[IPV6]: Define constants for link-local multicast addresses.
[safe/jmp/linux-2.6] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16
17 /*
18  *      Changes:
19  *
20  *      Janos Farkas                    :       delete timer on ifdown
21  *      <chexum@bankinf.banki.hu>
22  *      Andi Kleen                      :       kill double kfree on module
23  *                                              unload.
24  *      Maciej W. Rozycki               :       FDDI support
25  *      sekiya@USAGI                    :       Don't send too many RS
26  *                                              packets.
27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
28  *                                              packets.
29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
30  *                                              address validation timer.
31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
32  *                                              support.
33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
34  *                                              address on a same interface.
35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
37  *                                              seq_file.
38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
39  *                                              selection; consider scope,
40  *                                              status etc.
41  */
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <asm/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99
100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
108 {
109 }
110
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 {
113 }
114 #endif
115
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
120
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
123
124 static int ipv6_count_addresses(struct inet6_dev *idev);
125
126 /*
127  *      Configured unicast address hash table
128  */
129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
131
132 static void addrconf_verify(unsigned long);
133
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
136
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139
140 static int addrconf_ifdown(struct net_device *dev, int how);
141
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
149
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151                                 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153                               struct net_device *dev);
154
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156
157 struct ipv6_devconf ipv6_devconf __read_mostly = {
158         .forwarding             = 0,
159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
160         .mtu6                   = IPV6_MIN_MTU,
161         .accept_ra              = 1,
162         .accept_redirects       = 1,
163         .autoconf               = 1,
164         .force_mld_version      = 0,
165         .dad_transmits          = 1,
166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170         .use_tempaddr           = 0,
171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
173         .regen_max_retry        = REGEN_MAX_RETRY,
174         .max_desync_factor      = MAX_DESYNC_FACTOR,
175 #endif
176         .max_addresses          = IPV6_MAX_ADDRESSES,
177         .accept_ra_defrtr       = 1,
178         .accept_ra_pinfo        = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180         .accept_ra_rtr_pref     = 1,
181         .rtr_probe_interval     = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183         .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186         .proxy_ndp              = 0,
187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
188 };
189
190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
191         .forwarding             = 0,
192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
193         .mtu6                   = IPV6_MIN_MTU,
194         .accept_ra              = 1,
195         .accept_redirects       = 1,
196         .autoconf               = 1,
197         .dad_transmits          = 1,
198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
201 #ifdef CONFIG_IPV6_PRIVACY
202         .use_tempaddr           = 0,
203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
205         .regen_max_retry        = REGEN_MAX_RETRY,
206         .max_desync_factor      = MAX_DESYNC_FACTOR,
207 #endif
208         .max_addresses          = IPV6_MAX_ADDRESSES,
209         .accept_ra_defrtr       = 1,
210         .accept_ra_pinfo        = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212         .accept_ra_rtr_pref     = 1,
213         .rtr_probe_interval     = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215         .accept_ra_rt_info_max_plen = 0,
216 #endif
217 #endif
218         .proxy_ndp              = 0,
219         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
220 };
221
222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
225 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
226 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
227
228 /* Check if a valid qdisc is available */
229 static inline int addrconf_qdisc_ok(struct net_device *dev)
230 {
231         return (dev->qdisc != &noop_qdisc);
232 }
233
234 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
235 {
236         if (del_timer(&ifp->timer))
237                 __in6_ifa_put(ifp);
238 }
239
240 enum addrconf_timer_t
241 {
242         AC_NONE,
243         AC_DAD,
244         AC_RS,
245 };
246
247 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
248                                enum addrconf_timer_t what,
249                                unsigned long when)
250 {
251         if (!del_timer(&ifp->timer))
252                 in6_ifa_hold(ifp);
253
254         switch (what) {
255         case AC_DAD:
256                 ifp->timer.function = addrconf_dad_timer;
257                 break;
258         case AC_RS:
259                 ifp->timer.function = addrconf_rs_timer;
260                 break;
261         default:;
262         }
263         ifp->timer.expires = jiffies + when;
264         add_timer(&ifp->timer);
265 }
266
267 static int snmp6_alloc_dev(struct inet6_dev *idev)
268 {
269         if (snmp_mib_init((void **)idev->stats.ipv6,
270                           sizeof(struct ipstats_mib)) < 0)
271                 goto err_ip;
272         if (snmp_mib_init((void **)idev->stats.icmpv6,
273                           sizeof(struct icmpv6_mib)) < 0)
274                 goto err_icmp;
275         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
276                           sizeof(struct icmpv6msg_mib)) < 0)
277                 goto err_icmpmsg;
278
279         return 0;
280
281 err_icmpmsg:
282         snmp_mib_free((void **)idev->stats.icmpv6);
283 err_icmp:
284         snmp_mib_free((void **)idev->stats.ipv6);
285 err_ip:
286         return -ENOMEM;
287 }
288
289 static void snmp6_free_dev(struct inet6_dev *idev)
290 {
291         snmp_mib_free((void **)idev->stats.icmpv6msg);
292         snmp_mib_free((void **)idev->stats.icmpv6);
293         snmp_mib_free((void **)idev->stats.ipv6);
294 }
295
296 /* Nobody refers to this device, we may destroy it. */
297
298 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
299 {
300         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
301         kfree(idev);
302 }
303
304 void in6_dev_finish_destroy(struct inet6_dev *idev)
305 {
306         struct net_device *dev = idev->dev;
307         BUG_TRAP(idev->addr_list==NULL);
308         BUG_TRAP(idev->mc_list==NULL);
309 #ifdef NET_REFCNT_DEBUG
310         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
311 #endif
312         dev_put(dev);
313         if (!idev->dead) {
314                 printk("Freeing alive inet6 device %p\n", idev);
315                 return;
316         }
317         snmp6_free_dev(idev);
318         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
319 }
320
321 EXPORT_SYMBOL(in6_dev_finish_destroy);
322
323 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
324 {
325         struct inet6_dev *ndev;
326
327         ASSERT_RTNL();
328
329         if (dev->mtu < IPV6_MIN_MTU)
330                 return NULL;
331
332         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
333
334         if (ndev == NULL)
335                 return NULL;
336
337         rwlock_init(&ndev->lock);
338         ndev->dev = dev;
339         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
340         ndev->cnf.mtu6 = dev->mtu;
341         ndev->cnf.sysctl = NULL;
342         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
343         if (ndev->nd_parms == NULL) {
344                 kfree(ndev);
345                 return NULL;
346         }
347         /* We refer to the device */
348         dev_hold(dev);
349
350         if (snmp6_alloc_dev(ndev) < 0) {
351                 ADBG((KERN_WARNING
352                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
353                         __func__, dev->name));
354                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
355                 ndev->dead = 1;
356                 in6_dev_finish_destroy(ndev);
357                 return NULL;
358         }
359
360         if (snmp6_register_dev(ndev) < 0) {
361                 ADBG((KERN_WARNING
362                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
363                         __func__, dev->name));
364                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
365                 ndev->dead = 1;
366                 in6_dev_finish_destroy(ndev);
367                 return NULL;
368         }
369
370         /* One reference from device.  We must do this before
371          * we invoke __ipv6_regen_rndid().
372          */
373         in6_dev_hold(ndev);
374
375 #ifdef CONFIG_IPV6_PRIVACY
376         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
377         if ((dev->flags&IFF_LOOPBACK) ||
378             dev->type == ARPHRD_TUNNEL ||
379 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
380             dev->type == ARPHRD_SIT ||
381 #endif
382             dev->type == ARPHRD_NONE) {
383                 printk(KERN_INFO
384                        "%s: Disabled Privacy Extensions\n",
385                        dev->name);
386                 ndev->cnf.use_tempaddr = -1;
387
388                 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
389                         printk(KERN_INFO
390                                "%s: Disabled Multicast RS\n",
391                                dev->name);
392                         ndev->cnf.rtr_solicits = 0;
393                 }
394         } else {
395                 in6_dev_hold(ndev);
396                 ipv6_regen_rndid((unsigned long) ndev);
397         }
398 #endif
399
400         if (netif_running(dev) && addrconf_qdisc_ok(dev))
401                 ndev->if_flags |= IF_READY;
402
403         ipv6_mc_init_dev(ndev);
404         ndev->tstamp = jiffies;
405         addrconf_sysctl_register(ndev);
406         /* protected by rtnl_lock */
407         rcu_assign_pointer(dev->ip6_ptr, ndev);
408
409         /* Join all-node multicast group */
410         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
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
437         if (!idev)
438                 return;
439         dev = idev->dev;
440         if (dev && (dev->flags & IFF_MULTICAST)) {
441                 if (idev->cnf.forwarding)
442                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
443                 else
444                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
445         }
446         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
447                 if (ifa->flags&IFA_F_TENTATIVE)
448                         continue;
449                 if (idev->cnf.forwarding)
450                         addrconf_join_anycast(ifa);
451                 else
452                         addrconf_leave_anycast(ifa);
453         }
454 }
455
456
457 static void addrconf_forward_change(struct net *net, __s32 newf)
458 {
459         struct net_device *dev;
460         struct inet6_dev *idev;
461
462         read_lock(&dev_base_lock);
463         for_each_netdev(net, dev) {
464                 rcu_read_lock();
465                 idev = __in6_dev_get(dev);
466                 if (idev) {
467                         int changed = (!idev->cnf.forwarding) ^ (!newf);
468                         idev->cnf.forwarding = newf;
469                         if (changed)
470                                 dev_forward_change(idev);
471                 }
472                 rcu_read_unlock();
473         }
474         read_unlock(&dev_base_lock);
475 }
476
477 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
478 {
479         struct net *net;
480
481         net = (struct net *)table->extra2;
482         if (p == &net->ipv6.devconf_dflt->forwarding)
483                 return;
484
485         if (p == &net->ipv6.devconf_all->forwarding) {
486                 __s32 newf = net->ipv6.devconf_all->forwarding;
487                 net->ipv6.devconf_dflt->forwarding = newf;
488                 addrconf_forward_change(net, newf);
489         } else if ((!*p) ^ (!old))
490                 dev_forward_change((struct inet6_dev *)table->extra1);
491
492         if (*p)
493                 rt6_purge_dflt_routers(net);
494 }
495 #endif
496
497 /* Nobody refers to this ifaddr, destroy it */
498
499 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
500 {
501         BUG_TRAP(ifp->if_next==NULL);
502         BUG_TRAP(ifp->lst_next==NULL);
503 #ifdef NET_REFCNT_DEBUG
504         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
505 #endif
506
507         in6_dev_put(ifp->idev);
508
509         if (del_timer(&ifp->timer))
510                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
511
512         if (!ifp->dead) {
513                 printk("Freeing alive inet6 address %p\n", ifp);
514                 return;
515         }
516         dst_release(&ifp->rt->u.dst);
517
518         kfree(ifp);
519 }
520
521 static void
522 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
523 {
524         struct inet6_ifaddr *ifa, **ifap;
525         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
526
527         /*
528          * Each device address list is sorted in order of scope -
529          * global before linklocal.
530          */
531         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
532              ifap = &ifa->if_next) {
533                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
534                         break;
535         }
536
537         ifp->if_next = *ifap;
538         *ifap = ifp;
539 }
540
541 /*
542  *      Hash function taken from net_alias.c
543  */
544 static u8 ipv6_addr_hash(const struct in6_addr *addr)
545 {
546         __u32 word;
547
548         /*
549          * We perform the hash function over the last 64 bits of the address
550          * This will include the IEEE address token on links that support it.
551          */
552
553         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
554         word ^= (word >> 16);
555         word ^= (word >> 8);
556
557         return ((word ^ (word >> 4)) & 0x0f);
558 }
559
560 /* On success it returns ifp with increased reference count */
561
562 static struct inet6_ifaddr *
563 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
564               int scope, u32 flags)
565 {
566         struct inet6_ifaddr *ifa = NULL;
567         struct rt6_info *rt;
568         int hash;
569         int err = 0;
570
571         rcu_read_lock_bh();
572         if (idev->dead) {
573                 err = -ENODEV;                  /*XXX*/
574                 goto out2;
575         }
576
577         write_lock(&addrconf_hash_lock);
578
579         /* Ignore adding duplicate addresses on an interface */
580         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
581                 ADBG(("ipv6_add_addr: already assigned\n"));
582                 err = -EEXIST;
583                 goto out;
584         }
585
586         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
587
588         if (ifa == NULL) {
589                 ADBG(("ipv6_add_addr: malloc failed\n"));
590                 err = -ENOBUFS;
591                 goto out;
592         }
593
594         rt = addrconf_dst_alloc(idev, addr, 0);
595         if (IS_ERR(rt)) {
596                 err = PTR_ERR(rt);
597                 goto out;
598         }
599
600         ipv6_addr_copy(&ifa->addr, addr);
601
602         spin_lock_init(&ifa->lock);
603         init_timer(&ifa->timer);
604         ifa->timer.data = (unsigned long) ifa;
605         ifa->scope = scope;
606         ifa->prefix_len = pfxlen;
607         ifa->flags = flags | IFA_F_TENTATIVE;
608         ifa->cstamp = ifa->tstamp = jiffies;
609
610         ifa->rt = rt;
611
612         /*
613          * part one of RFC 4429, section 3.3
614          * We should not configure an address as
615          * optimistic if we do not yet know the link
616          * layer address of our nexhop router
617          */
618
619         if (rt->rt6i_nexthop == NULL)
620                 ifa->flags &= ~IFA_F_OPTIMISTIC;
621
622         ifa->idev = idev;
623         in6_dev_hold(idev);
624         /* For caller */
625         in6_ifa_hold(ifa);
626
627         /* Add to big hash table */
628         hash = ipv6_addr_hash(addr);
629
630         ifa->lst_next = inet6_addr_lst[hash];
631         inet6_addr_lst[hash] = ifa;
632         in6_ifa_hold(ifa);
633         write_unlock(&addrconf_hash_lock);
634
635         write_lock(&idev->lock);
636         /* Add to inet6_dev unicast addr list. */
637         ipv6_link_dev_addr(idev, ifa);
638
639 #ifdef CONFIG_IPV6_PRIVACY
640         if (ifa->flags&IFA_F_TEMPORARY) {
641                 ifa->tmp_next = idev->tempaddr_list;
642                 idev->tempaddr_list = ifa;
643                 in6_ifa_hold(ifa);
644         }
645 #endif
646
647         in6_ifa_hold(ifa);
648         write_unlock(&idev->lock);
649 out2:
650         rcu_read_unlock_bh();
651
652         if (likely(err == 0))
653                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
654         else {
655                 kfree(ifa);
656                 ifa = ERR_PTR(err);
657         }
658
659         return ifa;
660 out:
661         write_unlock(&addrconf_hash_lock);
662         goto out2;
663 }
664
665 /* This function wants to get referenced ifp and releases it before return */
666
667 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
668 {
669         struct inet6_ifaddr *ifa, **ifap;
670         struct inet6_dev *idev = ifp->idev;
671         int hash;
672         int deleted = 0, onlink = 0;
673         unsigned long expires = jiffies;
674
675         hash = ipv6_addr_hash(&ifp->addr);
676
677         ifp->dead = 1;
678
679         write_lock_bh(&addrconf_hash_lock);
680         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
681              ifap = &ifa->lst_next) {
682                 if (ifa == ifp) {
683                         *ifap = ifa->lst_next;
684                         __in6_ifa_put(ifp);
685                         ifa->lst_next = NULL;
686                         break;
687                 }
688         }
689         write_unlock_bh(&addrconf_hash_lock);
690
691         write_lock_bh(&idev->lock);
692 #ifdef CONFIG_IPV6_PRIVACY
693         if (ifp->flags&IFA_F_TEMPORARY) {
694                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
695                      ifap = &ifa->tmp_next) {
696                         if (ifa == ifp) {
697                                 *ifap = ifa->tmp_next;
698                                 if (ifp->ifpub) {
699                                         in6_ifa_put(ifp->ifpub);
700                                         ifp->ifpub = NULL;
701                                 }
702                                 __in6_ifa_put(ifp);
703                                 ifa->tmp_next = NULL;
704                                 break;
705                         }
706                 }
707         }
708 #endif
709
710         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
711                 if (ifa == ifp) {
712                         *ifap = ifa->if_next;
713                         __in6_ifa_put(ifp);
714                         ifa->if_next = NULL;
715                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
716                                 break;
717                         deleted = 1;
718                         continue;
719                 } else if (ifp->flags & IFA_F_PERMANENT) {
720                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
721                                               ifp->prefix_len)) {
722                                 if (ifa->flags & IFA_F_PERMANENT) {
723                                         onlink = 1;
724                                         if (deleted)
725                                                 break;
726                                 } else {
727                                         unsigned long lifetime;
728
729                                         if (!onlink)
730                                                 onlink = -1;
731
732                                         spin_lock(&ifa->lock);
733                                         lifetime = min_t(unsigned long,
734                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
735                                         if (time_before(expires,
736                                                         ifa->tstamp + lifetime * HZ))
737                                                 expires = ifa->tstamp + lifetime * HZ;
738                                         spin_unlock(&ifa->lock);
739                                 }
740                         }
741                 }
742                 ifap = &ifa->if_next;
743         }
744         write_unlock_bh(&idev->lock);
745
746         ipv6_ifa_notify(RTM_DELADDR, ifp);
747
748         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
749
750         addrconf_del_timer(ifp);
751
752         /*
753          * Purge or update corresponding prefix
754          *
755          * 1) we don't purge prefix here if address was not permanent.
756          *    prefix is managed by its own lifetime.
757          * 2) if there're no addresses, delete prefix.
758          * 3) if there're still other permanent address(es),
759          *    corresponding prefix is still permanent.
760          * 4) otherwise, update prefix lifetime to the
761          *    longest valid lifetime among the corresponding
762          *    addresses on the device.
763          *    Note: subsequent RA will update lifetime.
764          *
765          * --yoshfuji
766          */
767         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
768                 struct in6_addr prefix;
769                 struct rt6_info *rt;
770                 struct net *net = dev_net(ifp->idev->dev);
771                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
772                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
773
774                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
775                         if (onlink == 0) {
776                                 ip6_del_rt(rt);
777                                 rt = NULL;
778                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
779                                 rt->rt6i_expires = expires;
780                                 rt->rt6i_flags |= RTF_EXPIRES;
781                         }
782                 }
783                 dst_release(&rt->u.dst);
784         }
785
786         in6_ifa_put(ifp);
787 }
788
789 #ifdef CONFIG_IPV6_PRIVACY
790 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
791 {
792         struct inet6_dev *idev = ifp->idev;
793         struct in6_addr addr, *tmpaddr;
794         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
795         unsigned long regen_advance;
796         int tmp_plen;
797         int ret = 0;
798         int max_addresses;
799         u32 addr_flags;
800
801         write_lock(&idev->lock);
802         if (ift) {
803                 spin_lock_bh(&ift->lock);
804                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
805                 spin_unlock_bh(&ift->lock);
806                 tmpaddr = &addr;
807         } else {
808                 tmpaddr = NULL;
809         }
810 retry:
811         in6_dev_hold(idev);
812         if (idev->cnf.use_tempaddr <= 0) {
813                 write_unlock(&idev->lock);
814                 printk(KERN_INFO
815                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
816                 in6_dev_put(idev);
817                 ret = -1;
818                 goto out;
819         }
820         spin_lock_bh(&ifp->lock);
821         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
822                 idev->cnf.use_tempaddr = -1;    /*XXX*/
823                 spin_unlock_bh(&ifp->lock);
824                 write_unlock(&idev->lock);
825                 printk(KERN_WARNING
826                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
827                 in6_dev_put(idev);
828                 ret = -1;
829                 goto out;
830         }
831         in6_ifa_hold(ifp);
832         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
833         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
834                 spin_unlock_bh(&ifp->lock);
835                 write_unlock(&idev->lock);
836                 printk(KERN_WARNING
837                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
838                 in6_ifa_put(ifp);
839                 in6_dev_put(idev);
840                 ret = -1;
841                 goto out;
842         }
843         memcpy(&addr.s6_addr[8], idev->rndid, 8);
844         tmp_valid_lft = min_t(__u32,
845                               ifp->valid_lft,
846                               idev->cnf.temp_valid_lft);
847         tmp_prefered_lft = min_t(__u32,
848                                  ifp->prefered_lft,
849                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
850         tmp_plen = ifp->prefix_len;
851         max_addresses = idev->cnf.max_addresses;
852         tmp_cstamp = ifp->cstamp;
853         tmp_tstamp = ifp->tstamp;
854         spin_unlock_bh(&ifp->lock);
855
856         regen_advance = idev->cnf.regen_max_retry *
857                         idev->cnf.dad_transmits *
858                         idev->nd_parms->retrans_time / HZ;
859         write_unlock(&idev->lock);
860
861         /* A temporary address is created only if this calculated Preferred
862          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
863          * an implementation must not create a temporary address with a zero
864          * Preferred Lifetime.
865          */
866         if (tmp_prefered_lft <= regen_advance) {
867                 in6_ifa_put(ifp);
868                 in6_dev_put(idev);
869                 ret = -1;
870                 goto out;
871         }
872
873         addr_flags = IFA_F_TEMPORARY;
874         /* set in addrconf_prefix_rcv() */
875         if (ifp->flags & IFA_F_OPTIMISTIC)
876                 addr_flags |= IFA_F_OPTIMISTIC;
877
878         ift = !max_addresses ||
879               ipv6_count_addresses(idev) < max_addresses ?
880                 ipv6_add_addr(idev, &addr, tmp_plen,
881                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
882                               addr_flags) : NULL;
883         if (!ift || IS_ERR(ift)) {
884                 in6_ifa_put(ifp);
885                 in6_dev_put(idev);
886                 printk(KERN_INFO
887                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
888                 tmpaddr = &addr;
889                 write_lock(&idev->lock);
890                 goto retry;
891         }
892
893         spin_lock_bh(&ift->lock);
894         ift->ifpub = ifp;
895         ift->valid_lft = tmp_valid_lft;
896         ift->prefered_lft = tmp_prefered_lft;
897         ift->cstamp = tmp_cstamp;
898         ift->tstamp = tmp_tstamp;
899         spin_unlock_bh(&ift->lock);
900
901         addrconf_dad_start(ift, 0);
902         in6_ifa_put(ift);
903         in6_dev_put(idev);
904 out:
905         return ret;
906 }
907 #endif
908
909 /*
910  *      Choose an appropriate source address (RFC3484)
911  */
912 enum {
913         IPV6_SADDR_RULE_INIT = 0,
914         IPV6_SADDR_RULE_LOCAL,
915         IPV6_SADDR_RULE_SCOPE,
916         IPV6_SADDR_RULE_PREFERRED,
917 #ifdef CONFIG_IPV6_MIP6
918         IPV6_SADDR_RULE_HOA,
919 #endif
920         IPV6_SADDR_RULE_OIF,
921         IPV6_SADDR_RULE_LABEL,
922 #ifdef CONFIG_IPV6_PRIVACY
923         IPV6_SADDR_RULE_PRIVACY,
924 #endif
925         IPV6_SADDR_RULE_ORCHID,
926         IPV6_SADDR_RULE_PREFIX,
927         IPV6_SADDR_RULE_MAX
928 };
929
930 struct ipv6_saddr_score {
931         int                     rule;
932         int                     addr_type;
933         struct inet6_ifaddr     *ifa;
934         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
935         int                     scopedist;
936         int                     matchlen;
937 };
938
939 struct ipv6_saddr_dst {
940         const struct in6_addr *addr;
941         int ifindex;
942         int scope;
943         int label;
944         unsigned int prefs;
945 };
946
947 static inline int ipv6_saddr_preferred(int type)
948 {
949         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
950                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
951                 return 1;
952         return 0;
953 }
954
955 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
956                                struct ipv6_saddr_dst *dst,
957                                int i)
958 {
959         int ret;
960
961         if (i <= score->rule) {
962                 switch (i) {
963                 case IPV6_SADDR_RULE_SCOPE:
964                         ret = score->scopedist;
965                         break;
966                 case IPV6_SADDR_RULE_PREFIX:
967                         ret = score->matchlen;
968                         break;
969                 default:
970                         ret = !!test_bit(i, score->scorebits);
971                 }
972                 goto out;
973         }
974
975         switch (i) {
976         case IPV6_SADDR_RULE_INIT:
977                 /* Rule 0: remember if hiscore is not ready yet */
978                 ret = !!score->ifa;
979                 break;
980         case IPV6_SADDR_RULE_LOCAL:
981                 /* Rule 1: Prefer same address */
982                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
983                 break;
984         case IPV6_SADDR_RULE_SCOPE:
985                 /* Rule 2: Prefer appropriate scope
986                  *
987                  *      ret
988                  *       ^
989                  *    -1 |  d 15
990                  *    ---+--+-+---> scope
991                  *       |
992                  *       |             d is scope of the destination.
993                  *  B-d  |  \
994                  *       |   \      <- smaller scope is better if
995                  *  B-15 |    \        if scope is enough for destinaion.
996                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
997                  * d-C-1 | /
998                  *       |/         <- greater is better
999                  *   -C  /             if scope is not enough for destination.
1000                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1001                  *
1002                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1003                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1004                  * Assume B = 0 and we get C > 29.
1005                  */
1006                 ret = __ipv6_addr_src_scope(score->addr_type);
1007                 if (ret >= dst->scope)
1008                         ret = -ret;
1009                 else
1010                         ret -= 128;     /* 30 is enough */
1011                 score->scopedist = ret;
1012                 break;
1013         case IPV6_SADDR_RULE_PREFERRED:
1014                 /* Rule 3: Avoid deprecated and optimistic addresses */
1015                 ret = ipv6_saddr_preferred(score->addr_type) ||
1016                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1017                 break;
1018 #ifdef CONFIG_IPV6_MIP6
1019         case IPV6_SADDR_RULE_HOA:
1020             {
1021                 /* Rule 4: Prefer home address */
1022                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1023                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1024                 break;
1025             }
1026 #endif
1027         case IPV6_SADDR_RULE_OIF:
1028                 /* Rule 5: Prefer outgoing interface */
1029                 ret = (!dst->ifindex ||
1030                        dst->ifindex == score->ifa->idev->dev->ifindex);
1031                 break;
1032         case IPV6_SADDR_RULE_LABEL:
1033                 /* Rule 6: Prefer matching label */
1034                 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
1035                                       score->ifa->idev->dev->ifindex) == dst->label;
1036                 break;
1037 #ifdef CONFIG_IPV6_PRIVACY
1038         case IPV6_SADDR_RULE_PRIVACY:
1039             {
1040                 /* Rule 7: Prefer public address
1041                  * Note: prefer temprary address if use_tempaddr >= 2
1042                  */
1043                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1044                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1045                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1046                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1047                 break;
1048             }
1049 #endif
1050         case IPV6_SADDR_RULE_ORCHID:
1051                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1052                  *          non-ORCHID vs non-ORCHID
1053                  */
1054                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1055                         ipv6_addr_orchid(dst->addr));
1056                 break;
1057         case IPV6_SADDR_RULE_PREFIX:
1058                 /* Rule 8: Use longest matching prefix */
1059                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1060                                                        dst->addr);
1061                 break;
1062         default:
1063                 ret = 0;
1064         }
1065
1066         if (ret)
1067                 __set_bit(i, score->scorebits);
1068         score->rule = i;
1069 out:
1070         return ret;
1071 }
1072
1073 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1074                        const struct in6_addr *daddr, unsigned int prefs,
1075                        struct in6_addr *saddr)
1076 {
1077         struct ipv6_saddr_score scores[2],
1078                                 *score = &scores[0], *hiscore = &scores[1];
1079         struct net *net = dev_net(dst_dev);
1080         struct ipv6_saddr_dst dst;
1081         struct net_device *dev;
1082         int dst_type;
1083
1084         dst_type = __ipv6_addr_type(daddr);
1085         dst.addr = daddr;
1086         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1087         dst.scope = __ipv6_addr_src_scope(dst_type);
1088         dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1089         dst.prefs = prefs;
1090
1091         hiscore->rule = -1;
1092         hiscore->ifa = NULL;
1093
1094         read_lock(&dev_base_lock);
1095         rcu_read_lock();
1096
1097         for_each_netdev(net, dev) {
1098                 struct inet6_dev *idev;
1099
1100                 /* Candidate Source Address (section 4)
1101                  *  - multicast and link-local destination address,
1102                  *    the set of candidate source address MUST only
1103                  *    include addresses assigned to interfaces
1104                  *    belonging to the same link as the outgoing
1105                  *    interface.
1106                  * (- For site-local destination addresses, the
1107                  *    set of candidate source addresses MUST only
1108                  *    include addresses assigned to interfaces
1109                  *    belonging to the same site as the outgoing
1110                  *    interface.)
1111                  */
1112                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1113                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1114                     dst.ifindex && dev->ifindex != dst.ifindex)
1115                         continue;
1116
1117                 idev = __in6_dev_get(dev);
1118                 if (!idev)
1119                         continue;
1120
1121                 read_lock_bh(&idev->lock);
1122                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1123                         int i;
1124
1125                         /*
1126                          * - Tentative Address (RFC2462 section 5.4)
1127                          *  - A tentative address is not considered
1128                          *    "assigned to an interface" in the traditional
1129                          *    sense, unless it is also flagged as optimistic.
1130                          * - Candidate Source Address (section 4)
1131                          *  - In any case, anycast addresses, multicast
1132                          *    addresses, and the unspecified address MUST
1133                          *    NOT be included in a candidate set.
1134                          */
1135                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1136                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1137                                 continue;
1138
1139                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1140
1141                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1142                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1143                                 LIMIT_NETDEBUG(KERN_DEBUG
1144                                                "ADDRCONF: unspecified / multicast address "
1145                                                "assigned as unicast address on %s",
1146                                                dev->name);
1147                                 continue;
1148                         }
1149
1150                         score->rule = -1;
1151                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1152
1153                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1154                                 int minihiscore, miniscore;
1155
1156                                 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1157                                 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1158
1159                                 if (minihiscore > miniscore) {
1160                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1161                                             score->scopedist > 0) {
1162                                                 /*
1163                                                  * special case:
1164                                                  * each remaining entry
1165                                                  * has too small (not enough)
1166                                                  * scope, because ifa entries
1167                                                  * are sorted by their scope
1168                                                  * values.
1169                                                  */
1170                                                 goto try_nextdev;
1171                                         }
1172                                         break;
1173                                 } else if (minihiscore < miniscore) {
1174                                         struct ipv6_saddr_score *tmp;
1175
1176                                         if (hiscore->ifa)
1177                                                 in6_ifa_put(hiscore->ifa);
1178
1179                                         in6_ifa_hold(score->ifa);
1180
1181                                         tmp = hiscore;
1182                                         hiscore = score;
1183                                         score = tmp;
1184
1185                                         /* restore our iterator */
1186                                         score->ifa = hiscore->ifa;
1187
1188                                         break;
1189                                 }
1190                         }
1191                 }
1192 try_nextdev:
1193                 read_unlock_bh(&idev->lock);
1194         }
1195         rcu_read_unlock();
1196         read_unlock(&dev_base_lock);
1197
1198         if (!hiscore->ifa)
1199                 return -EADDRNOTAVAIL;
1200
1201         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1202         in6_ifa_put(hiscore->ifa);
1203         return 0;
1204 }
1205
1206 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1207
1208 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1209                     unsigned char banned_flags)
1210 {
1211         struct inet6_dev *idev;
1212         int err = -EADDRNOTAVAIL;
1213
1214         rcu_read_lock();
1215         if ((idev = __in6_dev_get(dev)) != NULL) {
1216                 struct inet6_ifaddr *ifp;
1217
1218                 read_lock_bh(&idev->lock);
1219                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1220                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1221                                 ipv6_addr_copy(addr, &ifp->addr);
1222                                 err = 0;
1223                                 break;
1224                         }
1225                 }
1226                 read_unlock_bh(&idev->lock);
1227         }
1228         rcu_read_unlock();
1229         return err;
1230 }
1231
1232 static int ipv6_count_addresses(struct inet6_dev *idev)
1233 {
1234         int cnt = 0;
1235         struct inet6_ifaddr *ifp;
1236
1237         read_lock_bh(&idev->lock);
1238         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1239                 cnt++;
1240         read_unlock_bh(&idev->lock);
1241         return cnt;
1242 }
1243
1244 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1245                   struct net_device *dev, int strict)
1246 {
1247         struct inet6_ifaddr * ifp;
1248         u8 hash = ipv6_addr_hash(addr);
1249
1250         read_lock_bh(&addrconf_hash_lock);
1251         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1252                 if (!net_eq(dev_net(ifp->idev->dev), net))
1253                         continue;
1254                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1255                     !(ifp->flags&IFA_F_TENTATIVE)) {
1256                         if (dev == NULL || ifp->idev->dev == dev ||
1257                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1258                                 break;
1259                 }
1260         }
1261         read_unlock_bh(&addrconf_hash_lock);
1262         return ifp != NULL;
1263 }
1264 EXPORT_SYMBOL(ipv6_chk_addr);
1265
1266 static
1267 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1268                        struct net_device *dev)
1269 {
1270         struct inet6_ifaddr * ifp;
1271         u8 hash = ipv6_addr_hash(addr);
1272
1273         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1274                 if (!net_eq(dev_net(ifp->idev->dev), net))
1275                         continue;
1276                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1277                         if (dev == NULL || ifp->idev->dev == dev)
1278                                 break;
1279                 }
1280         }
1281         return ifp != NULL;
1282 }
1283
1284 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1285 {
1286         struct inet6_dev *idev;
1287         struct inet6_ifaddr *ifa;
1288         int     onlink;
1289
1290         onlink = 0;
1291         rcu_read_lock();
1292         idev = __in6_dev_get(dev);
1293         if (idev) {
1294                 read_lock_bh(&idev->lock);
1295                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1296                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1297                                                    ifa->prefix_len);
1298                         if (onlink)
1299                                 break;
1300                 }
1301                 read_unlock_bh(&idev->lock);
1302         }
1303         rcu_read_unlock();
1304         return onlink;
1305 }
1306
1307 EXPORT_SYMBOL(ipv6_chk_prefix);
1308
1309 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1310                                      struct net_device *dev, int strict)
1311 {
1312         struct inet6_ifaddr * ifp;
1313         u8 hash = ipv6_addr_hash(addr);
1314
1315         read_lock_bh(&addrconf_hash_lock);
1316         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1317                 if (!net_eq(dev_net(ifp->idev->dev), net))
1318                         continue;
1319                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1320                         if (dev == NULL || ifp->idev->dev == dev ||
1321                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1322                                 in6_ifa_hold(ifp);
1323                                 break;
1324                         }
1325                 }
1326         }
1327         read_unlock_bh(&addrconf_hash_lock);
1328
1329         return ifp;
1330 }
1331
1332 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1333 {
1334         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1335         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1336         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1337         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1338         int sk_ipv6only = ipv6_only_sock(sk);
1339         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1340         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1341         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1342
1343         if (!sk2_rcv_saddr && !sk_ipv6only)
1344                 return 1;
1345
1346         if (addr_type2 == IPV6_ADDR_ANY &&
1347             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1348                 return 1;
1349
1350         if (addr_type == IPV6_ADDR_ANY &&
1351             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1352                 return 1;
1353
1354         if (sk2_rcv_saddr6 &&
1355             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1356                 return 1;
1357
1358         if (addr_type == IPV6_ADDR_MAPPED &&
1359             !sk2_ipv6only &&
1360             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1361                 return 1;
1362
1363         return 0;
1364 }
1365
1366 /* Gets referenced address, destroys ifaddr */
1367
1368 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1369 {
1370         if (ifp->flags&IFA_F_PERMANENT) {
1371                 spin_lock_bh(&ifp->lock);
1372                 addrconf_del_timer(ifp);
1373                 ifp->flags |= IFA_F_TENTATIVE;
1374                 spin_unlock_bh(&ifp->lock);
1375                 in6_ifa_put(ifp);
1376 #ifdef CONFIG_IPV6_PRIVACY
1377         } else if (ifp->flags&IFA_F_TEMPORARY) {
1378                 struct inet6_ifaddr *ifpub;
1379                 spin_lock_bh(&ifp->lock);
1380                 ifpub = ifp->ifpub;
1381                 if (ifpub) {
1382                         in6_ifa_hold(ifpub);
1383                         spin_unlock_bh(&ifp->lock);
1384                         ipv6_create_tempaddr(ifpub, ifp);
1385                         in6_ifa_put(ifpub);
1386                 } else {
1387                         spin_unlock_bh(&ifp->lock);
1388                 }
1389                 ipv6_del_addr(ifp);
1390 #endif
1391         } else
1392                 ipv6_del_addr(ifp);
1393 }
1394
1395 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1396 {
1397         if (net_ratelimit())
1398                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1399         addrconf_dad_stop(ifp);
1400 }
1401
1402 /* Join to solicited addr multicast group. */
1403
1404 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1405 {
1406         struct in6_addr maddr;
1407
1408         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1409                 return;
1410
1411         addrconf_addr_solict_mult(addr, &maddr);
1412         ipv6_dev_mc_inc(dev, &maddr);
1413 }
1414
1415 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1416 {
1417         struct in6_addr maddr;
1418
1419         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1420                 return;
1421
1422         addrconf_addr_solict_mult(addr, &maddr);
1423         __ipv6_dev_mc_dec(idev, &maddr);
1424 }
1425
1426 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1427 {
1428         struct in6_addr addr;
1429         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1430         if (ipv6_addr_any(&addr))
1431                 return;
1432         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1433 }
1434
1435 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1436 {
1437         struct in6_addr addr;
1438         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1439         if (ipv6_addr_any(&addr))
1440                 return;
1441         __ipv6_dev_ac_dec(ifp->idev, &addr);
1442 }
1443
1444 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1445 {
1446         if (dev->addr_len != ETH_ALEN)
1447                 return -1;
1448         memcpy(eui, dev->dev_addr, 3);
1449         memcpy(eui + 5, dev->dev_addr + 3, 3);
1450
1451         /*
1452          * The zSeries OSA network cards can be shared among various
1453          * OS instances, but the OSA cards have only one MAC address.
1454          * This leads to duplicate address conflicts in conjunction
1455          * with IPv6 if more than one instance uses the same card.
1456          *
1457          * The driver for these cards can deliver a unique 16-bit
1458          * identifier for each instance sharing the same card.  It is
1459          * placed instead of 0xFFFE in the interface identifier.  The
1460          * "u" bit of the interface identifier is not inverted in this
1461          * case.  Hence the resulting interface identifier has local
1462          * scope according to RFC2373.
1463          */
1464         if (dev->dev_id) {
1465                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1466                 eui[4] = dev->dev_id & 0xFF;
1467         } else {
1468                 eui[3] = 0xFF;
1469                 eui[4] = 0xFE;
1470                 eui[0] ^= 2;
1471         }
1472         return 0;
1473 }
1474
1475 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1476 {
1477         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1478         if (dev->addr_len != ARCNET_ALEN)
1479                 return -1;
1480         memset(eui, 0, 7);
1481         eui[7] = *(u8*)dev->dev_addr;
1482         return 0;
1483 }
1484
1485 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1486 {
1487         if (dev->addr_len != INFINIBAND_ALEN)
1488                 return -1;
1489         memcpy(eui, dev->dev_addr + 12, 8);
1490         eui[0] |= 2;
1491         return 0;
1492 }
1493
1494 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1495 {
1496         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1497                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1498                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1499                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1500                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1501                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1502         eui[1] = 0;
1503         eui[2] = 0x5E;
1504         eui[3] = 0xFE;
1505         memcpy(eui + 4, &addr, 4);
1506         return 0;
1507 }
1508 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1509
1510 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1511 {
1512         if (dev->priv_flags & IFF_ISATAP)
1513                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1514         return -1;
1515 }
1516
1517 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1518 {
1519         switch (dev->type) {
1520         case ARPHRD_ETHER:
1521         case ARPHRD_FDDI:
1522         case ARPHRD_IEEE802_TR:
1523                 return addrconf_ifid_eui48(eui, dev);
1524         case ARPHRD_ARCNET:
1525                 return addrconf_ifid_arcnet(eui, dev);
1526         case ARPHRD_INFINIBAND:
1527                 return addrconf_ifid_infiniband(eui, dev);
1528         case ARPHRD_SIT:
1529                 return addrconf_ifid_sit(eui, dev);
1530         }
1531         return -1;
1532 }
1533
1534 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1535 {
1536         int err = -1;
1537         struct inet6_ifaddr *ifp;
1538
1539         read_lock_bh(&idev->lock);
1540         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1541                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1542                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1543                         err = 0;
1544                         break;
1545                 }
1546         }
1547         read_unlock_bh(&idev->lock);
1548         return err;
1549 }
1550
1551 #ifdef CONFIG_IPV6_PRIVACY
1552 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1553 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1554 {
1555 regen:
1556         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1557         idev->rndid[0] &= ~0x02;
1558
1559         /*
1560          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1561          * check if generated address is not inappropriate
1562          *
1563          *  - Reserved subnet anycast (RFC 2526)
1564          *      11111101 11....11 1xxxxxxx
1565          *  - ISATAP (RFC4214) 6.1
1566          *      00-00-5E-FE-xx-xx-xx-xx
1567          *  - value 0
1568          *  - XXX: already assigned to an address on the device
1569          */
1570         if (idev->rndid[0] == 0xfd &&
1571             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1572             (idev->rndid[7]&0x80))
1573                 goto regen;
1574         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1575                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1576                         goto regen;
1577                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1578                         goto regen;
1579         }
1580
1581         return 0;
1582 }
1583
1584 static void ipv6_regen_rndid(unsigned long data)
1585 {
1586         struct inet6_dev *idev = (struct inet6_dev *) data;
1587         unsigned long expires;
1588
1589         rcu_read_lock_bh();
1590         write_lock_bh(&idev->lock);
1591
1592         if (idev->dead)
1593                 goto out;
1594
1595         if (__ipv6_regen_rndid(idev) < 0)
1596                 goto out;
1597
1598         expires = jiffies +
1599                 idev->cnf.temp_prefered_lft * HZ -
1600                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1601         if (time_before(expires, jiffies)) {
1602                 printk(KERN_WARNING
1603                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1604                         idev->dev->name);
1605                 goto out;
1606         }
1607
1608         if (!mod_timer(&idev->regen_timer, expires))
1609                 in6_dev_hold(idev);
1610
1611 out:
1612         write_unlock_bh(&idev->lock);
1613         rcu_read_unlock_bh();
1614         in6_dev_put(idev);
1615 }
1616
1617 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1618         int ret = 0;
1619
1620         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1621                 ret = __ipv6_regen_rndid(idev);
1622         return ret;
1623 }
1624 #endif
1625
1626 /*
1627  *      Add prefix route.
1628  */
1629
1630 static void
1631 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1632                       unsigned long expires, u32 flags)
1633 {
1634         struct fib6_config cfg = {
1635                 .fc_table = RT6_TABLE_PREFIX,
1636                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1637                 .fc_ifindex = dev->ifindex,
1638                 .fc_expires = expires,
1639                 .fc_dst_len = plen,
1640                 .fc_flags = RTF_UP | flags,
1641                 .fc_nlinfo.nl_net = dev_net(dev),
1642         };
1643
1644         ipv6_addr_copy(&cfg.fc_dst, pfx);
1645
1646         /* Prevent useless cloning on PtP SIT.
1647            This thing is done here expecting that the whole
1648            class of non-broadcast devices need not cloning.
1649          */
1650 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1651         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1652                 cfg.fc_flags |= RTF_NONEXTHOP;
1653 #endif
1654
1655         ip6_route_add(&cfg);
1656 }
1657
1658 /* Create "default" multicast route to the interface */
1659
1660 static void addrconf_add_mroute(struct net_device *dev)
1661 {
1662         struct fib6_config cfg = {
1663                 .fc_table = RT6_TABLE_LOCAL,
1664                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1665                 .fc_ifindex = dev->ifindex,
1666                 .fc_dst_len = 8,
1667                 .fc_flags = RTF_UP,
1668                 .fc_nlinfo.nl_net = dev_net(dev),
1669         };
1670
1671         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1672
1673         ip6_route_add(&cfg);
1674 }
1675
1676 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1677 static void sit_route_add(struct net_device *dev)
1678 {
1679         struct fib6_config cfg = {
1680                 .fc_table = RT6_TABLE_MAIN,
1681                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1682                 .fc_ifindex = dev->ifindex,
1683                 .fc_dst_len = 96,
1684                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1685                 .fc_nlinfo.nl_net = dev_net(dev),
1686         };
1687
1688         /* prefix length - 96 bits "::d.d.d.d" */
1689         ip6_route_add(&cfg);
1690 }
1691 #endif
1692
1693 static void addrconf_add_lroute(struct net_device *dev)
1694 {
1695         struct in6_addr addr;
1696
1697         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1698         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1699 }
1700
1701 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1702 {
1703         struct inet6_dev *idev;
1704
1705         ASSERT_RTNL();
1706
1707         if ((idev = ipv6_find_idev(dev)) == NULL)
1708                 return NULL;
1709
1710         /* Add default multicast route */
1711         addrconf_add_mroute(dev);
1712
1713         /* Add link local route */
1714         addrconf_add_lroute(dev);
1715         return idev;
1716 }
1717
1718 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1719 {
1720         struct prefix_info *pinfo;
1721         __u32 valid_lft;
1722         __u32 prefered_lft;
1723         int addr_type;
1724         unsigned long rt_expires;
1725         struct inet6_dev *in6_dev;
1726
1727         pinfo = (struct prefix_info *) opt;
1728
1729         if (len < sizeof(struct prefix_info)) {
1730                 ADBG(("addrconf: prefix option too short\n"));
1731                 return;
1732         }
1733
1734         /*
1735          *      Validation checks ([ADDRCONF], page 19)
1736          */
1737
1738         addr_type = ipv6_addr_type(&pinfo->prefix);
1739
1740         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1741                 return;
1742
1743         valid_lft = ntohl(pinfo->valid);
1744         prefered_lft = ntohl(pinfo->prefered);
1745
1746         if (prefered_lft > valid_lft) {
1747                 if (net_ratelimit())
1748                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1749                 return;
1750         }
1751
1752         in6_dev = in6_dev_get(dev);
1753
1754         if (in6_dev == NULL) {
1755                 if (net_ratelimit())
1756                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1757                 return;
1758         }
1759
1760         /*
1761          *      Two things going on here:
1762          *      1) Add routes for on-link prefixes
1763          *      2) Configure prefixes with the auto flag set
1764          */
1765
1766         /* Avoid arithmetic overflow. Really, we could
1767            save rt_expires in seconds, likely valid_lft,
1768            but it would require division in fib gc, that it
1769            not good.
1770          */
1771         if (valid_lft >= 0x7FFFFFFF/HZ)
1772                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1773         else
1774                 rt_expires = valid_lft * HZ;
1775
1776         /*
1777          * We convert this (in jiffies) to clock_t later.
1778          * Avoid arithmetic overflow there as well.
1779          * Overflow can happen only if HZ < USER_HZ.
1780          */
1781         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1782                 rt_expires = 0x7FFFFFFF / USER_HZ;
1783
1784         if (pinfo->onlink) {
1785                 struct rt6_info *rt;
1786                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1787                                 dev->ifindex, 1);
1788
1789                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1790                         if (rt->rt6i_flags&RTF_EXPIRES) {
1791                                 if (valid_lft == 0) {
1792                                         ip6_del_rt(rt);
1793                                         rt = NULL;
1794                                 } else {
1795                                         rt->rt6i_expires = jiffies + rt_expires;
1796                                 }
1797                         }
1798                 } else if (valid_lft) {
1799                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1800                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1801                 }
1802                 if (rt)
1803                         dst_release(&rt->u.dst);
1804         }
1805
1806         /* Try to figure out our local address for this prefix */
1807
1808         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1809                 struct inet6_ifaddr * ifp;
1810                 struct in6_addr addr;
1811                 int create = 0, update_lft = 0;
1812
1813                 if (pinfo->prefix_len == 64) {
1814                         memcpy(&addr, &pinfo->prefix, 8);
1815                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1816                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1817                                 in6_dev_put(in6_dev);
1818                                 return;
1819                         }
1820                         goto ok;
1821                 }
1822                 if (net_ratelimit())
1823                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1824                                pinfo->prefix_len);
1825                 in6_dev_put(in6_dev);
1826                 return;
1827
1828 ok:
1829
1830                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1831
1832                 if (ifp == NULL && valid_lft) {
1833                         int max_addresses = in6_dev->cnf.max_addresses;
1834                         u32 addr_flags = 0;
1835
1836 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1837                         if (in6_dev->cnf.optimistic_dad &&
1838                             !ipv6_devconf.forwarding)
1839                                 addr_flags = IFA_F_OPTIMISTIC;
1840 #endif
1841
1842                         /* Do not allow to create too much of autoconfigured
1843                          * addresses; this would be too easy way to crash kernel.
1844                          */
1845                         if (!max_addresses ||
1846                             ipv6_count_addresses(in6_dev) < max_addresses)
1847                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1848                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1849                                                     addr_flags);
1850
1851                         if (!ifp || IS_ERR(ifp)) {
1852                                 in6_dev_put(in6_dev);
1853                                 return;
1854                         }
1855
1856                         update_lft = create = 1;
1857                         ifp->cstamp = jiffies;
1858                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1859                 }
1860
1861                 if (ifp) {
1862                         int flags;
1863                         unsigned long now;
1864 #ifdef CONFIG_IPV6_PRIVACY
1865                         struct inet6_ifaddr *ift;
1866 #endif
1867                         u32 stored_lft;
1868
1869                         /* update lifetime (RFC2462 5.5.3 e) */
1870                         spin_lock(&ifp->lock);
1871                         now = jiffies;
1872                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1873                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1874                         else
1875                                 stored_lft = 0;
1876                         if (!update_lft && stored_lft) {
1877                                 if (valid_lft > MIN_VALID_LIFETIME ||
1878                                     valid_lft > stored_lft)
1879                                         update_lft = 1;
1880                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1881                                         /* valid_lft <= stored_lft is always true */
1882                                         /* XXX: IPsec */
1883                                         update_lft = 0;
1884                                 } else {
1885                                         valid_lft = MIN_VALID_LIFETIME;
1886                                         if (valid_lft < prefered_lft)
1887                                                 prefered_lft = valid_lft;
1888                                         update_lft = 1;
1889                                 }
1890                         }
1891
1892                         if (update_lft) {
1893                                 ifp->valid_lft = valid_lft;
1894                                 ifp->prefered_lft = prefered_lft;
1895                                 ifp->tstamp = now;
1896                                 flags = ifp->flags;
1897                                 ifp->flags &= ~IFA_F_DEPRECATED;
1898                                 spin_unlock(&ifp->lock);
1899
1900                                 if (!(flags&IFA_F_TENTATIVE))
1901                                         ipv6_ifa_notify(0, ifp);
1902                         } else
1903                                 spin_unlock(&ifp->lock);
1904
1905 #ifdef CONFIG_IPV6_PRIVACY
1906                         read_lock_bh(&in6_dev->lock);
1907                         /* update all temporary addresses in the list */
1908                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1909                                 /*
1910                                  * When adjusting the lifetimes of an existing
1911                                  * temporary address, only lower the lifetimes.
1912                                  * Implementations must not increase the
1913                                  * lifetimes of an existing temporary address
1914                                  * when processing a Prefix Information Option.
1915                                  */
1916                                 if (ifp != ift->ifpub)
1917                                         continue;
1918
1919                                 spin_lock(&ift->lock);
1920                                 flags = ift->flags;
1921                                 if (ift->valid_lft > valid_lft &&
1922                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1923                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1924                                 if (ift->prefered_lft > prefered_lft &&
1925                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1926                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1927                                 spin_unlock(&ift->lock);
1928                                 if (!(flags&IFA_F_TENTATIVE))
1929                                         ipv6_ifa_notify(0, ift);
1930                         }
1931
1932                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1933                                 /*
1934                                  * When a new public address is created as described in [ADDRCONF],
1935                                  * also create a new temporary address.
1936                                  */
1937                                 read_unlock_bh(&in6_dev->lock);
1938                                 ipv6_create_tempaddr(ifp, NULL);
1939                         } else {
1940                                 read_unlock_bh(&in6_dev->lock);
1941                         }
1942 #endif
1943                         in6_ifa_put(ifp);
1944                         addrconf_verify(0);
1945                 }
1946         }
1947         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1948         in6_dev_put(in6_dev);
1949 }
1950
1951 /*
1952  *      Set destination address.
1953  *      Special case for SIT interfaces where we create a new "virtual"
1954  *      device.
1955  */
1956 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1957 {
1958         struct in6_ifreq ireq;
1959         struct net_device *dev;
1960         int err = -EINVAL;
1961
1962         rtnl_lock();
1963
1964         err = -EFAULT;
1965         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1966                 goto err_exit;
1967
1968         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1969
1970         err = -ENODEV;
1971         if (dev == NULL)
1972                 goto err_exit;
1973
1974 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1975         if (dev->type == ARPHRD_SIT) {
1976                 struct ifreq ifr;
1977                 mm_segment_t    oldfs;
1978                 struct ip_tunnel_parm p;
1979
1980                 err = -EADDRNOTAVAIL;
1981                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1982                         goto err_exit;
1983
1984                 memset(&p, 0, sizeof(p));
1985                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1986                 p.iph.saddr = 0;
1987                 p.iph.version = 4;
1988                 p.iph.ihl = 5;
1989                 p.iph.protocol = IPPROTO_IPV6;
1990                 p.iph.ttl = 64;
1991                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1992
1993                 oldfs = get_fs(); set_fs(KERNEL_DS);
1994                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1995                 set_fs(oldfs);
1996
1997                 if (err == 0) {
1998                         err = -ENOBUFS;
1999                         dev = __dev_get_by_name(net, p.name);
2000                         if (!dev)
2001                                 goto err_exit;
2002                         err = dev_open(dev);
2003                 }
2004         }
2005 #endif
2006
2007 err_exit:
2008         rtnl_unlock();
2009         return err;
2010 }
2011
2012 /*
2013  *      Manual configuration of address on an interface
2014  */
2015 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2016                           int plen, __u8 ifa_flags, __u32 prefered_lft,
2017                           __u32 valid_lft)
2018 {
2019         struct inet6_ifaddr *ifp;
2020         struct inet6_dev *idev;
2021         struct net_device *dev;
2022         int scope;
2023         u32 flags = RTF_EXPIRES;
2024
2025         ASSERT_RTNL();
2026
2027         /* check the lifetime */
2028         if (!valid_lft || prefered_lft > valid_lft)
2029                 return -EINVAL;
2030
2031         dev = __dev_get_by_index(net, ifindex);
2032         if (!dev)
2033                 return -ENODEV;
2034
2035         if ((idev = addrconf_add_dev(dev)) == NULL)
2036                 return -ENOBUFS;
2037
2038         scope = ipv6_addr_scope(pfx);
2039
2040         if (valid_lft == INFINITY_LIFE_TIME) {
2041                 ifa_flags |= IFA_F_PERMANENT;
2042                 flags = 0;
2043         } else if (valid_lft >= 0x7FFFFFFF/HZ)
2044                 valid_lft = 0x7FFFFFFF/HZ;
2045
2046         if (prefered_lft == 0)
2047                 ifa_flags |= IFA_F_DEPRECATED;
2048         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2049                  (prefered_lft != INFINITY_LIFE_TIME))
2050                 prefered_lft = 0x7FFFFFFF/HZ;
2051
2052         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2053
2054         if (!IS_ERR(ifp)) {
2055                 spin_lock_bh(&ifp->lock);
2056                 ifp->valid_lft = valid_lft;
2057                 ifp->prefered_lft = prefered_lft;
2058                 ifp->tstamp = jiffies;
2059                 spin_unlock_bh(&ifp->lock);
2060
2061                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2062                                       jiffies_to_clock_t(valid_lft * HZ), flags);
2063                 /*
2064                  * Note that section 3.1 of RFC 4429 indicates
2065                  * that the Optimistic flag should not be set for
2066                  * manually configured addresses
2067                  */
2068                 addrconf_dad_start(ifp, 0);
2069                 in6_ifa_put(ifp);
2070                 addrconf_verify(0);
2071                 return 0;
2072         }
2073
2074         return PTR_ERR(ifp);
2075 }
2076
2077 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2078                           int plen)
2079 {
2080         struct inet6_ifaddr *ifp;
2081         struct inet6_dev *idev;
2082         struct net_device *dev;
2083
2084         dev = __dev_get_by_index(net, ifindex);
2085         if (!dev)
2086                 return -ENODEV;
2087
2088         if ((idev = __in6_dev_get(dev)) == NULL)
2089                 return -ENXIO;
2090
2091         read_lock_bh(&idev->lock);
2092         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2093                 if (ifp->prefix_len == plen &&
2094                     ipv6_addr_equal(pfx, &ifp->addr)) {
2095                         in6_ifa_hold(ifp);
2096                         read_unlock_bh(&idev->lock);
2097
2098                         ipv6_del_addr(ifp);
2099
2100                         /* If the last address is deleted administratively,
2101                            disable IPv6 on this interface.
2102                          */
2103                         if (idev->addr_list == NULL)
2104                                 addrconf_ifdown(idev->dev, 1);
2105                         return 0;
2106                 }
2107         }
2108         read_unlock_bh(&idev->lock);
2109         return -EADDRNOTAVAIL;
2110 }
2111
2112
2113 int addrconf_add_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_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2126                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2127                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2128         rtnl_unlock();
2129         return err;
2130 }
2131
2132 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2133 {
2134         struct in6_ifreq ireq;
2135         int err;
2136
2137         if (!capable(CAP_NET_ADMIN))
2138                 return -EPERM;
2139
2140         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2141                 return -EFAULT;
2142
2143         rtnl_lock();
2144         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2145                              ireq.ifr6_prefixlen);
2146         rtnl_unlock();
2147         return err;
2148 }
2149
2150 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2151 static void sit_add_v4_addrs(struct inet6_dev *idev)
2152 {
2153         struct inet6_ifaddr * ifp;
2154         struct in6_addr addr;
2155         struct net_device *dev;
2156         struct net *net = dev_net(idev->dev);
2157         int scope;
2158
2159         ASSERT_RTNL();
2160
2161         memset(&addr, 0, sizeof(struct in6_addr));
2162         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2163
2164         if (idev->dev->flags&IFF_POINTOPOINT) {
2165                 addr.s6_addr32[0] = htonl(0xfe800000);
2166                 scope = IFA_LINK;
2167         } else {
2168                 scope = IPV6_ADDR_COMPATv4;
2169         }
2170
2171         if (addr.s6_addr32[3]) {
2172                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2173                 if (!IS_ERR(ifp)) {
2174                         spin_lock_bh(&ifp->lock);
2175                         ifp->flags &= ~IFA_F_TENTATIVE;
2176                         spin_unlock_bh(&ifp->lock);
2177                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2178                         in6_ifa_put(ifp);
2179                 }
2180                 return;
2181         }
2182
2183         for_each_netdev(net, dev) {
2184                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2185                 if (in_dev && (dev->flags & IFF_UP)) {
2186                         struct in_ifaddr * ifa;
2187
2188                         int flag = scope;
2189
2190                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2191                                 int plen;
2192
2193                                 addr.s6_addr32[3] = ifa->ifa_local;
2194
2195                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2196                                         continue;
2197                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2198                                         if (idev->dev->flags&IFF_POINTOPOINT)
2199                                                 continue;
2200                                         flag |= IFA_HOST;
2201                                 }
2202                                 if (idev->dev->flags&IFF_POINTOPOINT)
2203                                         plen = 64;
2204                                 else
2205                                         plen = 96;
2206
2207                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2208                                                     IFA_F_PERMANENT);
2209                                 if (!IS_ERR(ifp)) {
2210                                         spin_lock_bh(&ifp->lock);
2211                                         ifp->flags &= ~IFA_F_TENTATIVE;
2212                                         spin_unlock_bh(&ifp->lock);
2213                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2214                                         in6_ifa_put(ifp);
2215                                 }
2216                         }
2217                 }
2218         }
2219 }
2220 #endif
2221
2222 static void init_loopback(struct net_device *dev)
2223 {
2224         struct inet6_dev  *idev;
2225         struct inet6_ifaddr * ifp;
2226
2227         /* ::1 */
2228
2229         ASSERT_RTNL();
2230
2231         if ((idev = ipv6_find_idev(dev)) == NULL) {
2232                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2233                 return;
2234         }
2235
2236         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2237         if (!IS_ERR(ifp)) {
2238                 spin_lock_bh(&ifp->lock);
2239                 ifp->flags &= ~IFA_F_TENTATIVE;
2240                 spin_unlock_bh(&ifp->lock);
2241                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2242                 in6_ifa_put(ifp);
2243         }
2244 }
2245
2246 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2247 {
2248         struct inet6_ifaddr * ifp;
2249         u32 addr_flags = IFA_F_PERMANENT;
2250
2251 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2252         if (idev->cnf.optimistic_dad &&
2253             !ipv6_devconf.forwarding)
2254                 addr_flags |= IFA_F_OPTIMISTIC;
2255 #endif
2256
2257
2258         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2259         if (!IS_ERR(ifp)) {
2260                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2261                 addrconf_dad_start(ifp, 0);
2262                 in6_ifa_put(ifp);
2263         }
2264 }
2265
2266 static void addrconf_dev_config(struct net_device *dev)
2267 {
2268         struct in6_addr addr;
2269         struct inet6_dev    * idev;
2270
2271         ASSERT_RTNL();
2272
2273         if ((dev->type != ARPHRD_ETHER) &&
2274             (dev->type != ARPHRD_FDDI) &&
2275             (dev->type != ARPHRD_IEEE802_TR) &&
2276             (dev->type != ARPHRD_ARCNET) &&
2277             (dev->type != ARPHRD_INFINIBAND)) {
2278                 /* Alas, we support only Ethernet autoconfiguration. */
2279                 return;
2280         }
2281
2282         idev = addrconf_add_dev(dev);
2283         if (idev == NULL)
2284                 return;
2285
2286         memset(&addr, 0, sizeof(struct in6_addr));
2287         addr.s6_addr32[0] = htonl(0xFE800000);
2288
2289         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2290                 addrconf_add_linklocal(idev, &addr);
2291 }
2292
2293 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2294 static void addrconf_sit_config(struct net_device *dev)
2295 {
2296         struct inet6_dev *idev;
2297
2298         ASSERT_RTNL();
2299
2300         /*
2301          * Configure the tunnel with one of our IPv4
2302          * addresses... we should configure all of
2303          * our v4 addrs in the tunnel
2304          */
2305
2306         if ((idev = ipv6_find_idev(dev)) == NULL) {
2307                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2308                 return;
2309         }
2310
2311         if (dev->priv_flags & IFF_ISATAP) {
2312                 struct in6_addr addr;
2313
2314                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2315                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2316                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2317                         addrconf_add_linklocal(idev, &addr);
2318                 return;
2319         }
2320
2321         sit_add_v4_addrs(idev);
2322
2323         if (dev->flags&IFF_POINTOPOINT) {
2324                 addrconf_add_mroute(dev);
2325                 addrconf_add_lroute(dev);
2326         } else
2327                 sit_route_add(dev);
2328 }
2329 #endif
2330
2331 static inline int
2332 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2333 {
2334         struct in6_addr lladdr;
2335
2336         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2337                 addrconf_add_linklocal(idev, &lladdr);
2338                 return 0;
2339         }
2340         return -1;
2341 }
2342
2343 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2344 {
2345         struct net_device *link_dev;
2346         struct net *net = dev_net(idev->dev);
2347
2348         /* first try to inherit the link-local address from the link device */
2349         if (idev->dev->iflink &&
2350             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2351                 if (!ipv6_inherit_linklocal(idev, link_dev))
2352                         return;
2353         }
2354         /* then try to inherit it from any device */
2355         for_each_netdev(net, link_dev) {
2356                 if (!ipv6_inherit_linklocal(idev, link_dev))
2357                         return;
2358         }
2359         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2360 }
2361
2362 /*
2363  * Autoconfigure tunnel with a link-local address so routing protocols,
2364  * DHCPv6, MLD etc. can be run over the virtual link
2365  */
2366
2367 static void addrconf_ip6_tnl_config(struct net_device *dev)
2368 {
2369         struct inet6_dev *idev;
2370
2371         ASSERT_RTNL();
2372
2373         if ((idev = addrconf_add_dev(dev)) == NULL) {
2374                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2375                 return;
2376         }
2377         ip6_tnl_add_linklocal(idev);
2378 }
2379
2380 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2381                            void * data)
2382 {
2383         struct net_device *dev = (struct net_device *) data;
2384         struct inet6_dev *idev = __in6_dev_get(dev);
2385         int run_pending = 0;
2386         int err;
2387
2388         switch(event) {
2389         case NETDEV_REGISTER:
2390                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2391                         idev = ipv6_add_dev(dev);
2392                         if (!idev)
2393                                 return notifier_from_errno(-ENOMEM);
2394                 }
2395                 break;
2396         case NETDEV_UP:
2397         case NETDEV_CHANGE:
2398                 if (dev->flags & IFF_SLAVE)
2399                         break;
2400
2401                 if (event == NETDEV_UP) {
2402                         if (!addrconf_qdisc_ok(dev)) {
2403                                 /* device is not ready yet. */
2404                                 printk(KERN_INFO
2405                                         "ADDRCONF(NETDEV_UP): %s: "
2406                                         "link is not ready\n",
2407                                         dev->name);
2408                                 break;
2409                         }
2410
2411                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2412                                 idev = ipv6_add_dev(dev);
2413
2414                         if (idev)
2415                                 idev->if_flags |= IF_READY;
2416                 } else {
2417                         if (!addrconf_qdisc_ok(dev)) {
2418                                 /* device is still not ready. */
2419                                 break;
2420                         }
2421
2422                         if (idev) {
2423                                 if (idev->if_flags & IF_READY) {
2424                                         /* device is already configured. */
2425                                         break;
2426                                 }
2427                                 idev->if_flags |= IF_READY;
2428                         }
2429
2430                         printk(KERN_INFO
2431                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2432                                         "link becomes ready\n",
2433                                         dev->name);
2434
2435                         run_pending = 1;
2436                 }
2437
2438                 switch(dev->type) {
2439 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2440                 case ARPHRD_SIT:
2441                         addrconf_sit_config(dev);
2442                         break;
2443 #endif
2444                 case ARPHRD_TUNNEL6:
2445                         addrconf_ip6_tnl_config(dev);
2446                         break;
2447                 case ARPHRD_LOOPBACK:
2448                         init_loopback(dev);
2449                         break;
2450
2451                 default:
2452                         addrconf_dev_config(dev);
2453                         break;
2454                 }
2455                 if (idev) {
2456                         if (run_pending)
2457                                 addrconf_dad_run(idev);
2458
2459                         /* If the MTU changed during the interface down, when the
2460                            interface up, the changed MTU must be reflected in the
2461                            idev as well as routers.
2462                          */
2463                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2464                                 rt6_mtu_change(dev, dev->mtu);
2465                                 idev->cnf.mtu6 = dev->mtu;
2466                         }
2467                         idev->tstamp = jiffies;
2468                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2469                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2470                            stop IPv6 on this interface.
2471                          */
2472                         if (dev->mtu < IPV6_MIN_MTU)
2473                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2474                 }
2475                 break;
2476
2477         case NETDEV_CHANGEMTU:
2478                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2479                         rt6_mtu_change(dev, dev->mtu);
2480                         idev->cnf.mtu6 = dev->mtu;
2481                         break;
2482                 }
2483
2484                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2485                         idev = ipv6_add_dev(dev);
2486                         if (idev)
2487                                 break;
2488                 }
2489
2490                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2491
2492         case NETDEV_DOWN:
2493         case NETDEV_UNREGISTER:
2494                 /*
2495                  *      Remove all addresses from this interface.
2496                  */
2497                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2498                 break;
2499
2500         case NETDEV_CHANGENAME:
2501                 if (idev) {
2502                         snmp6_unregister_dev(idev);
2503                         addrconf_sysctl_unregister(idev);
2504                         addrconf_sysctl_register(idev);
2505                         err = snmp6_register_dev(idev);
2506                         if (err)
2507                                 return notifier_from_errno(err);
2508                 }
2509                 break;
2510         }
2511
2512         return NOTIFY_OK;
2513 }
2514
2515 /*
2516  *      addrconf module should be notified of a device going up
2517  */
2518 static struct notifier_block ipv6_dev_notf = {
2519         .notifier_call = addrconf_notify,
2520         .priority = 0
2521 };
2522
2523 static int addrconf_ifdown(struct net_device *dev, int how)
2524 {
2525         struct inet6_dev *idev;
2526         struct inet6_ifaddr *ifa, **bifa;
2527         struct net *net = dev_net(dev);
2528         int i;
2529
2530         ASSERT_RTNL();
2531
2532         if (dev == init_net.loopback_dev && how == 1)
2533                 how = 0;
2534
2535         rt6_ifdown(net, dev);
2536         neigh_ifdown(&nd_tbl, dev);
2537
2538         idev = __in6_dev_get(dev);
2539         if (idev == NULL)
2540                 return -ENODEV;
2541
2542         /* Step 1: remove reference to ipv6 device from parent device.
2543                    Do not dev_put!
2544          */
2545         if (how == 1) {
2546                 idev->dead = 1;
2547
2548                 /* protected by rtnl_lock */
2549                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2550
2551                 /* Step 1.5: remove snmp6 entry */
2552                 snmp6_unregister_dev(idev);
2553
2554         }
2555
2556         /* Step 2: clear hash table */
2557         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2558                 bifa = &inet6_addr_lst[i];
2559
2560                 write_lock_bh(&addrconf_hash_lock);
2561                 while ((ifa = *bifa) != NULL) {
2562                         if (ifa->idev == idev) {
2563                                 *bifa = ifa->lst_next;
2564                                 ifa->lst_next = NULL;
2565                                 addrconf_del_timer(ifa);
2566                                 in6_ifa_put(ifa);
2567                                 continue;
2568                         }
2569                         bifa = &ifa->lst_next;
2570                 }
2571                 write_unlock_bh(&addrconf_hash_lock);
2572         }
2573
2574         write_lock_bh(&idev->lock);
2575
2576         /* Step 3: clear flags for stateless addrconf */
2577         if (how != 1)
2578                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2579
2580         /* Step 4: clear address list */
2581 #ifdef CONFIG_IPV6_PRIVACY
2582         if (how == 1 && del_timer(&idev->regen_timer))
2583                 in6_dev_put(idev);
2584
2585         /* clear tempaddr list */
2586         while ((ifa = idev->tempaddr_list) != NULL) {
2587                 idev->tempaddr_list = ifa->tmp_next;
2588                 ifa->tmp_next = NULL;
2589                 ifa->dead = 1;
2590                 write_unlock_bh(&idev->lock);
2591                 spin_lock_bh(&ifa->lock);
2592
2593                 if (ifa->ifpub) {
2594                         in6_ifa_put(ifa->ifpub);
2595                         ifa->ifpub = NULL;
2596                 }
2597                 spin_unlock_bh(&ifa->lock);
2598                 in6_ifa_put(ifa);
2599                 write_lock_bh(&idev->lock);
2600         }
2601 #endif
2602         while ((ifa = idev->addr_list) != NULL) {
2603                 idev->addr_list = ifa->if_next;
2604                 ifa->if_next = NULL;
2605                 ifa->dead = 1;
2606                 addrconf_del_timer(ifa);
2607                 write_unlock_bh(&idev->lock);
2608
2609                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2610                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2611                 in6_ifa_put(ifa);
2612
2613                 write_lock_bh(&idev->lock);
2614         }
2615         write_unlock_bh(&idev->lock);
2616
2617         /* Step 5: Discard multicast list */
2618
2619         if (how == 1)
2620                 ipv6_mc_destroy_dev(idev);
2621         else
2622                 ipv6_mc_down(idev);
2623
2624         idev->tstamp = jiffies;
2625
2626         /* Shot the device (if unregistered) */
2627
2628         if (how == 1) {
2629                 addrconf_sysctl_unregister(idev);
2630                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2631                 neigh_ifdown(&nd_tbl, dev);
2632                 in6_dev_put(idev);
2633         }
2634         return 0;
2635 }
2636
2637 static void addrconf_rs_timer(unsigned long data)
2638 {
2639         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2640
2641         if (ifp->idev->cnf.forwarding)
2642                 goto out;
2643
2644         if (ifp->idev->if_flags & IF_RA_RCVD) {
2645                 /*
2646                  *      Announcement received after solicitation
2647                  *      was sent
2648                  */
2649                 goto out;
2650         }
2651
2652         spin_lock(&ifp->lock);
2653         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2654                 /* The wait after the last probe can be shorter */
2655                 addrconf_mod_timer(ifp, AC_RS,
2656                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2657                                    ifp->idev->cnf.rtr_solicit_delay :
2658                                    ifp->idev->cnf.rtr_solicit_interval);
2659                 spin_unlock(&ifp->lock);
2660
2661                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2662         } else {
2663                 spin_unlock(&ifp->lock);
2664                 /*
2665                  * Note: we do not support deprecated "all on-link"
2666                  * assumption any longer.
2667                  */
2668                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2669                        ifp->idev->dev->name);
2670         }
2671
2672 out:
2673         in6_ifa_put(ifp);
2674 }
2675
2676 /*
2677  *      Duplicate Address Detection
2678  */
2679 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2680 {
2681         unsigned long rand_num;
2682         struct inet6_dev *idev = ifp->idev;
2683
2684         if (ifp->flags & IFA_F_OPTIMISTIC)
2685                 rand_num = 0;
2686         else
2687                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2688
2689         ifp->probes = idev->cnf.dad_transmits;
2690         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2691 }
2692
2693 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2694 {
2695         struct inet6_dev *idev = ifp->idev;
2696         struct net_device *dev = idev->dev;
2697
2698         addrconf_join_solict(dev, &ifp->addr);
2699
2700         net_srandom(ifp->addr.s6_addr32[3]);
2701
2702         read_lock_bh(&idev->lock);
2703         if (ifp->dead)
2704                 goto out;
2705         spin_lock_bh(&ifp->lock);
2706
2707         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2708             !(ifp->flags&IFA_F_TENTATIVE) ||
2709             ifp->flags & IFA_F_NODAD) {
2710                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2711                 spin_unlock_bh(&ifp->lock);
2712                 read_unlock_bh(&idev->lock);
2713
2714                 addrconf_dad_completed(ifp);
2715                 return;
2716         }
2717
2718         if (!(idev->if_flags & IF_READY)) {
2719                 spin_unlock_bh(&ifp->lock);
2720                 read_unlock_bh(&idev->lock);
2721                 /*
2722                  * If the defice is not ready:
2723                  * - keep it tentative if it is a permanent address.
2724                  * - otherwise, kill it.
2725                  */
2726                 in6_ifa_hold(ifp);
2727                 addrconf_dad_stop(ifp);
2728                 return;
2729         }
2730
2731         /*
2732          * Optimistic nodes can start receiving
2733          * Frames right away
2734          */
2735         if(ifp->flags & IFA_F_OPTIMISTIC)
2736                 ip6_ins_rt(ifp->rt);
2737
2738         addrconf_dad_kick(ifp);
2739         spin_unlock_bh(&ifp->lock);
2740 out:
2741         read_unlock_bh(&idev->lock);
2742 }
2743
2744 static void addrconf_dad_timer(unsigned long data)
2745 {
2746         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2747         struct inet6_dev *idev = ifp->idev;
2748         struct in6_addr unspec;
2749         struct in6_addr mcaddr;
2750
2751         read_lock_bh(&idev->lock);
2752         if (idev->dead) {
2753                 read_unlock_bh(&idev->lock);
2754                 goto out;
2755         }
2756         spin_lock_bh(&ifp->lock);
2757         if (ifp->probes == 0) {
2758                 /*
2759                  * DAD was successful
2760                  */
2761
2762                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2763                 spin_unlock_bh(&ifp->lock);
2764                 read_unlock_bh(&idev->lock);
2765
2766                 addrconf_dad_completed(ifp);
2767
2768                 goto out;
2769         }
2770
2771         ifp->probes--;
2772         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2773         spin_unlock_bh(&ifp->lock);
2774         read_unlock_bh(&idev->lock);
2775
2776         /* send a neighbour solicitation for our addr */
2777         memset(&unspec, 0, sizeof(unspec));
2778         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2779         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2780 out:
2781         in6_ifa_put(ifp);
2782 }
2783
2784 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2785 {
2786         struct net_device *     dev = ifp->idev->dev;
2787
2788         /*
2789          *      Configure the address for reception. Now it is valid.
2790          */
2791
2792         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2793
2794         /* If added prefix is link local and forwarding is off,
2795            start sending router solicitations.
2796          */
2797
2798         if (ifp->idev->cnf.forwarding == 0 &&
2799             ifp->idev->cnf.rtr_solicits > 0 &&
2800             (dev->flags&IFF_LOOPBACK) == 0 &&
2801             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2802                 /*
2803                  *      If a host as already performed a random delay
2804                  *      [...] as part of DAD [...] there is no need
2805                  *      to delay again before sending the first RS
2806                  */
2807                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2808
2809                 spin_lock_bh(&ifp->lock);
2810                 ifp->probes = 1;
2811                 ifp->idev->if_flags |= IF_RS_SENT;
2812                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2813                 spin_unlock_bh(&ifp->lock);
2814         }
2815 }
2816
2817 static void addrconf_dad_run(struct inet6_dev *idev) {
2818         struct inet6_ifaddr *ifp;
2819
2820         read_lock_bh(&idev->lock);
2821         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2822                 spin_lock_bh(&ifp->lock);
2823                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2824                         spin_unlock_bh(&ifp->lock);
2825                         continue;
2826                 }
2827                 spin_unlock_bh(&ifp->lock);
2828                 addrconf_dad_kick(ifp);
2829         }
2830         read_unlock_bh(&idev->lock);
2831 }
2832
2833 #ifdef CONFIG_PROC_FS
2834 struct if6_iter_state {
2835         struct seq_net_private p;
2836         int bucket;
2837 };
2838
2839 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2840 {
2841         struct inet6_ifaddr *ifa = NULL;
2842         struct if6_iter_state *state = seq->private;
2843         struct net *net = seq_file_net(seq);
2844
2845         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2846                 ifa = inet6_addr_lst[state->bucket];
2847
2848                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2849                         ifa = ifa->lst_next;
2850                 if (ifa)
2851                         break;
2852         }
2853         return ifa;
2854 }
2855
2856 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2857 {
2858         struct if6_iter_state *state = seq->private;
2859         struct net *net = seq_file_net(seq);
2860
2861         ifa = ifa->lst_next;
2862 try_again:
2863         if (ifa) {
2864                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2865                         ifa = ifa->lst_next;
2866                         goto try_again;
2867                 }
2868         }
2869
2870         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2871                 ifa = inet6_addr_lst[state->bucket];
2872                 goto try_again;
2873         }
2874
2875         return ifa;
2876 }
2877
2878 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2879 {
2880         struct inet6_ifaddr *ifa = if6_get_first(seq);
2881
2882         if (ifa)
2883                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2884                         --pos;
2885         return pos ? NULL : ifa;
2886 }
2887
2888 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2889         __acquires(addrconf_hash_lock)
2890 {
2891         read_lock_bh(&addrconf_hash_lock);
2892         return if6_get_idx(seq, *pos);
2893 }
2894
2895 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2896 {
2897         struct inet6_ifaddr *ifa;
2898
2899         ifa = if6_get_next(seq, v);
2900         ++*pos;
2901         return ifa;
2902 }
2903
2904 static void if6_seq_stop(struct seq_file *seq, void *v)
2905         __releases(addrconf_hash_lock)
2906 {
2907         read_unlock_bh(&addrconf_hash_lock);
2908 }
2909
2910 static int if6_seq_show(struct seq_file *seq, void *v)
2911 {
2912         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2913         seq_printf(seq,
2914                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2915                    NIP6(ifp->addr),
2916                    ifp->idev->dev->ifindex,
2917                    ifp->prefix_len,
2918                    ifp->scope,
2919                    ifp->flags,
2920                    ifp->idev->dev->name);
2921         return 0;
2922 }
2923
2924 static const struct seq_operations if6_seq_ops = {
2925         .start  = if6_seq_start,
2926         .next   = if6_seq_next,
2927         .show   = if6_seq_show,
2928         .stop   = if6_seq_stop,
2929 };
2930
2931 static int if6_seq_open(struct inode *inode, struct file *file)
2932 {
2933         return seq_open_net(inode, file, &if6_seq_ops,
2934                             sizeof(struct if6_iter_state));
2935 }
2936
2937 static const struct file_operations if6_fops = {
2938         .owner          = THIS_MODULE,
2939         .open           = if6_seq_open,
2940         .read           = seq_read,
2941         .llseek         = seq_lseek,
2942         .release        = seq_release_net,
2943 };
2944
2945 static int if6_proc_net_init(struct net *net)
2946 {
2947         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2948                 return -ENOMEM;
2949         return 0;
2950 }
2951
2952 static void if6_proc_net_exit(struct net *net)
2953 {
2954        proc_net_remove(net, "if_inet6");
2955 }
2956
2957 static struct pernet_operations if6_proc_net_ops = {
2958        .init = if6_proc_net_init,
2959        .exit = if6_proc_net_exit,
2960 };
2961
2962 int __init if6_proc_init(void)
2963 {
2964         return register_pernet_subsys(&if6_proc_net_ops);
2965 }
2966
2967 void if6_proc_exit(void)
2968 {
2969         unregister_pernet_subsys(&if6_proc_net_ops);
2970 }
2971 #endif  /* CONFIG_PROC_FS */
2972
2973 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2974 /* Check if address is a home address configured on any interface. */
2975 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2976 {
2977         int ret = 0;
2978         struct inet6_ifaddr * ifp;
2979         u8 hash = ipv6_addr_hash(addr);
2980         read_lock_bh(&addrconf_hash_lock);
2981         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2982                 if (!net_eq(dev_net(ifp->idev->dev), net))
2983                         continue;
2984                 if (ipv6_addr_equal(&ifp->addr, addr) &&
2985                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2986                         ret = 1;
2987                         break;
2988                 }
2989         }
2990         read_unlock_bh(&addrconf_hash_lock);
2991         return ret;
2992 }
2993 #endif
2994
2995 /*
2996  *      Periodic address status verification
2997  */
2998
2999 static void addrconf_verify(unsigned long foo)
3000 {
3001         struct inet6_ifaddr *ifp;
3002         unsigned long now, next;
3003         int i;
3004
3005         spin_lock_bh(&addrconf_verify_lock);
3006         now = jiffies;
3007         next = now + ADDR_CHECK_FREQUENCY;
3008
3009         del_timer(&addr_chk_timer);
3010
3011         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3012
3013 restart:
3014                 read_lock(&addrconf_hash_lock);
3015                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3016                         unsigned long age;
3017 #ifdef CONFIG_IPV6_PRIVACY
3018                         unsigned long regen_advance;
3019 #endif
3020
3021                         if (ifp->flags & IFA_F_PERMANENT)
3022                                 continue;
3023
3024                         spin_lock(&ifp->lock);
3025                         age = (now - ifp->tstamp) / HZ;
3026
3027 #ifdef CONFIG_IPV6_PRIVACY
3028                         regen_advance = ifp->idev->cnf.regen_max_retry *
3029                                         ifp->idev->cnf.dad_transmits *
3030                                         ifp->idev->nd_parms->retrans_time / HZ;
3031 #endif
3032
3033                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3034                             age >= ifp->valid_lft) {
3035                                 spin_unlock(&ifp->lock);
3036                                 in6_ifa_hold(ifp);
3037                                 read_unlock(&addrconf_hash_lock);
3038                                 ipv6_del_addr(ifp);
3039                                 goto restart;
3040                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3041                                 spin_unlock(&ifp->lock);
3042                                 continue;
3043                         } else if (age >= ifp->prefered_lft) {
3044                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3045                                 int deprecate = 0;
3046
3047                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3048                                         deprecate = 1;
3049                                         ifp->flags |= IFA_F_DEPRECATED;
3050                                 }
3051
3052                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3053                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3054
3055                                 spin_unlock(&ifp->lock);
3056
3057                                 if (deprecate) {
3058                                         in6_ifa_hold(ifp);
3059                                         read_unlock(&addrconf_hash_lock);
3060
3061                                         ipv6_ifa_notify(0, ifp);
3062                                         in6_ifa_put(ifp);
3063                                         goto restart;
3064                                 }
3065 #ifdef CONFIG_IPV6_PRIVACY
3066                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3067                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3068                                 if (age >= ifp->prefered_lft - regen_advance) {
3069                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3070                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3071                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3072                                         if (!ifp->regen_count && ifpub) {
3073                                                 ifp->regen_count++;
3074                                                 in6_ifa_hold(ifp);
3075                                                 in6_ifa_hold(ifpub);
3076                                                 spin_unlock(&ifp->lock);
3077                                                 read_unlock(&addrconf_hash_lock);
3078                                                 spin_lock(&ifpub->lock);
3079                                                 ifpub->regen_count = 0;
3080                                                 spin_unlock(&ifpub->lock);
3081                                                 ipv6_create_tempaddr(ifpub, ifp);
3082                                                 in6_ifa_put(ifpub);
3083                                                 in6_ifa_put(ifp);
3084                                                 goto restart;
3085                                         }
3086                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3087                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3088                                 spin_unlock(&ifp->lock);
3089 #endif
3090                         } else {
3091                                 /* ifp->prefered_lft <= ifp->valid_lft */
3092                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3093                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3094                                 spin_unlock(&ifp->lock);
3095                         }
3096                 }
3097                 read_unlock(&addrconf_hash_lock);
3098         }
3099
3100         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3101         add_timer(&addr_chk_timer);
3102         spin_unlock_bh(&addrconf_verify_lock);
3103 }
3104
3105 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3106 {
3107         struct in6_addr *pfx = NULL;
3108
3109         if (addr)
3110                 pfx = nla_data(addr);
3111
3112         if (local) {
3113                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3114                         pfx = NULL;
3115                 else
3116                         pfx = nla_data(local);
3117         }
3118
3119         return pfx;
3120 }
3121
3122 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3123         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3124         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3125         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3126 };
3127
3128 static int
3129 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3130 {
3131         struct net *net = sock_net(skb->sk);
3132         struct ifaddrmsg *ifm;
3133         struct nlattr *tb[IFA_MAX+1];
3134         struct in6_addr *pfx;
3135         int err;
3136
3137         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3138         if (err < 0)
3139                 return err;
3140
3141         ifm = nlmsg_data(nlh);
3142         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3143         if (pfx == NULL)
3144                 return -EINVAL;
3145
3146         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3147 }
3148
3149 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3150                              u32 prefered_lft, u32 valid_lft)
3151 {
3152         u32 flags = RTF_EXPIRES;
3153
3154         if (!valid_lft || (prefered_lft > valid_lft))
3155                 return -EINVAL;
3156
3157         if (valid_lft == INFINITY_LIFE_TIME) {
3158                 ifa_flags |= IFA_F_PERMANENT;
3159                 flags = 0;
3160         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3161                 valid_lft = 0x7FFFFFFF/HZ;
3162
3163         if (prefered_lft == 0)
3164                 ifa_flags |= IFA_F_DEPRECATED;
3165         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3166                  (prefered_lft != INFINITY_LIFE_TIME))
3167                 prefered_lft = 0x7FFFFFFF/HZ;
3168
3169         spin_lock_bh(&ifp->lock);
3170         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3171         ifp->tstamp = jiffies;
3172         ifp->valid_lft = valid_lft;
3173         ifp->prefered_lft = prefered_lft;
3174
3175         spin_unlock_bh(&ifp->lock);
3176         if (!(ifp->flags&IFA_F_TENTATIVE))
3177                 ipv6_ifa_notify(0, ifp);
3178
3179         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3180                               jiffies_to_clock_t(valid_lft * HZ), flags);
3181         addrconf_verify(0);
3182
3183         return 0;
3184 }
3185
3186 static int
3187 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3188 {
3189         struct net *net = sock_net(skb->sk);
3190         struct ifaddrmsg *ifm;
3191         struct nlattr *tb[IFA_MAX+1];
3192         struct in6_addr *pfx;
3193         struct inet6_ifaddr *ifa;
3194         struct net_device *dev;
3195         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3196         u8 ifa_flags;
3197         int err;
3198
3199         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3200         if (err < 0)
3201                 return err;
3202
3203         ifm = nlmsg_data(nlh);
3204         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3205         if (pfx == NULL)
3206                 return -EINVAL;
3207
3208         if (tb[IFA_CACHEINFO]) {
3209                 struct ifa_cacheinfo *ci;
3210
3211                 ci = nla_data(tb[IFA_CACHEINFO]);
3212                 valid_lft = ci->ifa_valid;
3213                 preferred_lft = ci->ifa_prefered;
3214         } else {
3215                 preferred_lft = INFINITY_LIFE_TIME;
3216                 valid_lft = INFINITY_LIFE_TIME;
3217         }
3218
3219         dev =  __dev_get_by_index(net, ifm->ifa_index);
3220         if (dev == NULL)
3221                 return -ENODEV;
3222
3223         /* We ignore other flags so far. */
3224         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3225
3226         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3227         if (ifa == NULL) {
3228                 /*
3229                  * It would be best to check for !NLM_F_CREATE here but
3230                  * userspace alreay relies on not having to provide this.
3231                  */
3232                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3233                                       ifm->ifa_prefixlen, ifa_flags,
3234                                       preferred_lft, valid_lft);
3235         }
3236
3237         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3238             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3239                 err = -EEXIST;
3240         else
3241                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3242
3243         in6_ifa_put(ifa);
3244
3245         return err;
3246 }
3247
3248 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3249                           u8 scope, int ifindex)
3250 {
3251         struct ifaddrmsg *ifm;
3252
3253         ifm = nlmsg_data(nlh);
3254         ifm->ifa_family = AF_INET6;
3255         ifm->ifa_prefixlen = prefixlen;
3256         ifm->ifa_flags = flags;
3257         ifm->ifa_scope = scope;
3258         ifm->ifa_index = ifindex;
3259 }
3260
3261 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3262                          unsigned long tstamp, u32 preferred, u32 valid)
3263 {
3264         struct ifa_cacheinfo ci;
3265
3266         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3267                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3268         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3269                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3270         ci.ifa_prefered = preferred;
3271         ci.ifa_valid = valid;
3272
3273         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3274 }
3275
3276 static inline int rt_scope(int ifa_scope)
3277 {
3278         if (ifa_scope & IFA_HOST)
3279                 return RT_SCOPE_HOST;
3280         else if (ifa_scope & IFA_LINK)
3281                 return RT_SCOPE_LINK;
3282         else if (ifa_scope & IFA_SITE)
3283                 return RT_SCOPE_SITE;
3284         else
3285                 return RT_SCOPE_UNIVERSE;
3286 }
3287
3288 static inline int inet6_ifaddr_msgsize(void)
3289 {
3290         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3291                + nla_total_size(16) /* IFA_ADDRESS */
3292                + nla_total_size(sizeof(struct ifa_cacheinfo));
3293 }
3294
3295 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3296                              u32 pid, u32 seq, int event, unsigned int flags)
3297 {
3298         struct nlmsghdr  *nlh;
3299         u32 preferred, valid;
3300
3301         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3302         if (nlh == NULL)
3303                 return -EMSGSIZE;
3304
3305         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3306                       ifa->idev->dev->ifindex);
3307
3308         if (!(ifa->flags&IFA_F_PERMANENT)) {
3309                 preferred = ifa->prefered_lft;
3310                 valid = ifa->valid_lft;
3311                 if (preferred != INFINITY_LIFE_TIME) {
3312                         long tval = (jiffies - ifa->tstamp)/HZ;
3313                         preferred -= tval;
3314                         if (valid != INFINITY_LIFE_TIME)
3315                                 valid -= tval;
3316                 }
3317         } else {
3318                 preferred = INFINITY_LIFE_TIME;
3319                 valid = INFINITY_LIFE_TIME;
3320         }
3321
3322         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3323             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3324                 nlmsg_cancel(skb, nlh);
3325                 return -EMSGSIZE;
3326         }
3327
3328         return nlmsg_end(skb, nlh);
3329 }
3330
3331 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3332                                 u32 pid, u32 seq, int event, u16 flags)
3333 {
3334         struct nlmsghdr  *nlh;
3335         u8 scope = RT_SCOPE_UNIVERSE;
3336         int ifindex = ifmca->idev->dev->ifindex;
3337
3338         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3339                 scope = RT_SCOPE_SITE;
3340
3341         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3342         if (nlh == NULL)
3343                 return -EMSGSIZE;
3344
3345         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3346         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3347             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3348                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3349                 nlmsg_cancel(skb, nlh);
3350                 return -EMSGSIZE;
3351         }
3352
3353         return nlmsg_end(skb, nlh);
3354 }
3355
3356 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3357                                 u32 pid, u32 seq, int event, unsigned int flags)
3358 {
3359         struct nlmsghdr  *nlh;
3360         u8 scope = RT_SCOPE_UNIVERSE;
3361         int ifindex = ifaca->aca_idev->dev->ifindex;
3362
3363         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3364                 scope = RT_SCOPE_SITE;
3365
3366         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3367         if (nlh == NULL)
3368                 return -EMSGSIZE;
3369
3370         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3371         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3372             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3373                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3374                 nlmsg_cancel(skb, nlh);
3375                 return -EMSGSIZE;
3376         }
3377
3378         return nlmsg_end(skb, nlh);
3379 }
3380
3381 enum addr_type_t
3382 {
3383         UNICAST_ADDR,
3384         MULTICAST_ADDR,
3385         ANYCAST_ADDR,
3386 };
3387
3388 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3389                            enum addr_type_t type)
3390 {
3391         int idx, ip_idx;
3392         int s_idx, s_ip_idx;
3393         int err = 1;
3394         struct net_device *dev;
3395         struct inet6_dev *idev = NULL;
3396         struct inet6_ifaddr *ifa;
3397         struct ifmcaddr6 *ifmca;
3398         struct ifacaddr6 *ifaca;
3399         struct net *net = sock_net(skb->sk);
3400
3401         s_idx = cb->args[0];
3402         s_ip_idx = ip_idx = cb->args[1];
3403
3404         idx = 0;
3405         for_each_netdev(net, dev) {
3406                 if (idx < s_idx)
3407                         goto cont;
3408                 if (idx > s_idx)
3409                         s_ip_idx = 0;
3410                 ip_idx = 0;
3411                 if ((idev = in6_dev_get(dev)) == NULL)
3412                         goto cont;
3413                 read_lock_bh(&idev->lock);
3414                 switch (type) {
3415                 case UNICAST_ADDR:
3416                         /* unicast address incl. temp addr */
3417                         for (ifa = idev->addr_list; ifa;
3418                              ifa = ifa->if_next, ip_idx++) {
3419                                 if (ip_idx < s_ip_idx)
3420                                         continue;
3421                                 err = inet6_fill_ifaddr(skb, ifa,
3422                                                         NETLINK_CB(cb->skb).pid,
3423                                                         cb->nlh->nlmsg_seq,
3424                                                         RTM_NEWADDR,
3425                                                         NLM_F_MULTI);
3426                         }
3427                         break;
3428                 case MULTICAST_ADDR:
3429                         /* multicast address */
3430                         for (ifmca = idev->mc_list; ifmca;
3431                              ifmca = ifmca->next, ip_idx++) {
3432                                 if (ip_idx < s_ip_idx)
3433                                         continue;
3434                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3435                                                           NETLINK_CB(cb->skb).pid,
3436                                                           cb->nlh->nlmsg_seq,
3437                                                           RTM_GETMULTICAST,
3438                                                           NLM_F_MULTI);
3439                         }
3440                         break;
3441                 case ANYCAST_ADDR:
3442                         /* anycast address */
3443                         for (ifaca = idev->ac_list; ifaca;
3444                              ifaca = ifaca->aca_next, ip_idx++) {
3445                                 if (ip_idx < s_ip_idx)
3446                                         continue;
3447                                 err = inet6_fill_ifacaddr(skb, ifaca,
3448                                                           NETLINK_CB(cb->skb).pid,
3449                                                           cb->nlh->nlmsg_seq,
3450                                                           RTM_GETANYCAST,
3451                                                           NLM_F_MULTI);
3452                         }
3453                         break;
3454                 default:
3455                         break;
3456                 }
3457                 read_unlock_bh(&idev->lock);
3458                 in6_dev_put(idev);
3459
3460                 if (err <= 0)
3461                         break;
3462 cont:
3463                 idx++;
3464         }
3465         cb->args[0] = idx;
3466         cb->args[1] = ip_idx;
3467         return skb->len;
3468 }
3469
3470 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3471 {
3472         enum addr_type_t type = UNICAST_ADDR;
3473
3474         return inet6_dump_addr(skb, cb, type);
3475 }
3476
3477 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3478 {
3479         enum addr_type_t type = MULTICAST_ADDR;
3480
3481         return inet6_dump_addr(skb, cb, type);
3482 }
3483
3484
3485 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3486 {
3487         enum addr_type_t type = ANYCAST_ADDR;
3488
3489         return inet6_dump_addr(skb, cb, type);
3490 }
3491
3492 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3493                              void *arg)
3494 {
3495         struct net *net = sock_net(in_skb->sk);
3496         struct ifaddrmsg *ifm;
3497         struct nlattr *tb[IFA_MAX+1];
3498         struct in6_addr *addr = NULL;
3499         struct net_device *dev = NULL;
3500         struct inet6_ifaddr *ifa;
3501         struct sk_buff *skb;
3502         int err;
3503
3504         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3505         if (err < 0)
3506                 goto errout;
3507
3508         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3509         if (addr == NULL) {
3510                 err = -EINVAL;
3511                 goto errout;
3512         }
3513
3514         ifm = nlmsg_data(nlh);
3515         if (ifm->ifa_index)
3516                 dev = __dev_get_by_index(net, ifm->ifa_index);
3517
3518         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3519                 err = -EADDRNOTAVAIL;
3520                 goto errout;
3521         }
3522
3523         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3524                 err = -ENOBUFS;
3525                 goto errout_ifa;
3526         }
3527
3528         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3529                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3530         if (err < 0) {
3531                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3532                 WARN_ON(err == -EMSGSIZE);
3533                 kfree_skb(skb);
3534                 goto errout_ifa;
3535         }
3536         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3537 errout_ifa:
3538         in6_ifa_put(ifa);
3539 errout:
3540         return err;
3541 }
3542
3543 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3544 {
3545         struct sk_buff *skb;
3546         struct net *net = dev_net(ifa->idev->dev);
3547         int err = -ENOBUFS;
3548
3549         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3550         if (skb == NULL)
3551                 goto errout;
3552
3553         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3554         if (err < 0) {
3555                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3556                 WARN_ON(err == -EMSGSIZE);
3557                 kfree_skb(skb);
3558                 goto errout;
3559         }
3560         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3561 errout:
3562         if (err < 0)
3563                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3564 }
3565
3566 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3567                                 __s32 *array, int bytes)
3568 {
3569         BUG_ON(bytes < (DEVCONF_MAX * 4));
3570
3571         memset(array, 0, bytes);
3572         array[DEVCONF_FORWARDING] = cnf->forwarding;
3573         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3574         array[DEVCONF_MTU6] = cnf->mtu6;
3575         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3576         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3577         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3578         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3579         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3580         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3581         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3582         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3583 #ifdef CONFIG_IPV6_PRIVACY
3584         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3585         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3586         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3587         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3588         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3589 #endif
3590         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3591         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3592         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3593 #ifdef CONFIG_IPV6_ROUTER_PREF
3594         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3595         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3596 #ifdef CONFIG_IPV6_ROUTE_INFO
3597         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3598 #endif
3599 #endif
3600         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3601         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3602 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3603         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3604 #endif
3605 #ifdef CONFIG_IPV6_MROUTE
3606         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3607 #endif
3608 }
3609
3610 static inline size_t inet6_if_nlmsg_size(void)
3611 {
3612         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3613                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3614                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3615                + nla_total_size(4) /* IFLA_MTU */
3616                + nla_total_size(4) /* IFLA_LINK */
3617                + nla_total_size( /* IFLA_PROTINFO */
3618                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3619                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3620                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3621                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3622                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3623                  );
3624 }
3625
3626 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3627                                       int bytes)
3628 {
3629         int i;
3630         int pad = bytes - sizeof(u64) * items;
3631         BUG_ON(pad < 0);
3632
3633         /* Use put_unaligned() because stats may not be aligned for u64. */
3634         put_unaligned(items, &stats[0]);
3635         for (i = 1; i < items; i++)
3636                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3637
3638         memset(&stats[items], 0, pad);
3639 }
3640
3641 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3642                              int bytes)
3643 {
3644         switch(attrtype) {
3645         case IFLA_INET6_STATS:
3646                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3647                 break;
3648         case IFLA_INET6_ICMP6STATS:
3649                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3650                 break;
3651         }
3652 }
3653
3654 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3655                              u32 pid, u32 seq, int event, unsigned int flags)
3656 {
3657         struct net_device *dev = idev->dev;
3658         struct nlattr *nla;
3659         struct ifinfomsg *hdr;
3660         struct nlmsghdr *nlh;
3661         void *protoinfo;
3662         struct ifla_cacheinfo ci;
3663
3664         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3665         if (nlh == NULL)
3666                 return -EMSGSIZE;
3667
3668         hdr = nlmsg_data(nlh);
3669         hdr->ifi_family = AF_INET6;
3670         hdr->__ifi_pad = 0;
3671         hdr->ifi_type = dev->type;
3672         hdr->ifi_index = dev->ifindex;
3673         hdr->ifi_flags = dev_get_flags(dev);
3674         hdr->ifi_change = 0;
3675
3676         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3677
3678         if (dev->addr_len)
3679                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3680
3681         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3682         if (dev->ifindex != dev->iflink)
3683                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3684
3685         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3686         if (protoinfo == NULL)
3687                 goto nla_put_failure;
3688
3689         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3690
3691         ci.max_reasm_len = IPV6_MAXPLEN;
3692         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3693                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3694         ci.reachable_time = idev->nd_parms->reachable_time;
3695         ci.retrans_time = idev->nd_parms->retrans_time;
3696         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3697
3698         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3699         if (nla == NULL)
3700                 goto nla_put_failure;
3701         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3702
3703         /* XXX - MC not implemented */
3704
3705         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3706         if (nla == NULL)
3707                 goto nla_put_failure;
3708         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3709
3710         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3711         if (nla == NULL)
3712                 goto nla_put_failure;
3713         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3714
3715         nla_nest_end(skb, protoinfo);
3716         return nlmsg_end(skb, nlh);
3717
3718 nla_put_failure:
3719         nlmsg_cancel(skb, nlh);
3720         return -EMSGSIZE;
3721 }
3722
3723 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3724 {
3725         struct net *net = sock_net(skb->sk);
3726         int idx, err;
3727         int s_idx = cb->args[0];
3728         struct net_device *dev;
3729         struct inet6_dev *idev;
3730
3731         read_lock(&dev_base_lock);
3732         idx = 0;
3733         for_each_netdev(net, dev) {
3734                 if (idx < s_idx)
3735                         goto cont;
3736                 if ((idev = in6_dev_get(dev)) == NULL)
3737                         goto cont;
3738                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3739                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3740                 in6_dev_put(idev);
3741                 if (err <= 0)
3742                         break;
3743 cont:
3744                 idx++;
3745         }
3746         read_unlock(&dev_base_lock);
3747         cb->args[0] = idx;
3748
3749         return skb->len;
3750 }
3751
3752 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3753 {
3754         struct sk_buff *skb;
3755         struct net *net = dev_net(idev->dev);
3756         int err = -ENOBUFS;
3757
3758         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3759         if (skb == NULL)
3760                 goto errout;
3761
3762         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3763         if (err < 0) {
3764                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3765                 WARN_ON(err == -EMSGSIZE);
3766                 kfree_skb(skb);
3767                 goto errout;
3768         }
3769         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3770 errout:
3771         if (err < 0)
3772                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3773 }
3774
3775 static inline size_t inet6_prefix_nlmsg_size(void)
3776 {
3777         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3778                + nla_total_size(sizeof(struct in6_addr))
3779                + nla_total_size(sizeof(struct prefix_cacheinfo));
3780 }
3781
3782 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3783                              struct prefix_info *pinfo, u32 pid, u32 seq,
3784                              int event, unsigned int flags)
3785 {
3786         struct prefixmsg *pmsg;
3787         struct nlmsghdr *nlh;
3788         struct prefix_cacheinfo ci;
3789
3790         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3791         if (nlh == NULL)
3792                 return -EMSGSIZE;
3793
3794         pmsg = nlmsg_data(nlh);
3795         pmsg->prefix_family = AF_INET6;
3796         pmsg->prefix_pad1 = 0;
3797         pmsg->prefix_pad2 = 0;
3798         pmsg->prefix_ifindex = idev->dev->ifindex;
3799         pmsg->prefix_len = pinfo->prefix_len;
3800         pmsg->prefix_type = pinfo->type;
3801         pmsg->prefix_pad3 = 0;
3802         pmsg->prefix_flags = 0;
3803         if (pinfo->onlink)
3804                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3805         if (pinfo->autoconf)
3806                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3807
3808         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3809
3810         ci.preferred_time = ntohl(pinfo->prefered);
3811         ci.valid_time = ntohl(pinfo->valid);
3812         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3813
3814         return nlmsg_end(skb, nlh);
3815
3816 nla_put_failure:
3817         nlmsg_cancel(skb, nlh);
3818         return -EMSGSIZE;
3819 }
3820
3821 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3822                          struct prefix_info *pinfo)
3823 {
3824         struct sk_buff *skb;
3825         struct net *net = dev_net(idev->dev);
3826         int err = -ENOBUFS;
3827
3828         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3829         if (skb == NULL)
3830                 goto errout;
3831
3832         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3833         if (err < 0) {
3834                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3835                 WARN_ON(err == -EMSGSIZE);
3836                 kfree_skb(skb);
3837                 goto errout;
3838         }
3839         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3840 errout:
3841         if (err < 0)
3842                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3843 }
3844
3845 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3846 {
3847         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3848
3849         switch (event) {
3850         case RTM_NEWADDR:
3851                 /*
3852                  * If the address was optimistic
3853                  * we inserted the route at the start of
3854                  * our DAD process, so we don't need
3855                  * to do it again
3856                  */
3857                 if (!(ifp->rt->rt6i_node))
3858                         ip6_ins_rt(ifp->rt);
3859                 if (ifp->idev->cnf.forwarding)
3860                         addrconf_join_anycast(ifp);
3861                 break;
3862         case RTM_DELADDR:
3863                 if (ifp->idev->cnf.forwarding)
3864                         addrconf_leave_anycast(ifp);
3865                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3866                 dst_hold(&ifp->rt->u.dst);
3867                 if (ip6_del_rt(ifp->rt))
3868                         dst_free(&ifp->rt->u.dst);
3869                 break;
3870         }
3871 }
3872
3873 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3874 {
3875         rcu_read_lock_bh();
3876         if (likely(ifp->idev->dead == 0))
3877                 __ipv6_ifa_notify(event, ifp);
3878         rcu_read_unlock_bh();
3879 }
3880
3881 #ifdef CONFIG_SYSCTL
3882
3883 static
3884 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3885                            void __user *buffer, size_t *lenp, loff_t *ppos)
3886 {
3887         int *valp = ctl->data;
3888         int val = *valp;
3889         int ret;
3890
3891         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3892
3893         if (write)
3894                 addrconf_fixup_forwarding(ctl, valp, val);
3895         return ret;
3896 }
3897
3898 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3899                                             int __user *name, int nlen,
3900                                             void __user *oldval,
3901                                             size_t __user *oldlenp,
3902                                             void __user *newval, size_t newlen)
3903 {
3904         int *valp = table->data;
3905         int val = *valp;
3906         int new;
3907
3908         if (!newval || !newlen)
3909                 return 0;
3910         if (newlen != sizeof(int))
3911                 return -EINVAL;
3912         if (get_user(new, (int __user *)newval))
3913                 return -EFAULT;
3914         if (new == *valp)
3915                 return 0;
3916         if (oldval && oldlenp) {
3917                 size_t len;
3918                 if (get_user(len, oldlenp))
3919                         return -EFAULT;
3920                 if (len) {
3921                         if (len > table->maxlen)
3922                                 len = table->maxlen;
3923                         if (copy_to_user(oldval, valp, len))
3924                                 return -EFAULT;
3925                         if (put_user(len, oldlenp))
3926                                 return -EFAULT;
3927                 }
3928         }
3929
3930         *valp = new;
3931         addrconf_fixup_forwarding(table, valp, val);
3932         return 1;
3933 }
3934
3935 static struct addrconf_sysctl_table
3936 {
3937         struct ctl_table_header *sysctl_header;
3938         ctl_table addrconf_vars[DEVCONF_MAX+1];
3939         char *dev_name;
3940 } addrconf_sysctl __read_mostly = {
3941         .sysctl_header = NULL,
3942         .addrconf_vars = {
3943                 {
3944                         .ctl_name       =       NET_IPV6_FORWARDING,
3945                         .procname       =       "forwarding",
3946                         .data           =       &ipv6_devconf.forwarding,
3947                         .maxlen         =       sizeof(int),
3948                         .mode           =       0644,
3949                         .proc_handler   =       &addrconf_sysctl_forward,
3950                         .strategy       =       &addrconf_sysctl_forward_strategy,
3951                 },
3952                 {
3953                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3954                         .procname       =       "hop_limit",
3955                         .data           =       &ipv6_devconf.hop_limit,
3956                         .maxlen         =       sizeof(int),
3957                         .mode           =       0644,
3958                         .proc_handler   =       proc_dointvec,
3959                 },
3960                 {
3961                         .ctl_name       =       NET_IPV6_MTU,
3962                         .procname       =       "mtu",
3963                         .data           =       &ipv6_devconf.mtu6,
3964                         .maxlen         =       sizeof(int),
3965                         .mode           =       0644,
3966                         .proc_handler   =       &proc_dointvec,
3967                 },
3968                 {
3969                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3970                         .procname       =       "accept_ra",
3971                         .data           =       &ipv6_devconf.accept_ra,
3972                         .maxlen         =       sizeof(int),
3973                         .mode           =       0644,
3974                         .proc_handler   =       &proc_dointvec,
3975                 },
3976                 {
3977                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3978                         .procname       =       "accept_redirects",
3979                         .data           =       &ipv6_devconf.accept_redirects,
3980                         .maxlen         =       sizeof(int),
3981                         .mode           =       0644,
3982                         .proc_handler   =       &proc_dointvec,
3983                 },
3984                 {
3985                         .ctl_name       =       NET_IPV6_AUTOCONF,
3986                         .procname       =       "autoconf",
3987                         .data           =       &ipv6_devconf.autoconf,
3988                         .maxlen         =       sizeof(int),
3989                         .mode           =       0644,
3990                         .proc_handler   =       &proc_dointvec,
3991                 },
3992                 {
3993                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3994                         .procname       =       "dad_transmits",
3995                         .data           =       &ipv6_devconf.dad_transmits,
3996                         .maxlen         =       sizeof(int),
3997                         .mode           =       0644,
3998                         .proc_handler   =       &proc_dointvec,
3999                 },
4000                 {
4001                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4002                         .procname       =       "router_solicitations",
4003                         .data           =       &ipv6_devconf.rtr_solicits,
4004                         .maxlen         =       sizeof(int),
4005                         .mode           =       0644,
4006                         .proc_handler   =       &proc_dointvec,
4007                 },
4008                 {
4009                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4010                         .procname       =       "router_solicitation_interval",
4011                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4012                         .maxlen         =       sizeof(int),
4013                         .mode           =       0644,
4014                         .proc_handler   =       &proc_dointvec_jiffies,
4015                         .strategy       =       &sysctl_jiffies,
4016                 },
4017                 {
4018                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4019                         .procname       =       "router_solicitation_delay",
4020                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4021                         .maxlen         =       sizeof(int),
4022                         .mode           =       0644,
4023                         .proc_handler   =       &proc_dointvec_jiffies,
4024                         .strategy       =       &sysctl_jiffies,
4025                 },
4026                 {
4027                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4028                         .procname       =       "force_mld_version",
4029                         .data           =       &ipv6_devconf.force_mld_version,
4030                         .maxlen         =       sizeof(int),
4031                         .mode           =       0644,
4032                         .proc_handler   =       &proc_dointvec,
4033                 },
4034 #ifdef CONFIG_IPV6_PRIVACY
4035                 {
4036                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4037                         .procname       =       "use_tempaddr",
4038                         .data           =       &ipv6_devconf.use_tempaddr,
4039                         .maxlen         =       sizeof(int),
4040                         .mode           =       0644,
4041                         .proc_handler   =       &proc_dointvec,
4042                 },
4043                 {
4044                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4045                         .procname       =       "temp_valid_lft",
4046                         .data           =       &ipv6_devconf.temp_valid_lft,
4047                         .maxlen         =       sizeof(int),
4048                         .mode           =       0644,
4049                         .proc_handler   =       &proc_dointvec,
4050                 },
4051                 {
4052                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4053                         .procname       =       "temp_prefered_lft",
4054                         .data           =       &ipv6_devconf.temp_prefered_lft,
4055                         .maxlen         =       sizeof(int),
4056                         .mode           =       0644,
4057                         .proc_handler   =       &proc_dointvec,
4058                 },
4059                 {
4060                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4061                         .procname       =       "regen_max_retry",
4062                         .data           =       &ipv6_devconf.regen_max_retry,
4063                         .maxlen         =       sizeof(int),
4064                         .mode           =       0644,
4065                         .proc_handler   =       &proc_dointvec,
4066                 },
4067                 {
4068                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4069                         .procname       =       "max_desync_factor",
4070                         .data           =       &ipv6_devconf.max_desync_factor,
4071                         .maxlen         =       sizeof(int),
4072                         .mode           =       0644,
4073                         .proc_handler   =       &proc_dointvec,
4074                 },
4075 #endif
4076                 {
4077                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4078                         .procname       =       "max_addresses",
4079                         .data           =       &ipv6_devconf.max_addresses,
4080                         .maxlen         =       sizeof(int),
4081                         .mode           =       0644,
4082                         .proc_handler   =       &proc_dointvec,
4083                 },
4084                 {
4085                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4086                         .procname       =       "accept_ra_defrtr",
4087                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4088                         .maxlen         =       sizeof(int),
4089                         .mode           =       0644,
4090                         .proc_handler   =       &proc_dointvec,
4091                 },
4092                 {
4093                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4094                         .procname       =       "accept_ra_pinfo",
4095                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4096                         .maxlen         =       sizeof(int),
4097                         .mode           =       0644,
4098                         .proc_handler   =       &proc_dointvec,
4099                 },
4100 #ifdef CONFIG_IPV6_ROUTER_PREF
4101                 {
4102                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4103                         .procname       =       "accept_ra_rtr_pref",
4104                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4105                         .maxlen         =       sizeof(int),
4106                         .mode           =       0644,
4107                         .proc_handler   =       &proc_dointvec,
4108                 },
4109                 {
4110                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4111                         .procname       =       "router_probe_interval",
4112                         .data           =       &ipv6_devconf.rtr_probe_interval,
4113                         .maxlen         =       sizeof(int),
4114                         .mode           =       0644,
4115                         .proc_handler   =       &proc_dointvec_jiffies,
4116                         .strategy       =       &sysctl_jiffies,
4117                 },
4118 #ifdef CONFIG_IPV6_ROUTE_INFO
4119                 {
4120                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4121                         .procname       =       "accept_ra_rt_info_max_plen",
4122                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4123                         .maxlen         =       sizeof(int),
4124                         .mode           =       0644,
4125                         .proc_handler   =       &proc_dointvec,
4126                 },
4127 #endif
4128 #endif
4129                 {
4130                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4131                         .procname       =       "proxy_ndp",
4132                         .data           =       &ipv6_devconf.proxy_ndp,
4133                         .maxlen         =       sizeof(int),
4134                         .mode           =       0644,
4135                         .proc_handler   =       &proc_dointvec,
4136                 },
4137                 {
4138                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4139                         .procname       =       "accept_source_route",
4140                         .data           =       &ipv6_devconf.accept_source_route,
4141                         .maxlen         =       sizeof(int),
4142                         .mode           =       0644,
4143                         .proc_handler   =       &proc_dointvec,
4144                 },
4145 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4146                 {
4147                         .ctl_name       =       CTL_UNNUMBERED,
4148                         .procname       =       "optimistic_dad",
4149                         .data           =       &ipv6_devconf.optimistic_dad,
4150                         .maxlen         =       sizeof(int),
4151                         .mode           =       0644,
4152                         .proc_handler   =       &proc_dointvec,
4153
4154                 },
4155 #endif
4156 #ifdef CONFIG_IPV6_MROUTE
4157                 {
4158                         .ctl_name       =       CTL_UNNUMBERED,
4159                         .procname       =       "mc_forwarding",
4160                         .data           =       &ipv6_devconf.mc_forwarding,
4161                         .maxlen         =       sizeof(int),
4162                         .mode           =       0644,
4163                         .proc_handler   =       &proc_dointvec,
4164                 },
4165 #endif
4166                 {
4167                         .ctl_name       =       0,      /* sentinel */
4168                 }
4169         },
4170 };
4171
4172 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4173                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4174 {
4175         int i;
4176         struct addrconf_sysctl_table *t;
4177
4178 #define ADDRCONF_CTL_PATH_DEV   3
4179
4180         struct ctl_path addrconf_ctl_path[] = {
4181                 { .procname = "net", .ctl_name = CTL_NET, },
4182                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4183                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4184                 { /* to be set */ },
4185                 { },
4186         };
4187
4188
4189         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4190         if (t == NULL)
4191                 goto out;
4192
4193         for (i=0; t->addrconf_vars[i].data; i++) {
4194                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4195                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4196                 t->addrconf_vars[i].extra2 = net;
4197         }
4198
4199         /*
4200          * Make a copy of dev_name, because '.procname' is regarded as const
4201          * by sysctl and we wouldn't want anyone to change it under our feet
4202          * (see SIOCSIFNAME).
4203          */
4204         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4205         if (!t->dev_name)
4206                 goto free;
4207
4208         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4209         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4210
4211         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4212                         t->addrconf_vars);
4213         if (t->sysctl_header == NULL)
4214                 goto free_procname;
4215
4216         p->sysctl = t;
4217         return 0;
4218
4219 free_procname:
4220         kfree(t->dev_name);
4221 free:
4222         kfree(t);
4223 out:
4224         return -ENOBUFS;
4225 }
4226
4227 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4228 {
4229         struct addrconf_sysctl_table *t;
4230
4231         if (p->sysctl == NULL)
4232                 return;
4233
4234         t = p->sysctl;
4235         p->sysctl = NULL;
4236         unregister_sysctl_table(t->sysctl_header);
4237         kfree(t->dev_name);
4238         kfree(t);
4239 }
4240
4241 static void addrconf_sysctl_register(struct inet6_dev *idev)
4242 {
4243         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4244                               NET_IPV6_NEIGH, "ipv6",
4245                               &ndisc_ifinfo_sysctl_change,
4246                               NULL);
4247         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4248                         idev->dev->ifindex, idev, &idev->cnf);
4249 }
4250
4251 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4252 {
4253         __addrconf_sysctl_unregister(&idev->cnf);
4254         neigh_sysctl_unregister(idev->nd_parms);
4255 }
4256
4257
4258 #endif
4259
4260 static int addrconf_init_net(struct net *net)
4261 {
4262         int err;
4263         struct ipv6_devconf *all, *dflt;
4264
4265         err = -ENOMEM;
4266         all = &ipv6_devconf;
4267         dflt = &ipv6_devconf_dflt;
4268
4269         if (net != &init_net) {
4270                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4271                 if (all == NULL)
4272                         goto err_alloc_all;
4273
4274                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4275                 if (dflt == NULL)
4276                         goto err_alloc_dflt;
4277         }
4278
4279         net->ipv6.devconf_all = all;
4280         net->ipv6.devconf_dflt = dflt;
4281
4282 #ifdef CONFIG_SYSCTL
4283         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4284                         NULL, all);
4285         if (err < 0)
4286                 goto err_reg_all;
4287
4288         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4289                         NULL, dflt);
4290         if (err < 0)
4291                 goto err_reg_dflt;
4292 #endif
4293         return 0;
4294
4295 #ifdef CONFIG_SYSCTL
4296 err_reg_dflt:
4297         __addrconf_sysctl_unregister(all);
4298 err_reg_all:
4299         kfree(dflt);
4300 #endif
4301 err_alloc_dflt:
4302         kfree(all);
4303 err_alloc_all:
4304         return err;
4305 }
4306
4307 static void addrconf_exit_net(struct net *net)
4308 {
4309 #ifdef CONFIG_SYSCTL
4310         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4311         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4312 #endif
4313         if (net != &init_net) {
4314                 kfree(net->ipv6.devconf_dflt);
4315                 kfree(net->ipv6.devconf_all);
4316         }
4317 }
4318
4319 static struct pernet_operations addrconf_ops = {
4320         .init = addrconf_init_net,
4321         .exit = addrconf_exit_net,
4322 };
4323
4324 /*
4325  *      Device notifier
4326  */
4327
4328 int register_inet6addr_notifier(struct notifier_block *nb)
4329 {
4330         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4331 }
4332
4333 EXPORT_SYMBOL(register_inet6addr_notifier);
4334
4335 int unregister_inet6addr_notifier(struct notifier_block *nb)
4336 {
4337         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4338 }
4339
4340 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4341
4342
4343 static int addrconf_net_init(struct net *net)
4344 {
4345         return 0;
4346 }
4347
4348 static void addrconf_net_exit(struct net *net)
4349 {
4350         struct net_device *dev;
4351
4352         rtnl_lock();
4353         /* clean dev list */
4354         for_each_netdev(net, dev) {
4355                 if (__in6_dev_get(dev) == NULL)
4356                         continue;
4357                 addrconf_ifdown(dev, 1);
4358         }
4359         addrconf_ifdown(net->loopback_dev, 2);
4360         rtnl_unlock();
4361 }
4362
4363 static struct pernet_operations addrconf_net_ops = {
4364         .init = addrconf_net_init,
4365         .exit = addrconf_net_exit,
4366 };
4367
4368 /*
4369  *      Init / cleanup code
4370  */
4371
4372 int __init addrconf_init(void)
4373 {
4374         int err;
4375
4376         if ((err = ipv6_addr_label_init()) < 0) {
4377                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4378                         err);
4379                 return err;
4380         }
4381
4382         register_pernet_subsys(&addrconf_ops);
4383
4384         /* The addrconf netdev notifier requires that loopback_dev
4385          * has it's ipv6 private information allocated and setup
4386          * before it can bring up and give link-local addresses
4387          * to other devices which are up.
4388          *
4389          * Unfortunately, loopback_dev is not necessarily the first
4390          * entry in the global dev_base list of net devices.  In fact,
4391          * it is likely to be the very last entry on that list.
4392          * So this causes the notifier registry below to try and
4393          * give link-local addresses to all devices besides loopback_dev
4394          * first, then loopback_dev, which cases all the non-loopback_dev
4395          * devices to fail to get a link-local address.
4396          *
4397          * So, as a temporary fix, allocate the ipv6 structure for
4398          * loopback_dev first by hand.
4399          * Longer term, all of the dependencies ipv6 has upon the loopback
4400          * device and it being up should be removed.
4401          */
4402         rtnl_lock();
4403         if (!ipv6_add_dev(init_net.loopback_dev))
4404                 err = -ENOMEM;
4405         rtnl_unlock();
4406         if (err)
4407                 goto errlo;
4408
4409         err = register_pernet_device(&addrconf_net_ops);
4410         if (err)
4411                 return err;
4412
4413         register_netdevice_notifier(&ipv6_dev_notf);
4414
4415         addrconf_verify(0);
4416
4417         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4418         if (err < 0)
4419                 goto errout;
4420
4421         /* Only the first call to __rtnl_register can fail */
4422         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4423         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4424         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4425         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4426         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4427
4428         ipv6_addr_label_rtnl_register();
4429
4430         return 0;
4431 errout:
4432         unregister_netdevice_notifier(&ipv6_dev_notf);
4433 errlo:
4434         unregister_pernet_subsys(&addrconf_ops);
4435
4436         return err;
4437 }
4438
4439 void addrconf_cleanup(void)
4440 {
4441         struct inet6_ifaddr *ifa;
4442         int i;
4443
4444         unregister_netdevice_notifier(&ipv6_dev_notf);
4445         unregister_pernet_device(&addrconf_net_ops);
4446
4447         unregister_pernet_subsys(&addrconf_ops);
4448
4449         rtnl_lock();
4450
4451         /*
4452          *      Check hash table.
4453          */
4454         write_lock_bh(&addrconf_hash_lock);
4455         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4456                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4457                         struct inet6_ifaddr *bifa;
4458
4459                         bifa = ifa;
4460                         ifa = ifa->lst_next;
4461                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4462                         /* Do not free it; something is wrong.
4463                            Now we can investigate it with debugger.
4464                          */
4465                 }
4466         }
4467         write_unlock_bh(&addrconf_hash_lock);
4468
4469         del_timer(&addr_chk_timer);
4470         rtnl_unlock();
4471
4472         unregister_pernet_subsys(&addrconf_net_ops);
4473 }