6_if_nlmsg_size(void)
3600 {
3601         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3602                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3603                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3604                + nla_total_size(4) /* IFLA_MTU */
3605                + nla_total_size(4) /* IFLA_LINK */
3606                + nla_total_size( /* IFLA_PROTINFO */
3607                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3608                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3609                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3610                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3611                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3612                  );
3613 }
3614
3615 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3616                                       int bytes)
3617 {
3618         int i;
3619         int pad = bytes - sizeof(u64) * items;
3620         BUG_ON(pad < 0);
3621
3622         /* Use put_unaligned() because stats may not be aligned for u64. */
3623         put_unaligned(items, &stats[0]);
3624         for (i = 1; i < items; i++)
3625                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3626
3627         memset(&stats[items], 0, pad);
3628 }
3629
3630 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3631                              int bytes)
3632 {
3633         switch(attrtype) {
3634         case IFLA_INET6_STATS:
3635                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3636                 break;
3637         case IFLA_INET6_ICMP6STATS:
3638                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3639                 break;
3640         }
3641 }
3642
3643 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3644                              u32 pid, u32 seq, int event, unsigned int flags)
3645 {
3646         struct net_device *dev = idev->dev;
3647         struct nlattr *nla;
3648         struct ifinfomsg *hdr;
3649         struct nlmsghdr *nlh;
3650         void *protoinfo;
3651         struct ifla_cacheinfo ci;
3652
3653         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3654         if (nlh == NULL)
3655                 return -EMSGSIZE;
3656
3657         hdr = nlmsg_data(nlh);
3658         hdr->ifi_family = AF_INET6;
3659         hdr->__ifi_pad = 0;
3660         hdr->ifi_type = dev->type;
3661         hdr->ifi_index = dev->ifindex;
3662         hdr->ifi_flags = dev_get_flags(dev);
3663         hdr->ifi_change = 0;
3664
3665         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3666
3667         if (dev->addr_len)
3668                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3669
3670         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3671         if (dev->ifindex != dev->iflink)
3672                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3673
3674         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3675         if (protoinfo == NULL)
3676                 goto nla_put_failure;
3677
3678         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3679
3680         ci.max_reasm_len = IPV6_MAXPLEN;
3681         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3682                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3683         ci.reachable_time = idev->nd_parms->reachable_time;
3684         ci.retrans_time = idev->nd_parms->retrans_time;
3685         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3686
3687         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3688         if (nla == NULL)
3689                 goto nla_put_failure;
3690         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3691
3692         /* XXX - MC not implemented */
3693
3694         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3695         if (nla == NULL)
3696                 goto nla_put_failure;
3697         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3698
3699         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3700         if (nla == NULL)
3701                 goto nla_put_failure;
3702         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3703
3704         nla_nest_end(skb, protoinfo);
3705         return nlmsg_end(skb, nlh);
3706
3707 nla_put_failure:
3708         nlmsg_cancel(skb, nlh);
3709         return -EMSGSIZE;
3710 }
3711
3712 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3713 {
3714         struct net *net = sock_net(skb->sk);
3715         int idx, err;
3716         int s_idx = cb->args[0];
3717         struct net_device *dev;
3718         struct inet6_dev *idev;
3719
3720         read_lock(&dev_base_lock);
3721         idx = 0;
3722         for_each_netdev(net, dev) {
3723                 if (idx < s_idx)
3724                         goto cont;
3725                 if ((idev = in6_dev_get(dev)) == NULL)
3726                         goto cont;
3727                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3728                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3729                 in6_dev_put(idev);
3730                 if (err <= 0)
3731                         break;
3732 cont:
3733                 idx++;
3734         }
3735         read_unlock(&dev_base_lock);
3736         cb->args[0] = idx;
3737
3738         return skb->len;
3739 }
3740
3741 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3742 {
3743         struct sk_buff *skb;
3744         struct net *net = dev_net(idev->dev);
3745         int err = -ENOBUFS;
3746
3747         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3748         if (skb == NULL)
3749                 goto errout;
3750
3751         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3752         if (err < 0) {
3753                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3754                 WARN_ON(err == -EMSGSIZE);
3755                 kfree_skb(skb);
3756                 goto errout;
3757         }
3758         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3759 errout:
3760         if (err < 0)
3761                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3762 }
3763
3764 static inline size_t inet6_prefix_nlmsg_size(void)
3765 {
3766         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3767                + nla_total_size(sizeof(struct in6_addr))
3768                + nla_total_size(sizeof(struct prefix_cacheinfo));
3769 }
3770
3771 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3772                              struct prefix_info *pinfo, u32 pid, u32 seq,
3773                              int event, unsigned int flags)
3774 {
3775         struct prefixmsg *pmsg;
3776         struct nlmsghdr *nlh;
3777         struct prefix_cacheinfo ci;
3778
3779         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3780         if (nlh == NULL)
3781                 return -EMSGSIZE;
3782
3783         pmsg = nlmsg_data(nlh);
3784         pmsg->prefix_family = AF_INET6;
3785         pmsg->prefix_pad1 = 0;
3786         pmsg->prefix_pad2 = 0;
3787         pmsg->prefix_ifindex = idev->dev->ifindex;
3788         pmsg->prefix_len = pinfo->prefix_len;
3789         pmsg->prefix_type = pinfo->type;
3790         pmsg->prefix_pad3 = 0;
3791         pmsg->prefix_flags = 0;
3792         if (pinfo->onlink)
3793                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3794         if (pinfo->autoconf)
3795                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3796
3797         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3798
3799         ci.preferred_time = ntohl(pinfo->prefered);
3800         ci.valid_time = ntohl(pinfo->valid);
3801         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3802
3803         return nlmsg_end(skb, nlh);
3804
3805 nla_put_failure:
3806         nlmsg_cancel(skb, nlh);
3807         return -EMSGSIZE;
3808 }
3809
3810 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3811                          struct prefix_info *pinfo)
3812 {
3813         struct sk_buff *skb;
3814         struct net *net = dev_net(idev->dev);
3815         int err = -ENOBUFS;
3816
3817         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3818         if (skb == NULL)
3819                 goto errout;
3820
3821         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3822         if (err < 0) {
3823                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3824                 WARN_ON(err == -EMSGSIZE);
3825                 kfree_skb(skb);
3826                 goto errout;
3827         }
3828         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3829 errout:
3830         if (err < 0)
3831                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3832 }
3833
3834 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3835 {
3836         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3837
3838         switch (event) {
3839         case RTM_NEWADDR:
3840                 /*
3841                  * If the address was optimistic
3842                  * we inserted the route at the start of
3843                  * our DAD process, so we don't need
3844                  * to do it again
3845                  */
3846                 if (!(ifp->rt->rt6i_node))
3847                         ip6_ins_rt(ifp->rt);
3848                 if (ifp->idev->cnf.forwarding)
3849                         addrconf_join_anycast(ifp);
3850                 break;
3851         case RTM_DELADDR:
3852                 if (ifp->idev->cnf.forwarding)
3853                         addrconf_leave_anycast(ifp);
3854                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3855                 dst_hold(&ifp->rt->u.dst);
3856                 if (ip6_del_rt(ifp->rt))
3857                         dst_free(&ifp->rt->u.dst);
3858                 break;
3859         }
3860 }
3861
3862 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3863 {
3864         rcu_read_lock_bh();
3865         if (likely(ifp->idev->dead == 0))
3866                 __ipv6_ifa_notify(event, ifp);
3867         rcu_read_unlock_bh();
3868 }
3869
3870 #ifdef CONFIG_SYSCTL
3871
3872 static
3873 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3874                            void __user *buffer, size_t *lenp, loff_t *ppos)
3875 {
3876         int *valp = ctl->data;
3877         int val = *valp;
3878         int ret;
3879
3880         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3881
3882         if (write)
3883                 addrconf_fixup_forwarding(ctl, valp, val);
3884         return ret;
3885 }
3886
3887 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3888                                             int __user *name, int nlen,
3889                                             void __user *oldval,
3890                                             size_t __user *oldlenp,
3891                                             void __user *newval, size_t newlen)
3892 {
3893         int *valp = table->data;
3894         int val = *valp;
3895         int new;
3896
3897         if (!newval || !newlen)
3898                 return 0;
3899         if (newlen != sizeof(int))
3900                 return -EINVAL;
3901         if (get_user(new, (int __user *)newval))
3902                 return -EFAULT;
3903         if (new == *valp)
3904                 return 0;
3905         if (oldval && oldlenp) {
3906                 size_t len;
3907                 if (get_user(len, oldlenp))
3908                         return -EFAULT;
3909                 if (len) {
3910                         if (len > table->maxlen)
3911                                 len = table->maxlen;
3912                         if (copy_to_user(oldval, valp, len))
3913                                 return -EFAULT;
3914                         if (put_user(len, oldlenp))
3915                                 return -EFAULT;
3916                 }
3917         }
3918
3919         *valp = new;
3920         addrconf_fixup_forwarding(table, valp, val);
3921         return 1;
3922 }
3923
3924 static struct addrconf_sysctl_table
3925 {
3926         struct ctl_table_header *sysctl_header;
3927         ctl_table addrconf_vars[DEVCONF_MAX+1];
3928         char *dev_name;
3929 } addrconf_sysctl __read_mostly = {
3930         .sysctl_header = NULL,
3931         .addrconf_vars = {
3932                 {
3933                         .ctl_name       =       NET_IPV6_FORWARDING,
3934                         .procname       =       "forwarding",
3935                         .data           =       &ipv6_devconf.forwarding,
3936                         .maxlen         =       sizeof(int),
3937                         .mode           =       0644,
3938                         .proc_handler   =       &addrconf_sysctl_forward,
3939                         .strategy       =       &addrconf_sysctl_forward_strategy,
3940                 },
3941                 {
3942                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3943                         .procname       =       "hop_limit",
3944                         .data           =       &ipv6_devconf.hop_limit,
3945                         .maxlen         =       sizeof(int),
3946                         .mode           =       0644,
3947                         .proc_handler   =       proc_dointvec,
3948                 },
3949                 {
3950                         .ctl_name       =       NET_IPV6_MTU,
3951                         .procname       =       "mtu",
3952                         .data           =       &ipv6_devconf.mtu6,
3953                         .maxlen         =       sizeof(int),
3954                         .mode           =       0644,
3955                         .proc_handler   =       &proc_dointvec,
3956                 },
3957                 {
3958                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3959                         .procname       =       "accept_ra",
3960                         .data           =       &ipv6_devconf.accept_ra,
3961                         .maxlen         =       sizeof(int),
3962                         .mode           =       0644,
3963                         .proc_handler   =       &proc_dointvec,
3964                 },
3965                 {
3966                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3967                         .procname       =       "accept_redirects",
3968                         .data           =       &ipv6_devconf.accept_redirects,
3969                         .maxlen         =       sizeof(int),
3970                         .mode           =       0644,
3971                         .proc_handler   =       &proc_dointvec,
3972                 },
3973                 {
3974                         .ctl_name       =       NET_IPV6_AUTOCONF,
3975                         .procname       =       "autoconf",
3976                         .data           =       &ipv6_devconf.autoconf,
3977                         .maxlen         =       sizeof(int),
3978                         .mode           =       0644,
3979                         .proc_handler   =       &proc_dointvec,
3980                 },
3981                 {
3982                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3983                         .procname       =       "dad_transmits",
3984                         .data           =       &ipv6_devconf.dad_transmits,
3985                         .maxlen         =       sizeof(int),
3986                         .mode           =       0644,
3987                         .proc_handler   =       &proc_dointvec,
3988                 },
3989                 {
3990                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3991                         .procname       =       "router_solicitations",
3992                         .data           =       &ipv6_devconf.rtr_solicits,
3993                         .maxlen         =       sizeof(int),
3994                         .mode           =       0644,
3995                         .proc_handler   =       &proc_dointvec,
3996                 },
3997                 {
3998                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3999                         .procname       =       "router_solicitation_interval",
4000                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4001                         .maxlen         =       sizeof(int),
4002                         .mode           =       0644,
4003                         .proc_handler   =       &proc_dointvec_jiffies,
4004                         .strategy       =       &sysctl_jiffies,
4005                 },
4006                 {
4007                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4008                         .procname       =       "router_solicitation_delay",
4009                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4010                         .maxlen         =       sizeof(int),
4011                         .mode           =       0644,
4012                         .proc_handler   =       &proc_dointvec_jiffies,
4013                         .strategy       =       &sysctl_jiffies,
4014                 },
4015                 {
4016                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4017                         .procname       =       "force_mld_version",
4018                         .data           =       &ipv6_devconf.force_mld_version,
4019                         .maxlen         =       sizeof(int),
4020                         .mode           =       0644,
4021                         .proc_handler   =       &proc_dointvec,
4022                 },
4023 #ifdef CONFIG_IPV6_PRIVACY
4024                 {
4025                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4026                         .procname       =       "use_tempaddr",
4027                         .data           =       &ipv6_devconf.use_tempaddr,
4028                         .maxlen         =       sizeof(int),
4029                         .mode           =       0644,
4030                         .proc_handler   =       &proc_dointvec,
4031                 },
4032                 {
4033                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4034                         .procname       =       "temp_valid_lft",
4035                         .data           =       &ipv6_devconf.temp_valid_lft,
4036                         .maxlen         =       sizeof(int),
4037                         .mode           =       0644,
4038                         .proc_handler   =       &proc_dointvec,
4039                 },
4040                 {
4041                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4042                         .procname       =       "temp_prefered_lft",
4043                         .data           =       &ipv6_devconf.temp_prefered_lft,
4044                         .maxlen         =       sizeof(int),
4045                         .mode           =       0644,
4046                         .proc_handler   =       &proc_dointvec,
4047                 },
4048                 {
4049                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4050                         .procname       =       "regen_max_retry",
4051                         .data           =       &ipv6_devconf.regen_max_retry,
4052                         .maxlen         =       sizeof(int),
4053                         .mode           =       0644,
4054                         .proc_handler   =       &proc_dointvec,
4055                 },
4056                 {
4057                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4058                         .procname       =       "max_desync_factor",
4059                         .data           =       &ipv6_devconf.max_desync_factor,
4060                         .maxlen         =       sizeof(int),
4061                         .mode           =       0644,
4062                         .proc_handler   =       &proc_dointvec,
4063                 },
4064 #endif
4065                 {
4066                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4067                         .procname       =       "max_addresses",
4068                         .data           =       &ipv6_devconf.max_addresses,
4069                         .maxlen         =       sizeof(int),
4070                         .mode           =       0644,
4071                         .proc_handler   =       &proc_dointvec,
4072                 },
4073                 {
4074                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4075                         .procname       =       "accept_ra_defrtr",
4076                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4077                         .maxlen         =       sizeof(int),
4078                         .mode           =       0644,
4079                         .proc_handler   =       &proc_dointvec,
4080                 },
4081                 {
4082                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4083                         .procname       =       "accept_ra_pinfo",
4084                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4085                         .maxlen         =       sizeof(int),
4086                         .mode           =       0644,
4087                         .proc_handler   =       &proc_dointvec,
4088                 },
4089 #ifdef CONFIG_IPV6_ROUTER_PREF
4090                 {
4091                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4092                         .procname       =       "accept_ra_rtr_pref",
4093                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4094                         .maxlen         =       sizeof(int),
4095                         .mode           =       0644,
4096                         .proc_handler   =       &proc_dointvec,
4097                 },
4098                 {
4099                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4100                         .procname       =       "router_probe_interval",
4101                         .data           =       &ipv6_devconf.rtr_probe_interval,
4102                         .maxlen         =       sizeof(int),
4103                         .mode           =       0644,
4104                         .proc_handler   =       &proc_dointvec_jiffies,
4105                         .strategy       =       &sysctl_jiffies,
4106                 },
4107 #ifdef CONFIG_IPV6_ROUTE_INFO
4108                 {
4109                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4110                         .procname       =       "accept_ra_rt_info_max_plen",
4111                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4112                         .maxlen         =       sizeof(int),
4113                         .mode           =       0644,
4114                         .proc_handler   =       &proc_dointvec,
4115                 },
4116 #endif
4117 #endif
4118                 {
4119                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4120                         .procname       =       "proxy_ndp",
4121                         .data           =       &ipv6_devconf.proxy_ndp,
4122                         .maxlen         =       sizeof(int),
4123                         .mode           =       0644,
4124                         .proc_handler   =       &proc_dointvec,
4125                 },
4126                 {
4127                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4128                         .procname       =       "accept_source_route",
4129                         .data           =       &ipv6_devconf.accept_source_route,
4130                         .maxlen         =       sizeof(int),
4131                         .mode           =       0644,
4132                         .proc_handler   =       &proc_dointvec,
4133                 },
4134 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4135                 {
4136                         .ctl_name       =       CTL_UNNUMBERED,
4137                         .procname       =       "optimistic_dad",
4138                         .data           =       &ipv6_devconf.optimistic_dad,
4139                         .maxlen         =       sizeof(int),
4140                         .mode           =       0644,
4141                         .proc_handler   =       &proc_dointvec,
4142
4143                 },
4144 #endif
4145 #ifdef CONFIG_IPV6_MROUTE
4146                 {
4147                         .ctl_name       =       CTL_UNNUMBERED,
4148                         .procname       =       "mc_forwarding",
4149                         .data           =       &ipv6_devconf.mc_forwarding,
4150                         .maxlen         =       sizeof(int),
4151                         .mode           =       0644,
4152                         .proc_handler   =       &proc_dointvec,
4153                 },
4154 #endif
4155                 {
4156                         .ctl_name       =       0,      /* sentinel */
4157                 }
4158         },
4159 };
4160
4161 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4162                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4163 {
4164         int i;
4165         struct addrconf_sysctl_table *t;
4166
4167 #define ADDRCONF_CTL_PATH_DEV   3
4168
4169         struct ctl_path addrconf_ctl_path[] = {
4170                 { .procname = "net", .ctl_name = CTL_NET, },
4171                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4172                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4173                 { /* to be set */ },
4174                 { },
4175         };
4176
4177
4178         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4179         if (t == NULL)
4180                 goto out;
4181
4182         for (i=0; t->addrconf_vars[i].data; i++) {
4183                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4184                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4185                 t->addrconf_vars[i].extra2 = net;
4186         }
4187
4188         /*
4189          * Make a copy of dev_name, because '.procname' is regarded as const
4190          * by sysctl and we wouldn't want anyone to change it under our feet
4191          * (see SIOCSIFNAME).
4192          */
4193         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4194         if (!t->dev_name)
4195                 goto free;
4196
4197         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4198         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4199
4200         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4201                         t->addrconf_vars);
4202         if (t->sysctl_header == NULL)
4203                 goto free_procname;
4204
4205         p->sysctl = t;
4206         return 0;
4207
4208 free_procname:
4209         kfree(t->dev_name);
4210 free:
4211         kfree(t);
4212 out:
4213         return -ENOBUFS;
4214 }
4215
4216 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4217 {
4218         struct addrconf_sysctl_table *t;
4219
4220         if (p->sysctl == NULL)
4221                 return;
4222
4223         t = p->sysctl;
4224         p->sysctl = NULL;
4225         unregister_sysctl_table(t->sysctl_header);
4226         kfree(t->dev_name);
4227         kfree(t);
4228 }
4229
4230 static void addrconf_sysctl_register(struct inet6_dev *idev)
4231 {
4232         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4233                               NET_IPV6_NEIGH, "ipv6",
4234                               &ndisc_ifinfo_sysctl_change,
4235                               NULL);
4236         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4237                         idev->dev->ifindex, idev, &idev->cnf);
4238 }
4239
4240 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4241 {
4242         __addrconf_sysctl_unregister(&idev->cnf);
4243         neigh_sysctl_unregister(idev->nd_parms);
4244 }
4245
4246
4247 #endif
4248
4249 static int addrconf_init_net(struct net *net)
4250 {
4251         int err;
4252         struct ipv6_devconf *all, *dflt;
4253
4254         err = -ENOMEM;
4255         all = &ipv6_devconf;
4256         dflt = &ipv6_devconf_dflt;
4257
4258         if (net != &init_net) {
4259                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4260                 if (all == NULL)
4261                         goto err_alloc_all;
4262
4263                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4264                 if (dflt == NULL)
4265                         goto err_alloc_dflt;
4266         }
4267
4268         net->ipv6.devconf_all = all;
4269         net->ipv6.devconf_dflt = dflt;
4270
4271 #ifdef CONFIG_SYSCTL
4272         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4273                         NULL, all);
4274         if (err < 0)
4275                 goto err_reg_all;
4276
4277         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4278                         NULL, dflt);
4279         if (err < 0)
4280                 goto err_reg_dflt;
4281 #endif
4282         return 0;
4283
4284 #ifdef CONFIG_SYSCTL
4285 err_reg_dflt:
4286         __addrconf_sysctl_unregister(all);
4287 err_reg_all:
4288         kfree(dflt);
4289 #endif
4290 err_alloc_dflt:
4291         kfree(all);
4292 err_alloc_all:
4293         return err;
4294 }
4295
4296 static void addrconf_exit_net(struct net *net)
4297 {
4298 #ifdef CONFIG_SYSCTL
4299         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4300         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4301 #endif
4302         if (net != &init_net) {
4303                 kfree(net->ipv6.devconf_dflt);
4304                 kfree(net->ipv6.devconf_all);
4305         }
4306 }
4307
4308 static struct pernet_operations addrconf_ops = {
4309         .init = addrconf_init_net,
4310         .exit = addrconf_exit_net,
4311 };
4312
4313 /*
4314  *      Device notifier
4315  */
4316
4317 int register_inet6addr_notifier(struct notifier_block *nb)
4318 {
4319         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4320 }
4321
4322 EXPORT_SYMBOL(register_inet6addr_notifier);
4323
4324 int unregister_inet6addr_notifier(struct notifier_block *nb)
4325 {
4326         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4327 }
4328
4329 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4330
4331
4332 static int addrconf_net_init(struct net *net)
4333 {
4334         return 0;
4335 }
4336
4337 static void addrconf_net_exit(struct net *net)
4338 {
4339         struct net_device *dev;
4340
4341         rtnl_lock();
4342         /* clean dev list */
4343         for_each_netdev(net, dev) {
4344                 if (__in6_dev_get(dev) == NULL)
4345                         continue;
4346                 addrconf_ifdown(dev, 1);
4347         }
4348         addrconf_ifdown(net->loopback_dev, 2);
4349         rtnl_unlock();
4350 }
4351
4352 static struct pernet_operations addrconf_net_ops = {
4353         .init = addrconf_net_init,
4354         .exit = addrconf_net_exit,
4355 };
4356
4357 /*
4358  *      Init / cleanup code
4359  */
4360
4361 int __init addrconf_init(void)
4362 {
4363         int err;
4364
4365         if ((err = ipv6_addr_label_init()) < 0) {
4366                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4367                         err);
4368                 return err;
4369         }
4370
4371         register_pernet_subsys(&addrconf_ops);
4372
4373         /* The addrconf netdev notifier requires that loopback_dev
4374          * has it's ipv6 private information allocated and setup
4375          * before it can bring up and give link-local addresses
4376          * to other devices which are up.
4377          *
4378          * Unfortunately, loopback_dev is not necessarily the first
4379          * entry in the global dev_base list of net devices.  In fact,
4380          * it is likely to be the very last entry on that list.
4381          * So this causes the notifier registry below to try and
4382          * give link-local addresses to all devices besides loopback_dev
4383          * first, then loopback_dev, which cases all the non-loopback_dev
4384          * devices to fail to get a link-local address.
4385          *
4386          * So, as a temporary fix, allocate the ipv6 structure for
4387          * loopback_dev first by hand.
4388          * Longer term, all of the dependencies ipv6 has upon the loopback
4389          * device and it being up should be removed.
4390          */
4391         rtnl_lock();
4392         if (!ipv6_add_dev(init_net.loopback_dev))
4393                 err = -ENOMEM;
4394         rtnl_unlock();
4395         if (err)
4396                 goto errlo;
4397
4398         err = register_pernet_device(&addrconf_net_ops);
4399         if (err)
4400                 return err;
4401
4402         register_netdevice_notifier(&ipv6_dev_notf);
4403
4404         addrconf_verify(0);
4405
4406         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4407         if (err < 0)
4408                 goto errout;
4409
4410         /* Only the first call to __rtnl_register can fail */
4411         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4412         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4413         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4414         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4415         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4416
4417         ipv6_addr_label_rtnl_register();
4418
4419         return 0;
4420 errout:
4421         unregister_netdevice_notifier(&ipv6_dev_notf);
4422 errlo:
4423         unregister_pernet_subsys(&addrconf_ops);
4424
4425         return err;
4426 }
4427
4428 void addrconf_cleanup(void)
4429 {
4430         struct inet6_ifaddr *ifa;
4431         int i;
4432
4433         unregister_netdevice_notifier(&ipv6_dev_notf);
4434         unregister_pernet_device(&addrconf_net_ops);
4435
4436         unregister_pernet_subsys(&addrconf_ops);
4437
4438         rtnl_lock();
4439
4440         /*
4441          *      Check hash table.
4442          */
4443         write_lock_bh(&addrconf_hash_lock);
4444         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4445                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4446                         struct inet6_ifaddr *bifa;
4447
4448                         bifa = ifa;
4449                         ifa = ifa->lst_next;
4450                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4451                         /* Do not free it; something is wrong.
4452                            Now we can investigate it with debugger.
4453                          */
4454                 }
4455         }
4456         write_unlock_bh(&addrconf_hash_lock);
4457
4458         del_timer(&addr_chk_timer);
4459         rtnl_unlock();
4460
4461         unregister_pernet_subsys(&addrconf_net_ops);
4462 }