Merge branch 'master' of github.com:davem330/net
[linux-2.6.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.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 <linux/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
102 static inline u32 cstamp_delta(unsigned long cstamp)
103 {
104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
105 }
106
107 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
108 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
109 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
110
111 #ifdef CONFIG_SYSCTL
112 static void addrconf_sysctl_register(struct inet6_dev *idev);
113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
114 #else
115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
116 {
117 }
118
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
120 {
121 }
122 #endif
123
124 #ifdef CONFIG_IPV6_PRIVACY
125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
127 static void ipv6_regen_rndid(unsigned long data);
128 #endif
129
130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
131 static int ipv6_count_addresses(struct inet6_dev *idev);
132
133 /*
134  *      Configured unicast address hash table
135  */
136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
137 static DEFINE_SPINLOCK(addrconf_hash_lock);
138
139 static void addrconf_verify(unsigned long);
140
141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
142 static DEFINE_SPINLOCK(addrconf_verify_lock);
143
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146
147 static void addrconf_type_change(struct net_device *dev,
148                                  unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
150
151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
152 static void addrconf_dad_timer(unsigned long data);
153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
154 static void addrconf_dad_run(struct inet6_dev *idev);
155 static void addrconf_rs_timer(unsigned long data);
156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158
159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
160                                 struct prefix_info *pinfo);
161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
162                                struct net_device *dev);
163
164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
165
166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
167         .forwarding             = 0,
168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
169         .mtu6                   = IPV6_MIN_MTU,
170         .accept_ra              = 1,
171         .accept_redirects       = 1,
172         .autoconf               = 1,
173         .force_mld_version      = 0,
174         .dad_transmits          = 1,
175         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
176         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
177         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
178 #ifdef CONFIG_IPV6_PRIVACY
179         .use_tempaddr           = 0,
180         .temp_valid_lft         = TEMP_VALID_LIFETIME,
181         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
182         .regen_max_retry        = REGEN_MAX_RETRY,
183         .max_desync_factor      = MAX_DESYNC_FACTOR,
184 #endif
185         .max_addresses          = IPV6_MAX_ADDRESSES,
186         .accept_ra_defrtr       = 1,
187         .accept_ra_pinfo        = 1,
188 #ifdef CONFIG_IPV6_ROUTER_PREF
189         .accept_ra_rtr_pref     = 1,
190         .rtr_probe_interval     = 60 * HZ,
191 #ifdef CONFIG_IPV6_ROUTE_INFO
192         .accept_ra_rt_info_max_plen = 0,
193 #endif
194 #endif
195         .proxy_ndp              = 0,
196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
197         .disable_ipv6           = 0,
198         .accept_dad             = 1,
199 };
200
201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
202         .forwarding             = 0,
203         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
204         .mtu6                   = IPV6_MIN_MTU,
205         .accept_ra              = 1,
206         .accept_redirects       = 1,
207         .autoconf               = 1,
208         .dad_transmits          = 1,
209         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
210         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
212 #ifdef CONFIG_IPV6_PRIVACY
213         .use_tempaddr           = 0,
214         .temp_valid_lft         = TEMP_VALID_LIFETIME,
215         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
216         .regen_max_retry        = REGEN_MAX_RETRY,
217         .max_desync_factor      = MAX_DESYNC_FACTOR,
218 #endif
219         .max_addresses          = IPV6_MAX_ADDRESSES,
220         .accept_ra_defrtr       = 1,
221         .accept_ra_pinfo        = 1,
222 #ifdef CONFIG_IPV6_ROUTER_PREF
223         .accept_ra_rtr_pref     = 1,
224         .rtr_probe_interval     = 60 * HZ,
225 #ifdef CONFIG_IPV6_ROUTE_INFO
226         .accept_ra_rt_info_max_plen = 0,
227 #endif
228 #endif
229         .proxy_ndp              = 0,
230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
231         .disable_ipv6           = 0,
232         .accept_dad             = 1,
233 };
234
235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
240
241 /* Check if a valid qdisc is available */
242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
243 {
244         return !qdisc_tx_is_noop(dev);
245 }
246
247 /* Check if a route is valid prefix route */
248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
249 {
250         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
251 }
252
253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
254 {
255         if (del_timer(&ifp->timer))
256                 __in6_ifa_put(ifp);
257 }
258
259 enum addrconf_timer_t {
260         AC_NONE,
261         AC_DAD,
262         AC_RS,
263 };
264
265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
266                                enum addrconf_timer_t what,
267                                unsigned long when)
268 {
269         if (!del_timer(&ifp->timer))
270                 in6_ifa_hold(ifp);
271
272         switch (what) {
273         case AC_DAD:
274                 ifp->timer.function = addrconf_dad_timer;
275                 break;
276         case AC_RS:
277                 ifp->timer.function = addrconf_rs_timer;
278                 break;
279         default:
280                 break;
281         }
282         ifp->timer.expires = jiffies + when;
283         add_timer(&ifp->timer);
284 }
285
286 static int snmp6_alloc_dev(struct inet6_dev *idev)
287 {
288         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
289                           sizeof(struct ipstats_mib),
290                           __alignof__(struct ipstats_mib)) < 0)
291                 goto err_ip;
292         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
293                                         GFP_KERNEL);
294         if (!idev->stats.icmpv6dev)
295                 goto err_icmp;
296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
297                                            GFP_KERNEL);
298         if (!idev->stats.icmpv6msgdev)
299                 goto err_icmpmsg;
300
301         return 0;
302
303 err_icmpmsg:
304         kfree(idev->stats.icmpv6dev);
305 err_icmp:
306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
307 err_ip:
308         return -ENOMEM;
309 }
310
311 static void snmp6_free_dev(struct inet6_dev *idev)
312 {
313         kfree(idev->stats.icmpv6msgdev);
314         kfree(idev->stats.icmpv6dev);
315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
316 }
317
318 /* Nobody refers to this device, we may destroy it. */
319
320 void in6_dev_finish_destroy(struct inet6_dev *idev)
321 {
322         struct net_device *dev = idev->dev;
323
324         WARN_ON(!list_empty(&idev->addr_list));
325         WARN_ON(idev->mc_list != NULL);
326
327 #ifdef NET_REFCNT_DEBUG
328         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
329 #endif
330         dev_put(dev);
331         if (!idev->dead) {
332                 pr_warning("Freeing alive inet6 device %p\n", idev);
333                 return;
334         }
335         snmp6_free_dev(idev);
336         kfree_rcu(idev, rcu);
337 }
338
339 EXPORT_SYMBOL(in6_dev_finish_destroy);
340
341 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
342 {
343         struct inet6_dev *ndev;
344
345         ASSERT_RTNL();
346
347         if (dev->mtu < IPV6_MIN_MTU)
348                 return NULL;
349
350         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
351
352         if (ndev == NULL)
353                 return NULL;
354
355         rwlock_init(&ndev->lock);
356         ndev->dev = dev;
357         INIT_LIST_HEAD(&ndev->addr_list);
358
359         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
360         ndev->cnf.mtu6 = dev->mtu;
361         ndev->cnf.sysctl = NULL;
362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
363         if (ndev->nd_parms == NULL) {
364                 kfree(ndev);
365                 return NULL;
366         }
367         if (ndev->cnf.forwarding)
368                 dev_disable_lro(dev);
369         /* We refer to the device */
370         dev_hold(dev);
371
372         if (snmp6_alloc_dev(ndev) < 0) {
373                 ADBG((KERN_WARNING
374                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
375                         __func__, dev->name));
376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
377                 dev_put(dev);
378                 kfree(ndev);
379                 return NULL;
380         }
381
382         if (snmp6_register_dev(ndev) < 0) {
383                 ADBG((KERN_WARNING
384                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
385                         __func__, dev->name));
386                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
387                 ndev->dead = 1;
388                 in6_dev_finish_destroy(ndev);
389                 return NULL;
390         }
391
392         /* One reference from device.  We must do this before
393          * we invoke __ipv6_regen_rndid().
394          */
395         in6_dev_hold(ndev);
396
397         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
398                 ndev->cnf.accept_dad = -1;
399
400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
401         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
402                 printk(KERN_INFO
403                        "%s: Disabled Multicast RS\n",
404                        dev->name);
405                 ndev->cnf.rtr_solicits = 0;
406         }
407 #endif
408
409 #ifdef CONFIG_IPV6_PRIVACY
410         INIT_LIST_HEAD(&ndev->tempaddr_list);
411         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
412         if ((dev->flags&IFF_LOOPBACK) ||
413             dev->type == ARPHRD_TUNNEL ||
414             dev->type == ARPHRD_TUNNEL6 ||
415             dev->type == ARPHRD_SIT ||
416             dev->type == ARPHRD_NONE) {
417                 ndev->cnf.use_tempaddr = -1;
418         } else {
419                 in6_dev_hold(ndev);
420                 ipv6_regen_rndid((unsigned long) ndev);
421         }
422 #endif
423
424         if (netif_running(dev) && addrconf_qdisc_ok(dev))
425                 ndev->if_flags |= IF_READY;
426
427         ipv6_mc_init_dev(ndev);
428         ndev->tstamp = jiffies;
429         addrconf_sysctl_register(ndev);
430         /* protected by rtnl_lock */
431         RCU_INIT_POINTER(dev->ip6_ptr, ndev);
432
433         /* Join all-node multicast group */
434         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
435
436         return ndev;
437 }
438
439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
440 {
441         struct inet6_dev *idev;
442
443         ASSERT_RTNL();
444
445         idev = __in6_dev_get(dev);
446         if (!idev) {
447                 idev = ipv6_add_dev(dev);
448                 if (!idev)
449                         return NULL;
450         }
451
452         if (dev->flags&IFF_UP)
453                 ipv6_mc_up(idev);
454         return idev;
455 }
456
457 #ifdef CONFIG_SYSCTL
458 static void dev_forward_change(struct inet6_dev *idev)
459 {
460         struct net_device *dev;
461         struct inet6_ifaddr *ifa;
462
463         if (!idev)
464                 return;
465         dev = idev->dev;
466         if (idev->cnf.forwarding)
467                 dev_disable_lro(dev);
468         if (dev && (dev->flags & IFF_MULTICAST)) {
469                 if (idev->cnf.forwarding)
470                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
471                 else
472                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
473         }
474
475         list_for_each_entry(ifa, &idev->addr_list, if_list) {
476                 if (ifa->flags&IFA_F_TENTATIVE)
477                         continue;
478                 if (idev->cnf.forwarding)
479                         addrconf_join_anycast(ifa);
480                 else
481                         addrconf_leave_anycast(ifa);
482         }
483 }
484
485
486 static void addrconf_forward_change(struct net *net, __s32 newf)
487 {
488         struct net_device *dev;
489         struct inet6_dev *idev;
490
491         rcu_read_lock();
492         for_each_netdev_rcu(net, dev) {
493                 idev = __in6_dev_get(dev);
494                 if (idev) {
495                         int changed = (!idev->cnf.forwarding) ^ (!newf);
496                         idev->cnf.forwarding = newf;
497                         if (changed)
498                                 dev_forward_change(idev);
499                 }
500         }
501         rcu_read_unlock();
502 }
503
504 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
505 {
506         struct net *net;
507
508         net = (struct net *)table->extra2;
509         if (p == &net->ipv6.devconf_dflt->forwarding)
510                 return 0;
511
512         if (!rtnl_trylock()) {
513                 /* Restore the original values before restarting */
514                 *p = old;
515                 return restart_syscall();
516         }
517
518         if (p == &net->ipv6.devconf_all->forwarding) {
519                 __s32 newf = net->ipv6.devconf_all->forwarding;
520                 net->ipv6.devconf_dflt->forwarding = newf;
521                 addrconf_forward_change(net, newf);
522         } else if ((!*p) ^ (!old))
523                 dev_forward_change((struct inet6_dev *)table->extra1);
524         rtnl_unlock();
525
526         if (*p)
527                 rt6_purge_dflt_routers(net);
528         return 1;
529 }
530 #endif
531
532 /* Nobody refers to this ifaddr, destroy it */
533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
534 {
535         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
536
537 #ifdef NET_REFCNT_DEBUG
538         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
539 #endif
540
541         in6_dev_put(ifp->idev);
542
543         if (del_timer(&ifp->timer))
544                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
545
546         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
547                 pr_warning("Freeing alive inet6 address %p\n", ifp);
548                 return;
549         }
550         dst_release(&ifp->rt->dst);
551
552         kfree_rcu(ifp, rcu);
553 }
554
555 static void
556 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
557 {
558         struct list_head *p;
559         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
560
561         /*
562          * Each device address list is sorted in order of scope -
563          * global before linklocal.
564          */
565         list_for_each(p, &idev->addr_list) {
566                 struct inet6_ifaddr *ifa
567                         = list_entry(p, struct inet6_ifaddr, if_list);
568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
569                         break;
570         }
571
572         list_add_tail(&ifp->if_list, p);
573 }
574
575 static u32 ipv6_addr_hash(const struct in6_addr *addr)
576 {
577         /*
578          * We perform the hash function over the last 64 bits of the address
579          * This will include the IEEE address token on links that support it.
580          */
581         return jhash_2words((__force u32)addr->s6_addr32[2],
582                             (__force u32)addr->s6_addr32[3], 0)
583                 & (IN6_ADDR_HSIZE - 1);
584 }
585
586 /* On success it returns ifp with increased reference count */
587
588 static struct inet6_ifaddr *
589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
590               int scope, u32 flags)
591 {
592         struct inet6_ifaddr *ifa = NULL;
593         struct rt6_info *rt;
594         unsigned int hash;
595         int err = 0;
596         int addr_type = ipv6_addr_type(addr);
597
598         if (addr_type == IPV6_ADDR_ANY ||
599             addr_type & IPV6_ADDR_MULTICAST ||
600             (!(idev->dev->flags & IFF_LOOPBACK) &&
601              addr_type & IPV6_ADDR_LOOPBACK))
602                 return ERR_PTR(-EADDRNOTAVAIL);
603
604         rcu_read_lock_bh();
605         if (idev->dead) {
606                 err = -ENODEV;                  /*XXX*/
607                 goto out2;
608         }
609
610         if (idev->cnf.disable_ipv6) {
611                 err = -EACCES;
612                 goto out2;
613         }
614
615         spin_lock(&addrconf_hash_lock);
616
617         /* Ignore adding duplicate addresses on an interface */
618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
619                 ADBG(("ipv6_add_addr: already assigned\n"));
620                 err = -EEXIST;
621                 goto out;
622         }
623
624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
625
626         if (ifa == NULL) {
627                 ADBG(("ipv6_add_addr: malloc failed\n"));
628                 err = -ENOBUFS;
629                 goto out;
630         }
631
632         rt = addrconf_dst_alloc(idev, addr, 0);
633         if (IS_ERR(rt)) {
634                 err = PTR_ERR(rt);
635                 goto out;
636         }
637
638         ipv6_addr_copy(&ifa->addr, addr);
639
640         spin_lock_init(&ifa->lock);
641         spin_lock_init(&ifa->state_lock);
642         init_timer(&ifa->timer);
643         INIT_HLIST_NODE(&ifa->addr_lst);
644         ifa->timer.data = (unsigned long) ifa;
645         ifa->scope = scope;
646         ifa->prefix_len = pfxlen;
647         ifa->flags = flags | IFA_F_TENTATIVE;
648         ifa->cstamp = ifa->tstamp = jiffies;
649
650         ifa->rt = rt;
651
652         /*
653          * part one of RFC 4429, section 3.3
654          * We should not configure an address as
655          * optimistic if we do not yet know the link
656          * layer address of our nexhop router
657          */
658
659         if (dst_get_neighbour_raw(&rt->dst) == NULL)
660                 ifa->flags &= ~IFA_F_OPTIMISTIC;
661
662         ifa->idev = idev;
663         in6_dev_hold(idev);
664         /* For caller */
665         in6_ifa_hold(ifa);
666
667         /* Add to big hash table */
668         hash = ipv6_addr_hash(addr);
669
670         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
671         spin_unlock(&addrconf_hash_lock);
672
673         write_lock(&idev->lock);
674         /* Add to inet6_dev unicast addr list. */
675         ipv6_link_dev_addr(idev, ifa);
676
677 #ifdef CONFIG_IPV6_PRIVACY
678         if (ifa->flags&IFA_F_TEMPORARY) {
679                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
680                 in6_ifa_hold(ifa);
681         }
682 #endif
683
684         in6_ifa_hold(ifa);
685         write_unlock(&idev->lock);
686 out2:
687         rcu_read_unlock_bh();
688
689         if (likely(err == 0))
690                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
691         else {
692                 kfree(ifa);
693                 ifa = ERR_PTR(err);
694         }
695
696         return ifa;
697 out:
698         spin_unlock(&addrconf_hash_lock);
699         goto out2;
700 }
701
702 /* This function wants to get referenced ifp and releases it before return */
703
704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
705 {
706         struct inet6_ifaddr *ifa, *ifn;
707         struct inet6_dev *idev = ifp->idev;
708         int state;
709         int deleted = 0, onlink = 0;
710         unsigned long expires = jiffies;
711
712         spin_lock_bh(&ifp->state_lock);
713         state = ifp->state;
714         ifp->state = INET6_IFADDR_STATE_DEAD;
715         spin_unlock_bh(&ifp->state_lock);
716
717         if (state == INET6_IFADDR_STATE_DEAD)
718                 goto out;
719
720         spin_lock_bh(&addrconf_hash_lock);
721         hlist_del_init_rcu(&ifp->addr_lst);
722         spin_unlock_bh(&addrconf_hash_lock);
723
724         write_lock_bh(&idev->lock);
725 #ifdef CONFIG_IPV6_PRIVACY
726         if (ifp->flags&IFA_F_TEMPORARY) {
727                 list_del(&ifp->tmp_list);
728                 if (ifp->ifpub) {
729                         in6_ifa_put(ifp->ifpub);
730                         ifp->ifpub = NULL;
731                 }
732                 __in6_ifa_put(ifp);
733         }
734 #endif
735
736         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
737                 if (ifa == ifp) {
738                         list_del_init(&ifp->if_list);
739                         __in6_ifa_put(ifp);
740
741                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
742                                 break;
743                         deleted = 1;
744                         continue;
745                 } else if (ifp->flags & IFA_F_PERMANENT) {
746                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
747                                               ifp->prefix_len)) {
748                                 if (ifa->flags & IFA_F_PERMANENT) {
749                                         onlink = 1;
750                                         if (deleted)
751                                                 break;
752                                 } else {
753                                         unsigned long lifetime;
754
755                                         if (!onlink)
756                                                 onlink = -1;
757
758                                         spin_lock(&ifa->lock);
759
760                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
761                                         /*
762                                          * Note: Because this address is
763                                          * not permanent, lifetime <
764                                          * LONG_MAX / HZ here.
765                                          */
766                                         if (time_before(expires,
767                                                         ifa->tstamp + lifetime * HZ))
768                                                 expires = ifa->tstamp + lifetime * HZ;
769                                         spin_unlock(&ifa->lock);
770                                 }
771                         }
772                 }
773         }
774         write_unlock_bh(&idev->lock);
775
776         addrconf_del_timer(ifp);
777
778         ipv6_ifa_notify(RTM_DELADDR, ifp);
779
780         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
781
782         /*
783          * Purge or update corresponding prefix
784          *
785          * 1) we don't purge prefix here if address was not permanent.
786          *    prefix is managed by its own lifetime.
787          * 2) if there're no addresses, delete prefix.
788          * 3) if there're still other permanent address(es),
789          *    corresponding prefix is still permanent.
790          * 4) otherwise, update prefix lifetime to the
791          *    longest valid lifetime among the corresponding
792          *    addresses on the device.
793          *    Note: subsequent RA will update lifetime.
794          *
795          * --yoshfuji
796          */
797         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
798                 struct in6_addr prefix;
799                 struct rt6_info *rt;
800                 struct net *net = dev_net(ifp->idev->dev);
801                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
802                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
803
804                 if (rt && addrconf_is_prefix_route(rt)) {
805                         if (onlink == 0) {
806                                 ip6_del_rt(rt);
807                                 rt = NULL;
808                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
809                                 rt->rt6i_expires = expires;
810                                 rt->rt6i_flags |= RTF_EXPIRES;
811                         }
812                 }
813                 dst_release(&rt->dst);
814         }
815
816         /* clean up prefsrc entries */
817         rt6_remove_prefsrc(ifp);
818 out:
819         in6_ifa_put(ifp);
820 }
821
822 #ifdef CONFIG_IPV6_PRIVACY
823 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
824 {
825         struct inet6_dev *idev = ifp->idev;
826         struct in6_addr addr, *tmpaddr;
827         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
828         unsigned long regen_advance;
829         int tmp_plen;
830         int ret = 0;
831         int max_addresses;
832         u32 addr_flags;
833         unsigned long now = jiffies;
834
835         write_lock(&idev->lock);
836         if (ift) {
837                 spin_lock_bh(&ift->lock);
838                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
839                 spin_unlock_bh(&ift->lock);
840                 tmpaddr = &addr;
841         } else {
842                 tmpaddr = NULL;
843         }
844 retry:
845         in6_dev_hold(idev);
846         if (idev->cnf.use_tempaddr <= 0) {
847                 write_unlock(&idev->lock);
848                 printk(KERN_INFO
849                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
850                 in6_dev_put(idev);
851                 ret = -1;
852                 goto out;
853         }
854         spin_lock_bh(&ifp->lock);
855         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
856                 idev->cnf.use_tempaddr = -1;    /*XXX*/
857                 spin_unlock_bh(&ifp->lock);
858                 write_unlock(&idev->lock);
859                 printk(KERN_WARNING
860                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
861                 in6_dev_put(idev);
862                 ret = -1;
863                 goto out;
864         }
865         in6_ifa_hold(ifp);
866         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
867         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
868                 spin_unlock_bh(&ifp->lock);
869                 write_unlock(&idev->lock);
870                 printk(KERN_WARNING
871                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
872                 in6_ifa_put(ifp);
873                 in6_dev_put(idev);
874                 ret = -1;
875                 goto out;
876         }
877         memcpy(&addr.s6_addr[8], idev->rndid, 8);
878         age = (now - ifp->tstamp) / HZ;
879         tmp_valid_lft = min_t(__u32,
880                               ifp->valid_lft,
881                               idev->cnf.temp_valid_lft + age);
882         tmp_prefered_lft = min_t(__u32,
883                                  ifp->prefered_lft,
884                                  idev->cnf.temp_prefered_lft + age -
885                                  idev->cnf.max_desync_factor);
886         tmp_plen = ifp->prefix_len;
887         max_addresses = idev->cnf.max_addresses;
888         tmp_tstamp = ifp->tstamp;
889         spin_unlock_bh(&ifp->lock);
890
891         regen_advance = idev->cnf.regen_max_retry *
892                         idev->cnf.dad_transmits *
893                         idev->nd_parms->retrans_time / HZ;
894         write_unlock(&idev->lock);
895
896         /* A temporary address is created only if this calculated Preferred
897          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
898          * an implementation must not create a temporary address with a zero
899          * Preferred Lifetime.
900          */
901         if (tmp_prefered_lft <= regen_advance) {
902                 in6_ifa_put(ifp);
903                 in6_dev_put(idev);
904                 ret = -1;
905                 goto out;
906         }
907
908         addr_flags = IFA_F_TEMPORARY;
909         /* set in addrconf_prefix_rcv() */
910         if (ifp->flags & IFA_F_OPTIMISTIC)
911                 addr_flags |= IFA_F_OPTIMISTIC;
912
913         ift = !max_addresses ||
914               ipv6_count_addresses(idev) < max_addresses ?
915                 ipv6_add_addr(idev, &addr, tmp_plen,
916                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
917                               addr_flags) : NULL;
918         if (!ift || IS_ERR(ift)) {
919                 in6_ifa_put(ifp);
920                 in6_dev_put(idev);
921                 printk(KERN_INFO
922                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
923                 tmpaddr = &addr;
924                 write_lock(&idev->lock);
925                 goto retry;
926         }
927
928         spin_lock_bh(&ift->lock);
929         ift->ifpub = ifp;
930         ift->valid_lft = tmp_valid_lft;
931         ift->prefered_lft = tmp_prefered_lft;
932         ift->cstamp = now;
933         ift->tstamp = tmp_tstamp;
934         spin_unlock_bh(&ift->lock);
935
936         addrconf_dad_start(ift, 0);
937         in6_ifa_put(ift);
938         in6_dev_put(idev);
939 out:
940         return ret;
941 }
942 #endif
943
944 /*
945  *      Choose an appropriate source address (RFC3484)
946  */
947 enum {
948         IPV6_SADDR_RULE_INIT = 0,
949         IPV6_SADDR_RULE_LOCAL,
950         IPV6_SADDR_RULE_SCOPE,
951         IPV6_SADDR_RULE_PREFERRED,
952 #ifdef CONFIG_IPV6_MIP6
953         IPV6_SADDR_RULE_HOA,
954 #endif
955         IPV6_SADDR_RULE_OIF,
956         IPV6_SADDR_RULE_LABEL,
957 #ifdef CONFIG_IPV6_PRIVACY
958         IPV6_SADDR_RULE_PRIVACY,
959 #endif
960         IPV6_SADDR_RULE_ORCHID,
961         IPV6_SADDR_RULE_PREFIX,
962         IPV6_SADDR_RULE_MAX
963 };
964
965 struct ipv6_saddr_score {
966         int                     rule;
967         int                     addr_type;
968         struct inet6_ifaddr     *ifa;
969         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
970         int                     scopedist;
971         int                     matchlen;
972 };
973
974 struct ipv6_saddr_dst {
975         const struct in6_addr *addr;
976         int ifindex;
977         int scope;
978         int label;
979         unsigned int prefs;
980 };
981
982 static inline int ipv6_saddr_preferred(int type)
983 {
984         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
985                 return 1;
986         return 0;
987 }
988
989 static int ipv6_get_saddr_eval(struct net *net,
990                                struct ipv6_saddr_score *score,
991                                struct ipv6_saddr_dst *dst,
992                                int i)
993 {
994         int ret;
995
996         if (i <= score->rule) {
997                 switch (i) {
998                 case IPV6_SADDR_RULE_SCOPE:
999                         ret = score->scopedist;
1000                         break;
1001                 case IPV6_SADDR_RULE_PREFIX:
1002                         ret = score->matchlen;
1003                         break;
1004                 default:
1005                         ret = !!test_bit(i, score->scorebits);
1006                 }
1007                 goto out;
1008         }
1009
1010         switch (i) {
1011         case IPV6_SADDR_RULE_INIT:
1012                 /* Rule 0: remember if hiscore is not ready yet */
1013                 ret = !!score->ifa;
1014                 break;
1015         case IPV6_SADDR_RULE_LOCAL:
1016                 /* Rule 1: Prefer same address */
1017                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018                 break;
1019         case IPV6_SADDR_RULE_SCOPE:
1020                 /* Rule 2: Prefer appropriate scope
1021                  *
1022                  *      ret
1023                  *       ^
1024                  *    -1 |  d 15
1025                  *    ---+--+-+---> scope
1026                  *       |
1027                  *       |             d is scope of the destination.
1028                  *  B-d  |  \
1029                  *       |   \      <- smaller scope is better if
1030                  *  B-15 |    \        if scope is enough for destinaion.
1031                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1032                  * d-C-1 | /
1033                  *       |/         <- greater is better
1034                  *   -C  /             if scope is not enough for destination.
1035                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1036                  *
1037                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039                  * Assume B = 0 and we get C > 29.
1040                  */
1041                 ret = __ipv6_addr_src_scope(score->addr_type);
1042                 if (ret >= dst->scope)
1043                         ret = -ret;
1044                 else
1045                         ret -= 128;     /* 30 is enough */
1046                 score->scopedist = ret;
1047                 break;
1048         case IPV6_SADDR_RULE_PREFERRED:
1049                 /* Rule 3: Avoid deprecated and optimistic addresses */
1050                 ret = ipv6_saddr_preferred(score->addr_type) ||
1051                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052                 break;
1053 #ifdef CONFIG_IPV6_MIP6
1054         case IPV6_SADDR_RULE_HOA:
1055             {
1056                 /* Rule 4: Prefer home address */
1057                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059                 break;
1060             }
1061 #endif
1062         case IPV6_SADDR_RULE_OIF:
1063                 /* Rule 5: Prefer outgoing interface */
1064                 ret = (!dst->ifindex ||
1065                        dst->ifindex == score->ifa->idev->dev->ifindex);
1066                 break;
1067         case IPV6_SADDR_RULE_LABEL:
1068                 /* Rule 6: Prefer matching label */
1069                 ret = ipv6_addr_label(net,
1070                                       &score->ifa->addr, score->addr_type,
1071                                       score->ifa->idev->dev->ifindex) == dst->label;
1072                 break;
1073 #ifdef CONFIG_IPV6_PRIVACY
1074         case IPV6_SADDR_RULE_PRIVACY:
1075             {
1076                 /* Rule 7: Prefer public address
1077                  * Note: prefer temporary address if use_tempaddr >= 2
1078                  */
1079                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1082                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083                 break;
1084             }
1085 #endif
1086         case IPV6_SADDR_RULE_ORCHID:
1087                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1088                  *          non-ORCHID vs non-ORCHID
1089                  */
1090                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091                         ipv6_addr_orchid(dst->addr));
1092                 break;
1093         case IPV6_SADDR_RULE_PREFIX:
1094                 /* Rule 8: Use longest matching prefix */
1095                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096                                                        dst->addr);
1097                 break;
1098         default:
1099                 ret = 0;
1100         }
1101
1102         if (ret)
1103                 __set_bit(i, score->scorebits);
1104         score->rule = i;
1105 out:
1106         return ret;
1107 }
1108
1109 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110                        const struct in6_addr *daddr, unsigned int prefs,
1111                        struct in6_addr *saddr)
1112 {
1113         struct ipv6_saddr_score scores[2],
1114                                 *score = &scores[0], *hiscore = &scores[1];
1115         struct ipv6_saddr_dst dst;
1116         struct net_device *dev;
1117         int dst_type;
1118
1119         dst_type = __ipv6_addr_type(daddr);
1120         dst.addr = daddr;
1121         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122         dst.scope = __ipv6_addr_src_scope(dst_type);
1123         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124         dst.prefs = prefs;
1125
1126         hiscore->rule = -1;
1127         hiscore->ifa = NULL;
1128
1129         rcu_read_lock();
1130
1131         for_each_netdev_rcu(net, dev) {
1132                 struct inet6_dev *idev;
1133
1134                 /* Candidate Source Address (section 4)
1135                  *  - multicast and link-local destination address,
1136                  *    the set of candidate source address MUST only
1137                  *    include addresses assigned to interfaces
1138                  *    belonging to the same link as the outgoing
1139                  *    interface.
1140                  * (- For site-local destination addresses, the
1141                  *    set of candidate source addresses MUST only
1142                  *    include addresses assigned to interfaces
1143                  *    belonging to the same site as the outgoing
1144                  *    interface.)
1145                  */
1146                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1147                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1148                     dst.ifindex && dev->ifindex != dst.ifindex)
1149                         continue;
1150
1151                 idev = __in6_dev_get(dev);
1152                 if (!idev)
1153                         continue;
1154
1155                 read_lock_bh(&idev->lock);
1156                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1157                         int i;
1158
1159                         /*
1160                          * - Tentative Address (RFC2462 section 5.4)
1161                          *  - A tentative address is not considered
1162                          *    "assigned to an interface" in the traditional
1163                          *    sense, unless it is also flagged as optimistic.
1164                          * - Candidate Source Address (section 4)
1165                          *  - In any case, anycast addresses, multicast
1166                          *    addresses, and the unspecified address MUST
1167                          *    NOT be included in a candidate set.
1168                          */
1169                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1170                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171                                 continue;
1172
1173                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174
1175                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1176                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1177                                 LIMIT_NETDEBUG(KERN_DEBUG
1178                                                "ADDRCONF: unspecified / multicast address "
1179                                                "assigned as unicast address on %s",
1180                                                dev->name);
1181                                 continue;
1182                         }
1183
1184                         score->rule = -1;
1185                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186
1187                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1188                                 int minihiscore, miniscore;
1189
1190                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1191                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192
1193                                 if (minihiscore > miniscore) {
1194                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1195                                             score->scopedist > 0) {
1196                                                 /*
1197                                                  * special case:
1198                                                  * each remaining entry
1199                                                  * has too small (not enough)
1200                                                  * scope, because ifa entries
1201                                                  * are sorted by their scope
1202                                                  * values.
1203                                                  */
1204                                                 goto try_nextdev;
1205                                         }
1206                                         break;
1207                                 } else if (minihiscore < miniscore) {
1208                                         if (hiscore->ifa)
1209                                                 in6_ifa_put(hiscore->ifa);
1210
1211                                         in6_ifa_hold(score->ifa);
1212
1213                                         swap(hiscore, score);
1214
1215                                         /* restore our iterator */
1216                                         score->ifa = hiscore->ifa;
1217
1218                                         break;
1219                                 }
1220                         }
1221                 }
1222 try_nextdev:
1223                 read_unlock_bh(&idev->lock);
1224         }
1225         rcu_read_unlock();
1226
1227         if (!hiscore->ifa)
1228                 return -EADDRNOTAVAIL;
1229
1230         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1231         in6_ifa_put(hiscore->ifa);
1232         return 0;
1233 }
1234 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1235
1236 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1237                     unsigned char banned_flags)
1238 {
1239         struct inet6_dev *idev;
1240         int err = -EADDRNOTAVAIL;
1241
1242         rcu_read_lock();
1243         idev = __in6_dev_get(dev);
1244         if (idev) {
1245                 struct inet6_ifaddr *ifp;
1246
1247                 read_lock_bh(&idev->lock);
1248                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1249                         if (ifp->scope == IFA_LINK &&
1250                             !(ifp->flags & banned_flags)) {
1251                                 ipv6_addr_copy(addr, &ifp->addr);
1252                                 err = 0;
1253                                 break;
1254                         }
1255                 }
1256                 read_unlock_bh(&idev->lock);
1257         }
1258         rcu_read_unlock();
1259         return err;
1260 }
1261
1262 static int ipv6_count_addresses(struct inet6_dev *idev)
1263 {
1264         int cnt = 0;
1265         struct inet6_ifaddr *ifp;
1266
1267         read_lock_bh(&idev->lock);
1268         list_for_each_entry(ifp, &idev->addr_list, if_list)
1269                 cnt++;
1270         read_unlock_bh(&idev->lock);
1271         return cnt;
1272 }
1273
1274 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1275                   struct net_device *dev, int strict)
1276 {
1277         struct inet6_ifaddr *ifp;
1278         struct hlist_node *node;
1279         unsigned int hash = ipv6_addr_hash(addr);
1280
1281         rcu_read_lock_bh();
1282         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1283                 if (!net_eq(dev_net(ifp->idev->dev), net))
1284                         continue;
1285                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1286                     !(ifp->flags&IFA_F_TENTATIVE) &&
1287                     (dev == NULL || ifp->idev->dev == dev ||
1288                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1289                         rcu_read_unlock_bh();
1290                         return 1;
1291                 }
1292         }
1293
1294         rcu_read_unlock_bh();
1295         return 0;
1296 }
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1300                                struct net_device *dev)
1301 {
1302         unsigned int hash = ipv6_addr_hash(addr);
1303         struct inet6_ifaddr *ifp;
1304         struct hlist_node *node;
1305
1306         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1307                 if (!net_eq(dev_net(ifp->idev->dev), net))
1308                         continue;
1309                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310                         if (dev == NULL || ifp->idev->dev == dev)
1311                                 return true;
1312                 }
1313         }
1314         return false;
1315 }
1316
1317 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1318 {
1319         struct inet6_dev *idev;
1320         struct inet6_ifaddr *ifa;
1321         int     onlink;
1322
1323         onlink = 0;
1324         rcu_read_lock();
1325         idev = __in6_dev_get(dev);
1326         if (idev) {
1327                 read_lock_bh(&idev->lock);
1328                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1329                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330                                                    ifa->prefix_len);
1331                         if (onlink)
1332                                 break;
1333                 }
1334                 read_unlock_bh(&idev->lock);
1335         }
1336         rcu_read_unlock();
1337         return onlink;
1338 }
1339
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1341
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343                                      struct net_device *dev, int strict)
1344 {
1345         struct inet6_ifaddr *ifp, *result = NULL;
1346         unsigned int hash = ipv6_addr_hash(addr);
1347         struct hlist_node *node;
1348
1349         rcu_read_lock_bh();
1350         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1351                 if (!net_eq(dev_net(ifp->idev->dev), net))
1352                         continue;
1353                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354                         if (dev == NULL || ifp->idev->dev == dev ||
1355                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356                                 result = ifp;
1357                                 in6_ifa_hold(ifp);
1358                                 break;
1359                         }
1360                 }
1361         }
1362         rcu_read_unlock_bh();
1363
1364         return result;
1365 }
1366
1367 /* Gets referenced address, destroys ifaddr */
1368
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1370 {
1371         if (ifp->flags&IFA_F_PERMANENT) {
1372                 spin_lock_bh(&ifp->lock);
1373                 addrconf_del_timer(ifp);
1374                 ifp->flags |= IFA_F_TENTATIVE;
1375                 if (dad_failed)
1376                         ifp->flags |= IFA_F_DADFAILED;
1377                 spin_unlock_bh(&ifp->lock);
1378                 if (dad_failed)
1379                         ipv6_ifa_notify(0, ifp);
1380                 in6_ifa_put(ifp);
1381 #ifdef CONFIG_IPV6_PRIVACY
1382         } else if (ifp->flags&IFA_F_TEMPORARY) {
1383                 struct inet6_ifaddr *ifpub;
1384                 spin_lock_bh(&ifp->lock);
1385                 ifpub = ifp->ifpub;
1386                 if (ifpub) {
1387                         in6_ifa_hold(ifpub);
1388                         spin_unlock_bh(&ifp->lock);
1389                         ipv6_create_tempaddr(ifpub, ifp);
1390                         in6_ifa_put(ifpub);
1391                 } else {
1392                         spin_unlock_bh(&ifp->lock);
1393                 }
1394                 ipv6_del_addr(ifp);
1395 #endif
1396         } else
1397                 ipv6_del_addr(ifp);
1398 }
1399
1400 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1401 {
1402         int err = -ENOENT;
1403
1404         spin_lock(&ifp->state_lock);
1405         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1406                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1407                 err = 0;
1408         }
1409         spin_unlock(&ifp->state_lock);
1410
1411         return err;
1412 }
1413
1414 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1415 {
1416         struct inet6_dev *idev = ifp->idev;
1417
1418         if (addrconf_dad_end(ifp)) {
1419                 in6_ifa_put(ifp);
1420                 return;
1421         }
1422
1423         if (net_ratelimit())
1424                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1425                         ifp->idev->dev->name, &ifp->addr);
1426
1427         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1428                 struct in6_addr addr;
1429
1430                 addr.s6_addr32[0] = htonl(0xfe800000);
1431                 addr.s6_addr32[1] = 0;
1432
1433                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1434                     ipv6_addr_equal(&ifp->addr, &addr)) {
1435                         /* DAD failed for link-local based on MAC address */
1436                         idev->cnf.disable_ipv6 = 1;
1437
1438                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1439                                 ifp->idev->dev->name);
1440                 }
1441         }
1442
1443         addrconf_dad_stop(ifp, 1);
1444 }
1445
1446 /* Join to solicited addr multicast group. */
1447
1448 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1449 {
1450         struct in6_addr maddr;
1451
1452         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453                 return;
1454
1455         addrconf_addr_solict_mult(addr, &maddr);
1456         ipv6_dev_mc_inc(dev, &maddr);
1457 }
1458
1459 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1460 {
1461         struct in6_addr maddr;
1462
1463         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464                 return;
1465
1466         addrconf_addr_solict_mult(addr, &maddr);
1467         __ipv6_dev_mc_dec(idev, &maddr);
1468 }
1469
1470 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1471 {
1472         struct in6_addr addr;
1473         if (ifp->prefix_len == 127) /* RFC 6164 */
1474                 return;
1475         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1476         if (ipv6_addr_any(&addr))
1477                 return;
1478         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1479 }
1480
1481 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1482 {
1483         struct in6_addr addr;
1484         if (ifp->prefix_len == 127) /* RFC 6164 */
1485                 return;
1486         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1487         if (ipv6_addr_any(&addr))
1488                 return;
1489         __ipv6_dev_ac_dec(ifp->idev, &addr);
1490 }
1491
1492 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1493 {
1494         if (dev->addr_len != ETH_ALEN)
1495                 return -1;
1496         memcpy(eui, dev->dev_addr, 3);
1497         memcpy(eui + 5, dev->dev_addr + 3, 3);
1498
1499         /*
1500          * The zSeries OSA network cards can be shared among various
1501          * OS instances, but the OSA cards have only one MAC address.
1502          * This leads to duplicate address conflicts in conjunction
1503          * with IPv6 if more than one instance uses the same card.
1504          *
1505          * The driver for these cards can deliver a unique 16-bit
1506          * identifier for each instance sharing the same card.  It is
1507          * placed instead of 0xFFFE in the interface identifier.  The
1508          * "u" bit of the interface identifier is not inverted in this
1509          * case.  Hence the resulting interface identifier has local
1510          * scope according to RFC2373.
1511          */
1512         if (dev->dev_id) {
1513                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1514                 eui[4] = dev->dev_id & 0xFF;
1515         } else {
1516                 eui[3] = 0xFF;
1517                 eui[4] = 0xFE;
1518                 eui[0] ^= 2;
1519         }
1520         return 0;
1521 }
1522
1523 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1524 {
1525         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1526         if (dev->addr_len != ARCNET_ALEN)
1527                 return -1;
1528         memset(eui, 0, 7);
1529         eui[7] = *(u8*)dev->dev_addr;
1530         return 0;
1531 }
1532
1533 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1534 {
1535         if (dev->addr_len != INFINIBAND_ALEN)
1536                 return -1;
1537         memcpy(eui, dev->dev_addr + 12, 8);
1538         eui[0] |= 2;
1539         return 0;
1540 }
1541
1542 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1543 {
1544         if (addr == 0)
1545                 return -1;
1546         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1547                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1548                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1549                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1550                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1551                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1552         eui[1] = 0;
1553         eui[2] = 0x5E;
1554         eui[3] = 0xFE;
1555         memcpy(eui + 4, &addr, 4);
1556         return 0;
1557 }
1558
1559 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1560 {
1561         if (dev->priv_flags & IFF_ISATAP)
1562                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563         return -1;
1564 }
1565
1566 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1567 {
1568         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569 }
1570
1571 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1572 {
1573         switch (dev->type) {
1574         case ARPHRD_ETHER:
1575         case ARPHRD_FDDI:
1576         case ARPHRD_IEEE802_TR:
1577                 return addrconf_ifid_eui48(eui, dev);
1578         case ARPHRD_ARCNET:
1579                 return addrconf_ifid_arcnet(eui, dev);
1580         case ARPHRD_INFINIBAND:
1581                 return addrconf_ifid_infiniband(eui, dev);
1582         case ARPHRD_SIT:
1583                 return addrconf_ifid_sit(eui, dev);
1584         case ARPHRD_IPGRE:
1585                 return addrconf_ifid_gre(eui, dev);
1586         }
1587         return -1;
1588 }
1589
1590 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1591 {
1592         int err = -1;
1593         struct inet6_ifaddr *ifp;
1594
1595         read_lock_bh(&idev->lock);
1596         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1597                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1598                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1599                         err = 0;
1600                         break;
1601                 }
1602         }
1603         read_unlock_bh(&idev->lock);
1604         return err;
1605 }
1606
1607 #ifdef CONFIG_IPV6_PRIVACY
1608 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1609 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1610 {
1611 regen:
1612         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1613         idev->rndid[0] &= ~0x02;
1614
1615         /*
1616          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1617          * check if generated address is not inappropriate
1618          *
1619          *  - Reserved subnet anycast (RFC 2526)
1620          *      11111101 11....11 1xxxxxxx
1621          *  - ISATAP (RFC4214) 6.1
1622          *      00-00-5E-FE-xx-xx-xx-xx
1623          *  - value 0
1624          *  - XXX: already assigned to an address on the device
1625          */
1626         if (idev->rndid[0] == 0xfd &&
1627             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1628             (idev->rndid[7]&0x80))
1629                 goto regen;
1630         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1631                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1632                         goto regen;
1633                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1634                         goto regen;
1635         }
1636
1637         return 0;
1638 }
1639
1640 static void ipv6_regen_rndid(unsigned long data)
1641 {
1642         struct inet6_dev *idev = (struct inet6_dev *) data;
1643         unsigned long expires;
1644
1645         rcu_read_lock_bh();
1646         write_lock_bh(&idev->lock);
1647
1648         if (idev->dead)
1649                 goto out;
1650
1651         if (__ipv6_regen_rndid(idev) < 0)
1652                 goto out;
1653
1654         expires = jiffies +
1655                 idev->cnf.temp_prefered_lft * HZ -
1656                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1657                 idev->cnf.max_desync_factor * HZ;
1658         if (time_before(expires, jiffies)) {
1659                 printk(KERN_WARNING
1660                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1661                         idev->dev->name);
1662                 goto out;
1663         }
1664
1665         if (!mod_timer(&idev->regen_timer, expires))
1666                 in6_dev_hold(idev);
1667
1668 out:
1669         write_unlock_bh(&idev->lock);
1670         rcu_read_unlock_bh();
1671         in6_dev_put(idev);
1672 }
1673
1674 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1675         int ret = 0;
1676
1677         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1678                 ret = __ipv6_regen_rndid(idev);
1679         return ret;
1680 }
1681 #endif
1682
1683 /*
1684  *      Add prefix route.
1685  */
1686
1687 static void
1688 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1689                       unsigned long expires, u32 flags)
1690 {
1691         struct fib6_config cfg = {
1692                 .fc_table = RT6_TABLE_PREFIX,
1693                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1694                 .fc_ifindex = dev->ifindex,
1695                 .fc_expires = expires,
1696                 .fc_dst_len = plen,
1697                 .fc_flags = RTF_UP | flags,
1698                 .fc_nlinfo.nl_net = dev_net(dev),
1699                 .fc_protocol = RTPROT_KERNEL,
1700         };
1701
1702         ipv6_addr_copy(&cfg.fc_dst, pfx);
1703
1704         /* Prevent useless cloning on PtP SIT.
1705            This thing is done here expecting that the whole
1706            class of non-broadcast devices need not cloning.
1707          */
1708 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1709         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1710                 cfg.fc_flags |= RTF_NONEXTHOP;
1711 #endif
1712
1713         ip6_route_add(&cfg);
1714 }
1715
1716 /* Create "default" multicast route to the interface */
1717
1718 static void addrconf_add_mroute(struct net_device *dev)
1719 {
1720         struct fib6_config cfg = {
1721                 .fc_table = RT6_TABLE_LOCAL,
1722                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1723                 .fc_ifindex = dev->ifindex,
1724                 .fc_dst_len = 8,
1725                 .fc_flags = RTF_UP,
1726                 .fc_nlinfo.nl_net = dev_net(dev),
1727         };
1728
1729         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1730
1731         ip6_route_add(&cfg);
1732 }
1733
1734 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1735 static void sit_route_add(struct net_device *dev)
1736 {
1737         struct fib6_config cfg = {
1738                 .fc_table = RT6_TABLE_MAIN,
1739                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1740                 .fc_ifindex = dev->ifindex,
1741                 .fc_dst_len = 96,
1742                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1743                 .fc_nlinfo.nl_net = dev_net(dev),
1744         };
1745
1746         /* prefix length - 96 bits "::d.d.d.d" */
1747         ip6_route_add(&cfg);
1748 }
1749 #endif
1750
1751 static void addrconf_add_lroute(struct net_device *dev)
1752 {
1753         struct in6_addr addr;
1754
1755         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1756         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1757 }
1758
1759 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1760 {
1761         struct inet6_dev *idev;
1762
1763         ASSERT_RTNL();
1764
1765         idev = ipv6_find_idev(dev);
1766         if (!idev)
1767                 return ERR_PTR(-ENOBUFS);
1768
1769         if (idev->cnf.disable_ipv6)
1770                 return ERR_PTR(-EACCES);
1771
1772         /* Add default multicast route */
1773         addrconf_add_mroute(dev);
1774
1775         /* Add link local route */
1776         addrconf_add_lroute(dev);
1777         return idev;
1778 }
1779
1780 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1781 {
1782         struct prefix_info *pinfo;
1783         __u32 valid_lft;
1784         __u32 prefered_lft;
1785         int addr_type;
1786         struct inet6_dev *in6_dev;
1787         struct net *net = dev_net(dev);
1788
1789         pinfo = (struct prefix_info *) opt;
1790
1791         if (len < sizeof(struct prefix_info)) {
1792                 ADBG(("addrconf: prefix option too short\n"));
1793                 return;
1794         }
1795
1796         /*
1797          *      Validation checks ([ADDRCONF], page 19)
1798          */
1799
1800         addr_type = ipv6_addr_type(&pinfo->prefix);
1801
1802         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1803                 return;
1804
1805         valid_lft = ntohl(pinfo->valid);
1806         prefered_lft = ntohl(pinfo->prefered);
1807
1808         if (prefered_lft > valid_lft) {
1809                 if (net_ratelimit())
1810                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1811                 return;
1812         }
1813
1814         in6_dev = in6_dev_get(dev);
1815
1816         if (in6_dev == NULL) {
1817                 if (net_ratelimit())
1818                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1819                 return;
1820         }
1821
1822         /*
1823          *      Two things going on here:
1824          *      1) Add routes for on-link prefixes
1825          *      2) Configure prefixes with the auto flag set
1826          */
1827
1828         if (pinfo->onlink) {
1829                 struct rt6_info *rt;
1830                 unsigned long rt_expires;
1831
1832                 /* Avoid arithmetic overflow. Really, we could
1833                  * save rt_expires in seconds, likely valid_lft,
1834                  * but it would require division in fib gc, that it
1835                  * not good.
1836                  */
1837                 if (HZ > USER_HZ)
1838                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1839                 else
1840                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1841
1842                 if (addrconf_finite_timeout(rt_expires))
1843                         rt_expires *= HZ;
1844
1845                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1846                                 dev->ifindex, 1);
1847
1848                 if (rt && addrconf_is_prefix_route(rt)) {
1849                         /* Autoconf prefix route */
1850                         if (valid_lft == 0) {
1851                                 ip6_del_rt(rt);
1852                                 rt = NULL;
1853                         } else if (addrconf_finite_timeout(rt_expires)) {
1854                                 /* not infinity */
1855                                 rt->rt6i_expires = jiffies + rt_expires;
1856                                 rt->rt6i_flags |= RTF_EXPIRES;
1857                         } else {
1858                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1859                                 rt->rt6i_expires = 0;
1860                         }
1861                 } else if (valid_lft) {
1862                         clock_t expires = 0;
1863                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1864                         if (addrconf_finite_timeout(rt_expires)) {
1865                                 /* not infinity */
1866                                 flags |= RTF_EXPIRES;
1867                                 expires = jiffies_to_clock_t(rt_expires);
1868                         }
1869                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1870                                               dev, expires, flags);
1871                 }
1872                 if (rt)
1873                         dst_release(&rt->dst);
1874         }
1875
1876         /* Try to figure out our local address for this prefix */
1877
1878         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1879                 struct inet6_ifaddr * ifp;
1880                 struct in6_addr addr;
1881                 int create = 0, update_lft = 0;
1882
1883                 if (pinfo->prefix_len == 64) {
1884                         memcpy(&addr, &pinfo->prefix, 8);
1885                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1886                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1887                                 in6_dev_put(in6_dev);
1888                                 return;
1889                         }
1890                         goto ok;
1891                 }
1892                 if (net_ratelimit())
1893                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1894                                pinfo->prefix_len);
1895                 in6_dev_put(in6_dev);
1896                 return;
1897
1898 ok:
1899
1900                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1901
1902                 if (ifp == NULL && valid_lft) {
1903                         int max_addresses = in6_dev->cnf.max_addresses;
1904                         u32 addr_flags = 0;
1905
1906 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1907                         if (in6_dev->cnf.optimistic_dad &&
1908                             !net->ipv6.devconf_all->forwarding)
1909                                 addr_flags = IFA_F_OPTIMISTIC;
1910 #endif
1911
1912                         /* Do not allow to create too much of autoconfigured
1913                          * addresses; this would be too easy way to crash kernel.
1914                          */
1915                         if (!max_addresses ||
1916                             ipv6_count_addresses(in6_dev) < max_addresses)
1917                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1918                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1919                                                     addr_flags);
1920
1921                         if (!ifp || IS_ERR(ifp)) {
1922                                 in6_dev_put(in6_dev);
1923                                 return;
1924                         }
1925
1926                         update_lft = create = 1;
1927                         ifp->cstamp = jiffies;
1928                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1929                 }
1930
1931                 if (ifp) {
1932                         int flags;
1933                         unsigned long now;
1934 #ifdef CONFIG_IPV6_PRIVACY
1935                         struct inet6_ifaddr *ift;
1936 #endif
1937                         u32 stored_lft;
1938
1939                         /* update lifetime (RFC2462 5.5.3 e) */
1940                         spin_lock(&ifp->lock);
1941                         now = jiffies;
1942                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1943                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1944                         else
1945                                 stored_lft = 0;
1946                         if (!update_lft && stored_lft) {
1947                                 if (valid_lft > MIN_VALID_LIFETIME ||
1948                                     valid_lft > stored_lft)
1949                                         update_lft = 1;
1950                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1951                                         /* valid_lft <= stored_lft is always true */
1952                                         /*
1953                                          * RFC 4862 Section 5.5.3e:
1954                                          * "Note that the preferred lifetime of
1955                                          *  the corresponding address is always
1956                                          *  reset to the Preferred Lifetime in
1957                                          *  the received Prefix Information
1958                                          *  option, regardless of whether the
1959                                          *  valid lifetime is also reset or
1960                                          *  ignored."
1961                                          *
1962                                          *  So if the preferred lifetime in
1963                                          *  this advertisement is different
1964                                          *  than what we have stored, but the
1965                                          *  valid lifetime is invalid, just
1966                                          *  reset prefered_lft.
1967                                          *
1968                                          *  We must set the valid lifetime
1969                                          *  to the stored lifetime since we'll
1970                                          *  be updating the timestamp below,
1971                                          *  else we'll set it back to the
1972                                          *  minimum.
1973                                          */
1974                                         if (prefered_lft != ifp->prefered_lft) {
1975                                                 valid_lft = stored_lft;
1976                                                 update_lft = 1;
1977                                         }
1978                                 } else {
1979                                         valid_lft = MIN_VALID_LIFETIME;
1980                                         if (valid_lft < prefered_lft)
1981                                                 prefered_lft = valid_lft;
1982                                         update_lft = 1;
1983                                 }
1984                         }
1985
1986                         if (update_lft) {
1987                                 ifp->valid_lft = valid_lft;
1988                                 ifp->prefered_lft = prefered_lft;
1989                                 ifp->tstamp = now;
1990                                 flags = ifp->flags;
1991                                 ifp->flags &= ~IFA_F_DEPRECATED;
1992                                 spin_unlock(&ifp->lock);
1993
1994                                 if (!(flags&IFA_F_TENTATIVE))
1995                                         ipv6_ifa_notify(0, ifp);
1996                         } else
1997                                 spin_unlock(&ifp->lock);
1998
1999 #ifdef CONFIG_IPV6_PRIVACY
2000                         read_lock_bh(&in6_dev->lock);
2001                         /* update all temporary addresses in the list */
2002                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2003                                             tmp_list) {
2004                                 int age, max_valid, max_prefered;
2005
2006                                 if (ifp != ift->ifpub)
2007                                         continue;
2008
2009                                 /*
2010                                  * RFC 4941 section 3.3:
2011                                  * If a received option will extend the lifetime
2012                                  * of a public address, the lifetimes of
2013                                  * temporary addresses should be extended,
2014                                  * subject to the overall constraint that no
2015                                  * temporary addresses should ever remain
2016                                  * "valid" or "preferred" for a time longer than
2017                                  * (TEMP_VALID_LIFETIME) or
2018                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2019                                  * respectively.
2020                                  */
2021                                 age = (now - ift->cstamp) / HZ;
2022                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2023                                 if (max_valid < 0)
2024                                         max_valid = 0;
2025
2026                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2027                                                in6_dev->cnf.max_desync_factor -
2028                                                age;
2029                                 if (max_prefered < 0)
2030                                         max_prefered = 0;
2031
2032                                 if (valid_lft > max_valid)
2033                                         valid_lft = max_valid;
2034
2035                                 if (prefered_lft > max_prefered)
2036                                         prefered_lft = max_prefered;
2037
2038                                 spin_lock(&ift->lock);
2039                                 flags = ift->flags;
2040                                 ift->valid_lft = valid_lft;
2041                                 ift->prefered_lft = prefered_lft;
2042                                 ift->tstamp = now;
2043                                 if (prefered_lft > 0)
2044                                         ift->flags &= ~IFA_F_DEPRECATED;
2045
2046                                 spin_unlock(&ift->lock);
2047                                 if (!(flags&IFA_F_TENTATIVE))
2048                                         ipv6_ifa_notify(0, ift);
2049                         }
2050
2051                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2052                                 /*
2053                                  * When a new public address is created as
2054                                  * described in [ADDRCONF], also create a new
2055                                  * temporary address. Also create a temporary
2056                                  * address if it's enabled but no temporary
2057                                  * address currently exists.
2058                                  */
2059                                 read_unlock_bh(&in6_dev->lock);
2060                                 ipv6_create_tempaddr(ifp, NULL);
2061                         } else {
2062                                 read_unlock_bh(&in6_dev->lock);
2063                         }
2064 #endif
2065                         in6_ifa_put(ifp);
2066                         addrconf_verify(0);
2067                 }
2068         }
2069         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2070         in6_dev_put(in6_dev);
2071 }
2072
2073 /*
2074  *      Set destination address.
2075  *      Special case for SIT interfaces where we create a new "virtual"
2076  *      device.
2077  */
2078 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2079 {
2080         struct in6_ifreq ireq;
2081         struct net_device *dev;
2082         int err = -EINVAL;
2083
2084         rtnl_lock();
2085
2086         err = -EFAULT;
2087         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2088                 goto err_exit;
2089
2090         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2091
2092         err = -ENODEV;
2093         if (dev == NULL)
2094                 goto err_exit;
2095
2096 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2097         if (dev->type == ARPHRD_SIT) {
2098                 const struct net_device_ops *ops = dev->netdev_ops;
2099                 struct ifreq ifr;
2100                 struct ip_tunnel_parm p;
2101
2102                 err = -EADDRNOTAVAIL;
2103                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2104                         goto err_exit;
2105
2106                 memset(&p, 0, sizeof(p));
2107                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2108                 p.iph.saddr = 0;
2109                 p.iph.version = 4;
2110                 p.iph.ihl = 5;
2111                 p.iph.protocol = IPPROTO_IPV6;
2112                 p.iph.ttl = 64;
2113                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2114
2115                 if (ops->ndo_do_ioctl) {
2116                         mm_segment_t oldfs = get_fs();
2117
2118                         set_fs(KERNEL_DS);
2119                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2120                         set_fs(oldfs);
2121                 } else
2122                         err = -EOPNOTSUPP;
2123
2124                 if (err == 0) {
2125                         err = -ENOBUFS;
2126                         dev = __dev_get_by_name(net, p.name);
2127                         if (!dev)
2128                                 goto err_exit;
2129                         err = dev_open(dev);
2130                 }
2131         }
2132 #endif
2133
2134 err_exit:
2135         rtnl_unlock();
2136         return err;
2137 }
2138
2139 /*
2140  *      Manual configuration of address on an interface
2141  */
2142 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2143                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2144                           __u32 valid_lft)
2145 {
2146         struct inet6_ifaddr *ifp;
2147         struct inet6_dev *idev;
2148         struct net_device *dev;
2149         int scope;
2150         u32 flags;
2151         clock_t expires;
2152         unsigned long timeout;
2153
2154         ASSERT_RTNL();
2155
2156         if (plen > 128)
2157                 return -EINVAL;
2158
2159         /* check the lifetime */
2160         if (!valid_lft || prefered_lft > valid_lft)
2161                 return -EINVAL;
2162
2163         dev = __dev_get_by_index(net, ifindex);
2164         if (!dev)
2165                 return -ENODEV;
2166
2167         idev = addrconf_add_dev(dev);
2168         if (IS_ERR(idev))
2169                 return PTR_ERR(idev);
2170
2171         scope = ipv6_addr_scope(pfx);
2172
2173         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2174         if (addrconf_finite_timeout(timeout)) {
2175                 expires = jiffies_to_clock_t(timeout * HZ);
2176                 valid_lft = timeout;
2177                 flags = RTF_EXPIRES;
2178         } else {
2179                 expires = 0;
2180                 flags = 0;
2181                 ifa_flags |= IFA_F_PERMANENT;
2182         }
2183
2184         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2185         if (addrconf_finite_timeout(timeout)) {
2186                 if (timeout == 0)
2187                         ifa_flags |= IFA_F_DEPRECATED;
2188                 prefered_lft = timeout;
2189         }
2190
2191         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2192
2193         if (!IS_ERR(ifp)) {
2194                 spin_lock_bh(&ifp->lock);
2195                 ifp->valid_lft = valid_lft;
2196                 ifp->prefered_lft = prefered_lft;
2197                 ifp->tstamp = jiffies;
2198                 spin_unlock_bh(&ifp->lock);
2199
2200                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2201                                       expires, flags);
2202                 /*
2203                  * Note that section 3.1 of RFC 4429 indicates
2204                  * that the Optimistic flag should not be set for
2205                  * manually configured addresses
2206                  */
2207                 addrconf_dad_start(ifp, 0);
2208                 in6_ifa_put(ifp);
2209                 addrconf_verify(0);
2210                 return 0;
2211         }
2212
2213         return PTR_ERR(ifp);
2214 }
2215
2216 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2217                           unsigned int plen)
2218 {
2219         struct inet6_ifaddr *ifp;
2220         struct inet6_dev *idev;
2221         struct net_device *dev;
2222
2223         if (plen > 128)
2224                 return -EINVAL;
2225
2226         dev = __dev_get_by_index(net, ifindex);
2227         if (!dev)
2228                 return -ENODEV;
2229
2230         if ((idev = __in6_dev_get(dev)) == NULL)
2231                 return -ENXIO;
2232
2233         read_lock_bh(&idev->lock);
2234         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2235                 if (ifp->prefix_len == plen &&
2236                     ipv6_addr_equal(pfx, &ifp->addr)) {
2237                         in6_ifa_hold(ifp);
2238                         read_unlock_bh(&idev->lock);
2239
2240                         ipv6_del_addr(ifp);
2241
2242                         /* If the last address is deleted administratively,
2243                            disable IPv6 on this interface.
2244                          */
2245                         if (list_empty(&idev->addr_list))
2246                                 addrconf_ifdown(idev->dev, 1);
2247                         return 0;
2248                 }
2249         }
2250         read_unlock_bh(&idev->lock);
2251         return -EADDRNOTAVAIL;
2252 }
2253
2254
2255 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2256 {
2257         struct in6_ifreq ireq;
2258         int err;
2259
2260         if (!capable(CAP_NET_ADMIN))
2261                 return -EPERM;
2262
2263         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2264                 return -EFAULT;
2265
2266         rtnl_lock();
2267         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2268                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2269                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2270         rtnl_unlock();
2271         return err;
2272 }
2273
2274 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2275 {
2276         struct in6_ifreq ireq;
2277         int err;
2278
2279         if (!capable(CAP_NET_ADMIN))
2280                 return -EPERM;
2281
2282         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2283                 return -EFAULT;
2284
2285         rtnl_lock();
2286         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2287                              ireq.ifr6_prefixlen);
2288         rtnl_unlock();
2289         return err;
2290 }
2291
2292 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2293                      int plen, int scope)
2294 {
2295         struct inet6_ifaddr *ifp;
2296
2297         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2298         if (!IS_ERR(ifp)) {
2299                 spin_lock_bh(&ifp->lock);
2300                 ifp->flags &= ~IFA_F_TENTATIVE;
2301                 spin_unlock_bh(&ifp->lock);
2302                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2303                 in6_ifa_put(ifp);
2304         }
2305 }
2306
2307 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2308 static void sit_add_v4_addrs(struct inet6_dev *idev)
2309 {
2310         struct in6_addr addr;
2311         struct net_device *dev;
2312         struct net *net = dev_net(idev->dev);
2313         int scope;
2314
2315         ASSERT_RTNL();
2316
2317         memset(&addr, 0, sizeof(struct in6_addr));
2318         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2319
2320         if (idev->dev->flags&IFF_POINTOPOINT) {
2321                 addr.s6_addr32[0] = htonl(0xfe800000);
2322                 scope = IFA_LINK;
2323         } else {
2324                 scope = IPV6_ADDR_COMPATv4;
2325         }
2326
2327         if (addr.s6_addr32[3]) {
2328                 add_addr(idev, &addr, 128, scope);
2329                 return;
2330         }
2331
2332         for_each_netdev(net, dev) {
2333                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2334                 if (in_dev && (dev->flags & IFF_UP)) {
2335                         struct in_ifaddr * ifa;
2336
2337                         int flag = scope;
2338
2339                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2340                                 int plen;
2341
2342                                 addr.s6_addr32[3] = ifa->ifa_local;
2343
2344                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2345                                         continue;
2346                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2347                                         if (idev->dev->flags&IFF_POINTOPOINT)
2348                                                 continue;
2349                                         flag |= IFA_HOST;
2350                                 }
2351                                 if (idev->dev->flags&IFF_POINTOPOINT)
2352                                         plen = 64;
2353                                 else
2354                                         plen = 96;
2355
2356                                 add_addr(idev, &addr, plen, flag);
2357                         }
2358                 }
2359         }
2360 }
2361 #endif
2362
2363 static void init_loopback(struct net_device *dev)
2364 {
2365         struct inet6_dev  *idev;
2366
2367         /* ::1 */
2368
2369         ASSERT_RTNL();
2370
2371         if ((idev = ipv6_find_idev(dev)) == NULL) {
2372                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2373                 return;
2374         }
2375
2376         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2377 }
2378
2379 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2380 {
2381         struct inet6_ifaddr * ifp;
2382         u32 addr_flags = IFA_F_PERMANENT;
2383
2384 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2385         if (idev->cnf.optimistic_dad &&
2386             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2387                 addr_flags |= IFA_F_OPTIMISTIC;
2388 #endif
2389
2390
2391         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2392         if (!IS_ERR(ifp)) {
2393                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2394                 addrconf_dad_start(ifp, 0);
2395                 in6_ifa_put(ifp);
2396         }
2397 }
2398
2399 static void addrconf_dev_config(struct net_device *dev)
2400 {
2401         struct in6_addr addr;
2402         struct inet6_dev    * idev;
2403
2404         ASSERT_RTNL();
2405
2406         if ((dev->type != ARPHRD_ETHER) &&
2407             (dev->type != ARPHRD_FDDI) &&
2408             (dev->type != ARPHRD_IEEE802_TR) &&
2409             (dev->type != ARPHRD_ARCNET) &&
2410             (dev->type != ARPHRD_INFINIBAND)) {
2411                 /* Alas, we support only Ethernet autoconfiguration. */
2412                 return;
2413         }
2414
2415         idev = addrconf_add_dev(dev);
2416         if (IS_ERR(idev))
2417                 return;
2418
2419         memset(&addr, 0, sizeof(struct in6_addr));
2420         addr.s6_addr32[0] = htonl(0xFE800000);
2421
2422         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2423                 addrconf_add_linklocal(idev, &addr);
2424 }
2425
2426 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2427 static void addrconf_sit_config(struct net_device *dev)
2428 {
2429         struct inet6_dev *idev;
2430
2431         ASSERT_RTNL();
2432
2433         /*
2434          * Configure the tunnel with one of our IPv4
2435          * addresses... we should configure all of
2436          * our v4 addrs in the tunnel
2437          */
2438
2439         if ((idev = ipv6_find_idev(dev)) == NULL) {
2440                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2441                 return;
2442         }
2443
2444         if (dev->priv_flags & IFF_ISATAP) {
2445                 struct in6_addr addr;
2446
2447                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2448                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2449                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2450                         addrconf_add_linklocal(idev, &addr);
2451                 return;
2452         }
2453
2454         sit_add_v4_addrs(idev);
2455
2456         if (dev->flags&IFF_POINTOPOINT) {
2457                 addrconf_add_mroute(dev);
2458                 addrconf_add_lroute(dev);
2459         } else
2460                 sit_route_add(dev);
2461 }
2462 #endif
2463
2464 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2465 static void addrconf_gre_config(struct net_device *dev)
2466 {
2467         struct inet6_dev *idev;
2468         struct in6_addr addr;
2469
2470         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2471
2472         ASSERT_RTNL();
2473
2474         if ((idev = ipv6_find_idev(dev)) == NULL) {
2475                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2476                 return;
2477         }
2478
2479         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2480         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2481
2482         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2483                 addrconf_add_linklocal(idev, &addr);
2484 }
2485 #endif
2486
2487 static inline int
2488 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2489 {
2490         struct in6_addr lladdr;
2491
2492         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2493                 addrconf_add_linklocal(idev, &lladdr);
2494                 return 0;
2495         }
2496         return -1;
2497 }
2498
2499 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2500 {
2501         struct net_device *link_dev;
2502         struct net *net = dev_net(idev->dev);
2503
2504         /* first try to inherit the link-local address from the link device */
2505         if (idev->dev->iflink &&
2506             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2507                 if (!ipv6_inherit_linklocal(idev, link_dev))
2508                         return;
2509         }
2510         /* then try to inherit it from any device */
2511         for_each_netdev(net, link_dev) {
2512                 if (!ipv6_inherit_linklocal(idev, link_dev))
2513                         return;
2514         }
2515         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2516 }
2517
2518 /*
2519  * Autoconfigure tunnel with a link-local address so routing protocols,
2520  * DHCPv6, MLD etc. can be run over the virtual link
2521  */
2522
2523 static void addrconf_ip6_tnl_config(struct net_device *dev)
2524 {
2525         struct inet6_dev *idev;
2526
2527         ASSERT_RTNL();
2528
2529         idev = addrconf_add_dev(dev);
2530         if (IS_ERR(idev)) {
2531                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2532                 return;
2533         }
2534         ip6_tnl_add_linklocal(idev);
2535 }
2536
2537 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2538                            void * data)
2539 {
2540         struct net_device *dev = (struct net_device *) data;
2541         struct inet6_dev *idev = __in6_dev_get(dev);
2542         int run_pending = 0;
2543         int err;
2544
2545         switch (event) {
2546         case NETDEV_REGISTER:
2547                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2548                         idev = ipv6_add_dev(dev);
2549                         if (!idev)
2550                                 return notifier_from_errno(-ENOMEM);
2551                 }
2552                 break;
2553
2554         case NETDEV_UP:
2555         case NETDEV_CHANGE:
2556                 if (dev->flags & IFF_SLAVE)
2557                         break;
2558
2559                 if (event == NETDEV_UP) {
2560                         if (!addrconf_qdisc_ok(dev)) {
2561                                 /* device is not ready yet. */
2562                                 printk(KERN_INFO
2563                                         "ADDRCONF(NETDEV_UP): %s: "
2564                                         "link is not ready\n",
2565                                         dev->name);
2566                                 break;
2567                         }
2568
2569                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2570                                 idev = ipv6_add_dev(dev);
2571
2572                         if (idev) {
2573                                 idev->if_flags |= IF_READY;
2574                                 run_pending = 1;
2575                         }
2576                 } else {
2577                         if (!addrconf_qdisc_ok(dev)) {
2578                                 /* device is still not ready. */
2579                                 break;
2580                         }
2581
2582                         if (idev) {
2583                                 if (idev->if_flags & IF_READY)
2584                                         /* device is already configured. */
2585                                         break;
2586                                 idev->if_flags |= IF_READY;
2587                         }
2588
2589                         printk(KERN_INFO
2590                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2591                                         "link becomes ready\n",
2592                                         dev->name);
2593
2594                         run_pending = 1;
2595                 }
2596
2597                 switch (dev->type) {
2598 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2599                 case ARPHRD_SIT:
2600                         addrconf_sit_config(dev);
2601                         break;
2602 #endif
2603 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2604                 case ARPHRD_IPGRE:
2605                         addrconf_gre_config(dev);
2606                         break;
2607 #endif
2608                 case ARPHRD_TUNNEL6:
2609                         addrconf_ip6_tnl_config(dev);
2610                         break;
2611                 case ARPHRD_LOOPBACK:
2612                         init_loopback(dev);
2613                         break;
2614
2615                 default:
2616                         addrconf_dev_config(dev);
2617                         break;
2618                 }
2619
2620                 if (idev) {
2621                         if (run_pending)
2622                                 addrconf_dad_run(idev);
2623
2624                         /*
2625                          * If the MTU changed during the interface down,
2626                          * when the interface up, the changed MTU must be
2627                          * reflected in the idev as well as routers.
2628                          */
2629                         if (idev->cnf.mtu6 != dev->mtu &&
2630                             dev->mtu >= IPV6_MIN_MTU) {
2631                                 rt6_mtu_change(dev, dev->mtu);
2632                                 idev->cnf.mtu6 = dev->mtu;
2633                         }
2634                         idev->tstamp = jiffies;
2635                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2636
2637                         /*
2638                          * If the changed mtu during down is lower than
2639                          * IPV6_MIN_MTU stop IPv6 on this interface.
2640                          */
2641                         if (dev->mtu < IPV6_MIN_MTU)
2642                                 addrconf_ifdown(dev, 1);
2643                 }
2644                 break;
2645
2646         case NETDEV_CHANGEMTU:
2647                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2648                         rt6_mtu_change(dev, dev->mtu);
2649                         idev->cnf.mtu6 = dev->mtu;
2650                         break;
2651                 }
2652
2653                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2654                         idev = ipv6_add_dev(dev);
2655                         if (idev)
2656                                 break;
2657                 }
2658
2659                 /*
2660                  * MTU falled under IPV6_MIN_MTU.
2661                  * Stop IPv6 on this interface.
2662                  */
2663
2664         case NETDEV_DOWN:
2665         case NETDEV_UNREGISTER:
2666                 /*
2667                  *      Remove all addresses from this interface.
2668                  */
2669                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2670                 break;
2671
2672         case NETDEV_CHANGENAME:
2673                 if (idev) {
2674                         snmp6_unregister_dev(idev);
2675                         addrconf_sysctl_unregister(idev);
2676                         addrconf_sysctl_register(idev);
2677                         err = snmp6_register_dev(idev);
2678                         if (err)
2679                                 return notifier_from_errno(err);
2680                 }
2681                 break;
2682
2683         case NETDEV_PRE_TYPE_CHANGE:
2684         case NETDEV_POST_TYPE_CHANGE:
2685                 addrconf_type_change(dev, event);
2686                 break;
2687         }
2688
2689         return NOTIFY_OK;
2690 }
2691
2692 /*
2693  *      addrconf module should be notified of a device going up
2694  */
2695 static struct notifier_block ipv6_dev_notf = {
2696         .notifier_call = addrconf_notify,
2697 };
2698
2699 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2700 {
2701         struct inet6_dev *idev;
2702         ASSERT_RTNL();
2703
2704         idev = __in6_dev_get(dev);
2705
2706         if (event == NETDEV_POST_TYPE_CHANGE)
2707                 ipv6_mc_remap(idev);
2708         else if (event == NETDEV_PRE_TYPE_CHANGE)
2709                 ipv6_mc_unmap(idev);
2710 }
2711
2712 static int addrconf_ifdown(struct net_device *dev, int how)
2713 {
2714         struct net *net = dev_net(dev);
2715         struct inet6_dev *idev;
2716         struct inet6_ifaddr *ifa;
2717         int state, i;
2718
2719         ASSERT_RTNL();
2720
2721         rt6_ifdown(net, dev);
2722         neigh_ifdown(&nd_tbl, dev);
2723
2724         idev = __in6_dev_get(dev);
2725         if (idev == NULL)
2726                 return -ENODEV;
2727
2728         /*
2729          * Step 1: remove reference to ipv6 device from parent device.
2730          *         Do not dev_put!
2731          */
2732         if (how) {
2733                 idev->dead = 1;
2734
2735                 /* protected by rtnl_lock */
2736                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2737
2738                 /* Step 1.5: remove snmp6 entry */
2739                 snmp6_unregister_dev(idev);
2740
2741         }
2742
2743         /* Step 2: clear hash table */
2744         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2745                 struct hlist_head *h = &inet6_addr_lst[i];
2746                 struct hlist_node *n;
2747
2748                 spin_lock_bh(&addrconf_hash_lock);
2749         restart:
2750                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2751                         if (ifa->idev == idev) {
2752                                 hlist_del_init_rcu(&ifa->addr_lst);
2753                                 addrconf_del_timer(ifa);
2754                                 goto restart;
2755                         }
2756                 }
2757                 spin_unlock_bh(&addrconf_hash_lock);
2758         }
2759
2760         write_lock_bh(&idev->lock);
2761
2762         /* Step 2: clear flags for stateless addrconf */
2763         if (!how)
2764                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2765
2766 #ifdef CONFIG_IPV6_PRIVACY
2767         if (how && del_timer(&idev->regen_timer))
2768                 in6_dev_put(idev);
2769
2770         /* Step 3: clear tempaddr list */
2771         while (!list_empty(&idev->tempaddr_list)) {
2772                 ifa = list_first_entry(&idev->tempaddr_list,
2773                                        struct inet6_ifaddr, tmp_list);
2774                 list_del(&ifa->tmp_list);
2775                 write_unlock_bh(&idev->lock);
2776                 spin_lock_bh(&ifa->lock);
2777
2778                 if (ifa->ifpub) {
2779                         in6_ifa_put(ifa->ifpub);
2780                         ifa->ifpub = NULL;
2781                 }
2782                 spin_unlock_bh(&ifa->lock);
2783                 in6_ifa_put(ifa);
2784                 write_lock_bh(&idev->lock);
2785         }
2786 #endif
2787
2788         while (!list_empty(&idev->addr_list)) {
2789                 ifa = list_first_entry(&idev->addr_list,
2790                                        struct inet6_ifaddr, if_list);
2791                 addrconf_del_timer(ifa);
2792
2793                 list_del(&ifa->if_list);
2794
2795                 write_unlock_bh(&idev->lock);
2796
2797                 spin_lock_bh(&ifa->state_lock);
2798                 state = ifa->state;
2799                 ifa->state = INET6_IFADDR_STATE_DEAD;
2800                 spin_unlock_bh(&ifa->state_lock);
2801
2802                 if (state != INET6_IFADDR_STATE_DEAD) {
2803                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2804                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2805                 }
2806                 in6_ifa_put(ifa);
2807
2808                 write_lock_bh(&idev->lock);
2809         }
2810
2811         write_unlock_bh(&idev->lock);
2812
2813         /* Step 5: Discard multicast list */
2814         if (how)
2815                 ipv6_mc_destroy_dev(idev);
2816         else
2817                 ipv6_mc_down(idev);
2818
2819         idev->tstamp = jiffies;
2820
2821         /* Last: Shot the device (if unregistered) */
2822         if (how) {
2823                 addrconf_sysctl_unregister(idev);
2824                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2825                 neigh_ifdown(&nd_tbl, dev);
2826                 in6_dev_put(idev);
2827         }
2828         return 0;
2829 }
2830
2831 static void addrconf_rs_timer(unsigned long data)
2832 {
2833         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2834         struct inet6_dev *idev = ifp->idev;
2835
2836         read_lock(&idev->lock);
2837         if (idev->dead || !(idev->if_flags & IF_READY))
2838                 goto out;
2839
2840         if (idev->cnf.forwarding)
2841                 goto out;
2842
2843         /* Announcement received after solicitation was sent */
2844         if (idev->if_flags & IF_RA_RCVD)
2845                 goto out;
2846
2847         spin_lock(&ifp->lock);
2848         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2849                 /* The wait after the last probe can be shorter */
2850                 addrconf_mod_timer(ifp, AC_RS,
2851                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2852                                    idev->cnf.rtr_solicit_delay :
2853                                    idev->cnf.rtr_solicit_interval);
2854                 spin_unlock(&ifp->lock);
2855
2856                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2857         } else {
2858                 spin_unlock(&ifp->lock);
2859                 /*
2860                  * Note: we do not support deprecated "all on-link"
2861                  * assumption any longer.
2862                  */
2863                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2864                        idev->dev->name);
2865         }
2866
2867 out:
2868         read_unlock(&idev->lock);
2869         in6_ifa_put(ifp);
2870 }
2871
2872 /*
2873  *      Duplicate Address Detection
2874  */
2875 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2876 {
2877         unsigned long rand_num;
2878         struct inet6_dev *idev = ifp->idev;
2879
2880         if (ifp->flags & IFA_F_OPTIMISTIC)
2881                 rand_num = 0;
2882         else
2883                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2884
2885         ifp->probes = idev->cnf.dad_transmits;
2886         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2887 }
2888
2889 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2890 {
2891         struct inet6_dev *idev = ifp->idev;
2892         struct net_device *dev = idev->dev;
2893
2894         addrconf_join_solict(dev, &ifp->addr);
2895
2896         net_srandom(ifp->addr.s6_addr32[3]);
2897
2898         read_lock_bh(&idev->lock);
2899         spin_lock(&ifp->lock);
2900         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2901                 goto out;
2902
2903         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2904             idev->cnf.accept_dad < 1 ||
2905             !(ifp->flags&IFA_F_TENTATIVE) ||
2906             ifp->flags & IFA_F_NODAD) {
2907                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2908                 spin_unlock(&ifp->lock);
2909                 read_unlock_bh(&idev->lock);
2910
2911                 addrconf_dad_completed(ifp);
2912                 return;
2913         }
2914
2915         if (!(idev->if_flags & IF_READY)) {
2916                 spin_unlock(&ifp->lock);
2917                 read_unlock_bh(&idev->lock);
2918                 /*
2919                  * If the device is not ready:
2920                  * - keep it tentative if it is a permanent address.
2921                  * - otherwise, kill it.
2922                  */
2923                 in6_ifa_hold(ifp);
2924                 addrconf_dad_stop(ifp, 0);
2925                 return;
2926         }
2927
2928         /*
2929          * Optimistic nodes can start receiving
2930          * Frames right away
2931          */
2932         if (ifp->flags & IFA_F_OPTIMISTIC)
2933                 ip6_ins_rt(ifp->rt);
2934
2935         addrconf_dad_kick(ifp);
2936 out:
2937         spin_unlock(&ifp->lock);
2938         read_unlock_bh(&idev->lock);
2939 }
2940
2941 static void addrconf_dad_timer(unsigned long data)
2942 {
2943         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2944         struct inet6_dev *idev = ifp->idev;
2945         struct in6_addr mcaddr;
2946
2947         if (!ifp->probes && addrconf_dad_end(ifp))
2948                 goto out;
2949
2950         read_lock(&idev->lock);
2951         if (idev->dead || !(idev->if_flags & IF_READY)) {
2952                 read_unlock(&idev->lock);
2953                 goto out;
2954         }
2955
2956         spin_lock(&ifp->lock);
2957         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2958                 spin_unlock(&ifp->lock);
2959                 read_unlock(&idev->lock);
2960                 goto out;
2961         }
2962
2963         if (ifp->probes == 0) {
2964                 /*
2965                  * DAD was successful
2966                  */
2967
2968                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2969                 spin_unlock(&ifp->lock);
2970                 read_unlock(&idev->lock);
2971
2972                 addrconf_dad_completed(ifp);
2973
2974                 goto out;
2975         }
2976
2977         ifp->probes--;
2978         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2979         spin_unlock(&ifp->lock);
2980         read_unlock(&idev->lock);
2981
2982         /* send a neighbour solicitation for our addr */
2983         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2984         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2985 out:
2986         in6_ifa_put(ifp);
2987 }
2988
2989 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2990 {
2991         struct net_device *dev = ifp->idev->dev;
2992
2993         /*
2994          *      Configure the address for reception. Now it is valid.
2995          */
2996
2997         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2998
2999         /* If added prefix is link local and we are prepared to process
3000            router advertisements, start sending router solicitations.
3001          */
3002
3003         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3004              ifp->idev->cnf.accept_ra == 2) &&
3005             ifp->idev->cnf.rtr_solicits > 0 &&
3006             (dev->flags&IFF_LOOPBACK) == 0 &&
3007             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3008                 /*
3009                  *      If a host as already performed a random delay
3010                  *      [...] as part of DAD [...] there is no need
3011                  *      to delay again before sending the first RS
3012                  */
3013                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3014
3015                 spin_lock_bh(&ifp->lock);
3016                 ifp->probes = 1;
3017                 ifp->idev->if_flags |= IF_RS_SENT;
3018                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3019                 spin_unlock_bh(&ifp->lock);
3020         }
3021 }
3022
3023 static void addrconf_dad_run(struct inet6_dev *idev)
3024 {
3025         struct inet6_ifaddr *ifp;
3026
3027         read_lock_bh(&idev->lock);
3028         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3029                 spin_lock(&ifp->lock);
3030                 if (ifp->flags & IFA_F_TENTATIVE &&
3031                     ifp->state == INET6_IFADDR_STATE_DAD)
3032                         addrconf_dad_kick(ifp);
3033                 spin_unlock(&ifp->lock);
3034         }
3035         read_unlock_bh(&idev->lock);
3036 }
3037
3038 #ifdef CONFIG_PROC_FS
3039 struct if6_iter_state {
3040         struct seq_net_private p;
3041         int bucket;
3042 };
3043
3044 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3045 {
3046         struct inet6_ifaddr *ifa = NULL;
3047         struct if6_iter_state *state = seq->private;
3048         struct net *net = seq_file_net(seq);
3049
3050         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3051                 struct hlist_node *n;
3052                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3053                                          addr_lst)
3054                         if (net_eq(dev_net(ifa->idev->dev), net))
3055                                 return ifa;
3056         }
3057         return NULL;
3058 }
3059
3060 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3061                                          struct inet6_ifaddr *ifa)
3062 {
3063         struct if6_iter_state *state = seq->private;
3064         struct net *net = seq_file_net(seq);
3065         struct hlist_node *n = &ifa->addr_lst;
3066
3067         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3068                 if (net_eq(dev_net(ifa->idev->dev), net))
3069                         return ifa;
3070
3071         while (++state->bucket < IN6_ADDR_HSIZE) {
3072                 hlist_for_each_entry_rcu_bh(ifa, n,
3073                                      &inet6_addr_lst[state->bucket], addr_lst) {
3074                         if (net_eq(dev_net(ifa->idev->dev), net))
3075                                 return ifa;
3076                 }
3077         }
3078
3079         return NULL;
3080 }
3081
3082 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3083 {
3084         struct inet6_ifaddr *ifa = if6_get_first(seq);
3085
3086         if (ifa)
3087                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3088                         --pos;
3089         return pos ? NULL : ifa;
3090 }
3091
3092 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3093         __acquires(rcu_bh)
3094 {
3095         rcu_read_lock_bh();
3096         return if6_get_idx(seq, *pos);
3097 }
3098
3099 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3100 {
3101         struct inet6_ifaddr *ifa;
3102
3103         ifa = if6_get_next(seq, v);
3104         ++*pos;
3105         return ifa;
3106 }
3107
3108 static void if6_seq_stop(struct seq_file *seq, void *v)
3109         __releases(rcu_bh)
3110 {
3111         rcu_read_unlock_bh();
3112 }
3113
3114 static int if6_seq_show(struct seq_file *seq, void *v)
3115 {
3116         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3117         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3118                    &ifp->addr,
3119                    ifp->idev->dev->ifindex,
3120                    ifp->prefix_len,
3121                    ifp->scope,
3122                    ifp->flags,
3123                    ifp->idev->dev->name);
3124         return 0;
3125 }
3126
3127 static const struct seq_operations if6_seq_ops = {
3128         .start  = if6_seq_start,
3129         .next   = if6_seq_next,
3130         .show   = if6_seq_show,
3131         .stop   = if6_seq_stop,
3132 };
3133
3134 static int if6_seq_open(struct inode *inode, struct file *file)
3135 {
3136         return seq_open_net(inode, file, &if6_seq_ops,
3137                             sizeof(struct if6_iter_state));
3138 }
3139
3140 static const struct file_operations if6_fops = {
3141         .owner          = THIS_MODULE,
3142         .open           = if6_seq_open,
3143         .read           = seq_read,
3144         .llseek         = seq_lseek,
3145         .release        = seq_release_net,
3146 };
3147
3148 static int __net_init if6_proc_net_init(struct net *net)
3149 {
3150         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3151                 return -ENOMEM;
3152         return 0;
3153 }
3154
3155 static void __net_exit if6_proc_net_exit(struct net *net)
3156 {
3157        proc_net_remove(net, "if_inet6");
3158 }
3159
3160 static struct pernet_operations if6_proc_net_ops = {
3161        .init = if6_proc_net_init,
3162        .exit = if6_proc_net_exit,
3163 };
3164
3165 int __init if6_proc_init(void)
3166 {
3167         return register_pernet_subsys(&if6_proc_net_ops);
3168 }
3169
3170 void if6_proc_exit(void)
3171 {
3172         unregister_pernet_subsys(&if6_proc_net_ops);
3173 }
3174 #endif  /* CONFIG_PROC_FS */
3175
3176 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3177 /* Check if address is a home address configured on any interface. */
3178 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3179 {
3180         int ret = 0;
3181         struct inet6_ifaddr *ifp = NULL;
3182         struct hlist_node *n;
3183         unsigned int hash = ipv6_addr_hash(addr);
3184
3185         rcu_read_lock_bh();
3186         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3187                 if (!net_eq(dev_net(ifp->idev->dev), net))
3188                         continue;
3189                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3190                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3191                         ret = 1;
3192                         break;
3193                 }
3194         }
3195         rcu_read_unlock_bh();
3196         return ret;
3197 }
3198 #endif
3199
3200 /*
3201  *      Periodic address status verification
3202  */
3203
3204 static void addrconf_verify(unsigned long foo)
3205 {
3206         unsigned long now, next, next_sec, next_sched;
3207         struct inet6_ifaddr *ifp;
3208         struct hlist_node *node;
3209         int i;
3210
3211         rcu_read_lock_bh();
3212         spin_lock(&addrconf_verify_lock);
3213         now = jiffies;
3214         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3215
3216         del_timer(&addr_chk_timer);
3217
3218         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3219 restart:
3220                 hlist_for_each_entry_rcu_bh(ifp, node,
3221                                          &inet6_addr_lst[i], addr_lst) {
3222                         unsigned long age;
3223
3224                         if (ifp->flags & IFA_F_PERMANENT)
3225                                 continue;
3226
3227                         spin_lock(&ifp->lock);
3228                         /* We try to batch several events at once. */
3229                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3230
3231                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3232                             age >= ifp->valid_lft) {
3233                                 spin_unlock(&ifp->lock);
3234                                 in6_ifa_hold(ifp);
3235                                 ipv6_del_addr(ifp);
3236                                 goto restart;
3237                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3238                                 spin_unlock(&ifp->lock);
3239                                 continue;
3240                         } else if (age >= ifp->prefered_lft) {
3241                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3242                                 int deprecate = 0;
3243
3244                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3245                                         deprecate = 1;
3246                                         ifp->flags |= IFA_F_DEPRECATED;
3247                                 }
3248
3249                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3250                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3251
3252                                 spin_unlock(&ifp->lock);
3253
3254                                 if (deprecate) {
3255                                         in6_ifa_hold(ifp);
3256
3257                                         ipv6_ifa_notify(0, ifp);
3258                                         in6_ifa_put(ifp);
3259                                         goto restart;
3260                                 }
3261 #ifdef CONFIG_IPV6_PRIVACY
3262                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3263                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3264                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3265                                         ifp->idev->cnf.dad_transmits *
3266                                         ifp->idev->nd_parms->retrans_time / HZ;
3267
3268                                 if (age >= ifp->prefered_lft - regen_advance) {
3269                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3270                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3271                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3272                                         if (!ifp->regen_count && ifpub) {
3273                                                 ifp->regen_count++;
3274                                                 in6_ifa_hold(ifp);
3275                                                 in6_ifa_hold(ifpub);
3276                                                 spin_unlock(&ifp->lock);
3277
3278                                                 spin_lock(&ifpub->lock);
3279                                                 ifpub->regen_count = 0;
3280                                                 spin_unlock(&ifpub->lock);
3281                                                 ipv6_create_tempaddr(ifpub, ifp);
3282                                                 in6_ifa_put(ifpub);
3283                                                 in6_ifa_put(ifp);
3284                                                 goto restart;
3285                                         }
3286                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3287                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3288                                 spin_unlock(&ifp->lock);
3289 #endif
3290                         } else {
3291                                 /* ifp->prefered_lft <= ifp->valid_lft */
3292                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3293                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3294                                 spin_unlock(&ifp->lock);
3295                         }
3296                 }
3297         }
3298
3299         next_sec = round_jiffies_up(next);
3300         next_sched = next;
3301
3302         /* If rounded timeout is accurate enough, accept it. */
3303         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3304                 next_sched = next_sec;
3305
3306         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3307         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3308                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3309
3310         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3311               now, next, next_sec, next_sched));
3312
3313         addr_chk_timer.expires = next_sched;
3314         add_timer(&addr_chk_timer);
3315         spin_unlock(&addrconf_verify_lock);
3316         rcu_read_unlock_bh();
3317 }
3318
3319 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3320 {
3321         struct in6_addr *pfx = NULL;
3322
3323         if (addr)
3324                 pfx = nla_data(addr);
3325
3326         if (local) {
3327                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3328                         pfx = NULL;
3329                 else
3330                         pfx = nla_data(local);
3331         }
3332
3333         return pfx;
3334 }
3335
3336 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3337         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3338         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3339         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3340 };
3341
3342 static int
3343 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3344 {
3345         struct net *net = sock_net(skb->sk);
3346         struct ifaddrmsg *ifm;
3347         struct nlattr *tb[IFA_MAX+1];
3348         struct in6_addr *pfx;
3349         int err;
3350
3351         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3352         if (err < 0)
3353                 return err;
3354
3355         ifm = nlmsg_data(nlh);
3356         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3357         if (pfx == NULL)
3358                 return -EINVAL;
3359
3360         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3361 }
3362
3363 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3364                              u32 prefered_lft, u32 valid_lft)
3365 {
3366         u32 flags;
3367         clock_t expires;
3368         unsigned long timeout;
3369
3370         if (!valid_lft || (prefered_lft > valid_lft))
3371                 return -EINVAL;
3372
3373         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3374         if (addrconf_finite_timeout(timeout)) {
3375                 expires = jiffies_to_clock_t(timeout * HZ);
3376                 valid_lft = timeout;
3377                 flags = RTF_EXPIRES;
3378         } else {
3379                 expires = 0;
3380                 flags = 0;
3381                 ifa_flags |= IFA_F_PERMANENT;
3382         }
3383
3384         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3385         if (addrconf_finite_timeout(timeout)) {
3386                 if (timeout == 0)
3387                         ifa_flags |= IFA_F_DEPRECATED;
3388                 prefered_lft = timeout;
3389         }
3390
3391         spin_lock_bh(&ifp->lock);
3392         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3393         ifp->tstamp = jiffies;
3394         ifp->valid_lft = valid_lft;
3395         ifp->prefered_lft = prefered_lft;
3396
3397         spin_unlock_bh(&ifp->lock);
3398         if (!(ifp->flags&IFA_F_TENTATIVE))
3399                 ipv6_ifa_notify(0, ifp);
3400
3401         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3402                               expires, flags);
3403         addrconf_verify(0);
3404
3405         return 0;
3406 }
3407
3408 static int
3409 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3410 {
3411         struct net *net = sock_net(skb->sk);
3412         struct ifaddrmsg *ifm;
3413         struct nlattr *tb[IFA_MAX+1];
3414         struct in6_addr *pfx;
3415         struct inet6_ifaddr *ifa;
3416         struct net_device *dev;
3417         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3418         u8 ifa_flags;
3419         int err;
3420
3421         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3422         if (err < 0)
3423                 return err;
3424
3425         ifm = nlmsg_data(nlh);
3426         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3427         if (pfx == NULL)
3428                 return -EINVAL;
3429
3430         if (tb[IFA_CACHEINFO]) {
3431                 struct ifa_cacheinfo *ci;
3432
3433                 ci = nla_data(tb[IFA_CACHEINFO]);
3434                 valid_lft = ci->ifa_valid;
3435                 preferred_lft = ci->ifa_prefered;
3436         } else {
3437                 preferred_lft = INFINITY_LIFE_TIME;
3438                 valid_lft = INFINITY_LIFE_TIME;
3439         }
3440
3441         dev =  __dev_get_by_index(net, ifm->ifa_index);
3442         if (dev == NULL)
3443                 return -ENODEV;
3444
3445         /* We ignore other flags so far. */
3446         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3447
3448         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3449         if (ifa == NULL) {
3450                 /*
3451                  * It would be best to check for !NLM_F_CREATE here but
3452                  * userspace alreay relies on not having to provide this.
3453                  */
3454                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3455                                       ifm->ifa_prefixlen, ifa_flags,
3456                                       preferred_lft, valid_lft);
3457         }
3458
3459         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3460             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3461                 err = -EEXIST;
3462         else
3463                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3464
3465         in6_ifa_put(ifa);
3466
3467         return err;
3468 }
3469
3470 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3471                           u8 scope, int ifindex)
3472 {
3473         struct ifaddrmsg *ifm;
3474
3475         ifm = nlmsg_data(nlh);
3476         ifm->ifa_family = AF_INET6;
3477         ifm->ifa_prefixlen = prefixlen;
3478         ifm->ifa_flags = flags;
3479         ifm->ifa_scope = scope;
3480         ifm->ifa_index = ifindex;
3481 }
3482
3483 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3484                          unsigned long tstamp, u32 preferred, u32 valid)
3485 {
3486         struct ifa_cacheinfo ci;
3487
3488         ci.cstamp = cstamp_delta(cstamp);
3489         ci.tstamp = cstamp_delta(tstamp);
3490         ci.ifa_prefered = preferred;
3491         ci.ifa_valid = valid;
3492
3493         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3494 }
3495
3496 static inline int rt_scope(int ifa_scope)
3497 {
3498         if (ifa_scope & IFA_HOST)
3499                 return RT_SCOPE_HOST;
3500         else if (ifa_scope & IFA_LINK)
3501                 return RT_SCOPE_LINK;
3502         else if (ifa_scope & IFA_SITE)
3503                 return RT_SCOPE_SITE;
3504         else
3505                 return RT_SCOPE_UNIVERSE;
3506 }
3507
3508 static inline int inet6_ifaddr_msgsize(void)
3509 {
3510         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3511                + nla_total_size(16) /* IFA_ADDRESS */
3512                + nla_total_size(sizeof(struct ifa_cacheinfo));
3513 }
3514
3515 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3516                              u32 pid, u32 seq, int event, unsigned int flags)
3517 {
3518         struct nlmsghdr  *nlh;
3519         u32 preferred, valid;
3520
3521         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3522         if (nlh == NULL)
3523                 return -EMSGSIZE;
3524
3525         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3526                       ifa->idev->dev->ifindex);
3527
3528         if (!(ifa->flags&IFA_F_PERMANENT)) {
3529                 preferred = ifa->prefered_lft;
3530                 valid = ifa->valid_lft;
3531                 if (preferred != INFINITY_LIFE_TIME) {
3532                         long tval = (jiffies - ifa->tstamp)/HZ;
3533                         if (preferred > tval)
3534                                 preferred -= tval;
3535                         else
3536                                 preferred = 0;
3537                         if (valid != INFINITY_LIFE_TIME) {
3538                                 if (valid > tval)
3539                                         valid -= tval;
3540                                 else
3541                                         valid = 0;
3542                         }
3543                 }
3544         } else {
3545                 preferred = INFINITY_LIFE_TIME;
3546                 valid = INFINITY_LIFE_TIME;
3547         }
3548
3549         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3550             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3551                 nlmsg_cancel(skb, nlh);
3552                 return -EMSGSIZE;
3553         }
3554
3555         return nlmsg_end(skb, nlh);
3556 }
3557
3558 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3559                                 u32 pid, u32 seq, int event, u16 flags)
3560 {
3561         struct nlmsghdr  *nlh;
3562         u8 scope = RT_SCOPE_UNIVERSE;
3563         int ifindex = ifmca->idev->dev->ifindex;
3564
3565         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3566                 scope = RT_SCOPE_SITE;
3567
3568         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3569         if (nlh == NULL)
3570                 return -EMSGSIZE;
3571
3572         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3573         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3574             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3575                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3576                 nlmsg_cancel(skb, nlh);
3577                 return -EMSGSIZE;
3578         }
3579
3580         return nlmsg_end(skb, nlh);
3581 }
3582
3583 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3584                                 u32 pid, u32 seq, int event, unsigned int flags)
3585 {
3586         struct nlmsghdr  *nlh;
3587         u8 scope = RT_SCOPE_UNIVERSE;
3588         int ifindex = ifaca->aca_idev->dev->ifindex;
3589
3590         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3591                 scope = RT_SCOPE_SITE;
3592
3593         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3594         if (nlh == NULL)
3595                 return -EMSGSIZE;
3596
3597         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3598         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3599             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3600                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3601                 nlmsg_cancel(skb, nlh);
3602                 return -EMSGSIZE;
3603         }
3604
3605         return nlmsg_end(skb, nlh);
3606 }
3607
3608 enum addr_type_t {
3609         UNICAST_ADDR,
3610         MULTICAST_ADDR,
3611         ANYCAST_ADDR,
3612 };
3613
3614 /* called with rcu_read_lock() */
3615 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3616                           struct netlink_callback *cb, enum addr_type_t type,
3617                           int s_ip_idx, int *p_ip_idx)
3618 {
3619         struct ifmcaddr6 *ifmca;
3620         struct ifacaddr6 *ifaca;
3621         int err = 1;
3622         int ip_idx = *p_ip_idx;
3623
3624         read_lock_bh(&idev->lock);
3625         switch (type) {
3626         case UNICAST_ADDR: {
3627                 struct inet6_ifaddr *ifa;
3628
3629                 /* unicast address incl. temp addr */
3630                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3631                         if (++ip_idx < s_ip_idx)
3632                                 continue;
3633                         err = inet6_fill_ifaddr(skb, ifa,
3634                                                 NETLINK_CB(cb->skb).pid,
3635                                                 cb->nlh->nlmsg_seq,
3636                                                 RTM_NEWADDR,
3637                                                 NLM_F_MULTI);
3638                         if (err <= 0)
3639                                 break;
3640                 }
3641                 break;
3642         }
3643         case MULTICAST_ADDR:
3644                 /* multicast address */
3645                 for (ifmca = idev->mc_list; ifmca;
3646                      ifmca = ifmca->next, ip_idx++) {
3647                         if (ip_idx < s_ip_idx)
3648                                 continue;
3649                         err = inet6_fill_ifmcaddr(skb, ifmca,
3650                                                   NETLINK_CB(cb->skb).pid,
3651                                                   cb->nlh->nlmsg_seq,
3652                                                   RTM_GETMULTICAST,
3653                                                   NLM_F_MULTI);
3654                         if (err <= 0)
3655                                 break;
3656                 }
3657                 break;
3658         case ANYCAST_ADDR:
3659                 /* anycast address */
3660                 for (ifaca = idev->ac_list; ifaca;
3661                      ifaca = ifaca->aca_next, ip_idx++) {
3662                         if (ip_idx < s_ip_idx)
3663                                 continue;
3664                         err = inet6_fill_ifacaddr(skb, ifaca,
3665                                                   NETLINK_CB(cb->skb).pid,
3666                                                   cb->nlh->nlmsg_seq,
3667                                                   RTM_GETANYCAST,
3668                                                   NLM_F_MULTI);
3669                         if (err <= 0)
3670                                 break;
3671                 }
3672                 break;
3673         default:
3674                 break;
3675         }
3676         read_unlock_bh(&idev->lock);
3677         *p_ip_idx = ip_idx;
3678         return err;
3679 }
3680
3681 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3682                            enum addr_type_t type)
3683 {
3684         struct net *net = sock_net(skb->sk);
3685         int h, s_h;
3686         int idx, ip_idx;
3687         int s_idx, s_ip_idx;
3688         struct net_device *dev;
3689         struct inet6_dev *idev;
3690         struct hlist_head *head;
3691         struct hlist_node *node;
3692
3693         s_h = cb->args[0];
3694         s_idx = idx = cb->args[1];
3695         s_ip_idx = ip_idx = cb->args[2];
3696
3697         rcu_read_lock();
3698         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3699                 idx = 0;
3700                 head = &net->dev_index_head[h];
3701                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3702                         if (idx < s_idx)
3703                                 goto cont;
3704                         if (h > s_h || idx > s_idx)
3705                                 s_ip_idx = 0;
3706                         ip_idx = 0;
3707                         idev = __in6_dev_get(dev);
3708                         if (!idev)
3709                                 goto cont;
3710
3711                         if (in6_dump_addrs(idev, skb, cb, type,
3712                                            s_ip_idx, &ip_idx) <= 0)
3713                                 goto done;
3714 cont:
3715                         idx++;
3716                 }
3717         }
3718 done:
3719         rcu_read_unlock();
3720         cb->args[0] = h;
3721         cb->args[1] = idx;
3722         cb->args[2] = ip_idx;
3723
3724         return skb->len;
3725 }
3726
3727 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3728 {
3729         enum addr_type_t type = UNICAST_ADDR;
3730
3731         return inet6_dump_addr(skb, cb, type);
3732 }
3733
3734 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3735 {
3736         enum addr_type_t type = MULTICAST_ADDR;
3737
3738         return inet6_dump_addr(skb, cb, type);
3739 }
3740
3741
3742 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3743 {
3744         enum addr_type_t type = ANYCAST_ADDR;
3745
3746         return inet6_dump_addr(skb, cb, type);
3747 }
3748
3749 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3750                              void *arg)
3751 {
3752         struct net *net = sock_net(in_skb->sk);
3753         struct ifaddrmsg *ifm;
3754         struct nlattr *tb[IFA_MAX+1];
3755         struct in6_addr *addr = NULL;
3756         struct net_device *dev = NULL;
3757         struct inet6_ifaddr *ifa;
3758         struct sk_buff *skb;
3759         int err;
3760
3761         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3762         if (err < 0)
3763                 goto errout;
3764
3765         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3766         if (addr == NULL) {
3767                 err = -EINVAL;
3768                 goto errout;
3769         }
3770
3771         ifm = nlmsg_data(nlh);
3772         if (ifm->ifa_index)
3773                 dev = __dev_get_by_index(net, ifm->ifa_index);
3774
3775         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3776         if (!ifa) {
3777                 err = -EADDRNOTAVAIL;
3778                 goto errout;
3779         }
3780
3781         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3782         if (!skb) {
3783                 err = -ENOBUFS;
3784                 goto errout_ifa;
3785         }
3786
3787         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3788                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3789         if (err < 0) {
3790                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3791                 WARN_ON(err == -EMSGSIZE);
3792                 kfree_skb(skb);
3793                 goto errout_ifa;
3794         }
3795         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3796 errout_ifa:
3797         in6_ifa_put(ifa);
3798 errout:
3799         return err;
3800 }
3801
3802 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3803 {
3804         struct sk_buff *skb;
3805         struct net *net = dev_net(ifa->idev->dev);
3806         int err = -ENOBUFS;
3807
3808         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3809         if (skb == NULL)
3810                 goto errout;
3811
3812         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3813         if (err < 0) {
3814                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3815                 WARN_ON(err == -EMSGSIZE);
3816                 kfree_skb(skb);
3817                 goto errout;
3818         }
3819         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3820         return;
3821 errout:
3822         if (err < 0)
3823                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3824 }
3825
3826 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3827                                 __s32 *array, int bytes)
3828 {
3829         BUG_ON(bytes < (DEVCONF_MAX * 4));
3830
3831         memset(array, 0, bytes);
3832         array[DEVCONF_FORWARDING] = cnf->forwarding;
3833         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3834         array[DEVCONF_MTU6] = cnf->mtu6;
3835         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3836         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3837         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3838         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3839         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3840         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3841                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3842         array[DEVCONF_RTR_SOLICIT_DELAY] =
3843                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3844         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3845 #ifdef CONFIG_IPV6_PRIVACY
3846         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3847         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3848         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3849         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3850         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3851 #endif
3852         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3853         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3854         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3855 #ifdef CONFIG_IPV6_ROUTER_PREF
3856         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3857         array[DEVCONF_RTR_PROBE_INTERVAL] =
3858                 jiffies_to_msecs(cnf->rtr_probe_interval);
3859 #ifdef CONFIG_IPV6_ROUTE_INFO
3860         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3861 #endif
3862 #endif
3863         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3864         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3865 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3866         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3867 #endif
3868 #ifdef CONFIG_IPV6_MROUTE
3869         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3870 #endif
3871         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3872         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3873         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3874 }
3875
3876 static inline size_t inet6_ifla6_size(void)
3877 {
3878         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3879              + nla_total_size(sizeof(struct ifla_cacheinfo))
3880              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3881              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3882              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3883 }
3884
3885 static inline size_t inet6_if_nlmsg_size(void)
3886 {
3887         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3888                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3889                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3890                + nla_total_size(4) /* IFLA_MTU */
3891                + nla_total_size(4) /* IFLA_LINK */
3892                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3893 }
3894
3895 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3896                                       int items, int bytes)
3897 {
3898         int i;
3899         int pad = bytes - sizeof(u64) * items;
3900         BUG_ON(pad < 0);
3901
3902         /* Use put_unaligned() because stats may not be aligned for u64. */
3903         put_unaligned(items, &stats[0]);
3904         for (i = 1; i < items; i++)
3905                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3906
3907         memset(&stats[items], 0, pad);
3908 }
3909
3910 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3911                                       int items, int bytes, size_t syncpoff)
3912 {
3913         int i;
3914         int pad = bytes - sizeof(u64) * items;
3915         BUG_ON(pad < 0);
3916
3917         /* Use put_unaligned() because stats may not be aligned for u64. */
3918         put_unaligned(items, &stats[0]);
3919         for (i = 1; i < items; i++)
3920                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3921
3922         memset(&stats[items], 0, pad);
3923 }
3924
3925 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3926                              int bytes)
3927 {
3928         switch (attrtype) {
3929         case IFLA_INET6_STATS:
3930                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3931                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3932                 break;
3933         case IFLA_INET6_ICMP6STATS:
3934                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3935                 break;
3936         }
3937 }
3938
3939 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3940 {
3941         struct nlattr *nla;
3942         struct ifla_cacheinfo ci;
3943
3944         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3945
3946         ci.max_reasm_len = IPV6_MAXPLEN;
3947         ci.tstamp = cstamp_delta(idev->tstamp);
3948         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3949         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3950         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3951
3952         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3953         if (nla == NULL)
3954                 goto nla_put_failure;
3955         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3956
3957         /* XXX - MC not implemented */
3958
3959         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3960         if (nla == NULL)
3961                 goto nla_put_failure;
3962         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3963
3964         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3965         if (nla == NULL)
3966                 goto nla_put_failure;
3967         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3968
3969         return 0;
3970
3971 nla_put_failure:
3972         return -EMSGSIZE;
3973 }
3974
3975 static size_t inet6_get_link_af_size(const struct net_device *dev)
3976 {
3977         if (!__in6_dev_get(dev))
3978                 return 0;
3979
3980         return inet6_ifla6_size();
3981 }
3982
3983 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3984 {
3985         struct inet6_dev *idev = __in6_dev_get(dev);
3986
3987         if (!idev)
3988                 return -ENODATA;
3989
3990         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3991                 return -EMSGSIZE;
3992
3993         return 0;
3994 }
3995
3996 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3997                              u32 pid, u32 seq, int event, unsigned int flags)
3998 {
3999         struct net_device *dev = idev->dev;
4000         struct ifinfomsg *hdr;
4001         struct nlmsghdr *nlh;
4002         void *protoinfo;
4003
4004         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4005         if (nlh == NULL)
4006                 return -EMSGSIZE;
4007
4008         hdr = nlmsg_data(nlh);
4009         hdr->ifi_family = AF_INET6;
4010         hdr->__ifi_pad = 0;
4011         hdr->ifi_type = dev->type;
4012         hdr->ifi_index = dev->ifindex;
4013         hdr->ifi_flags = dev_get_flags(dev);
4014         hdr->ifi_change = 0;
4015
4016         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4017
4018         if (dev->addr_len)
4019                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4020
4021         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4022         if (dev->ifindex != dev->iflink)
4023                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4024
4025         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4026         if (protoinfo == NULL)
4027                 goto nla_put_failure;
4028
4029         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4030                 goto nla_put_failure;
4031
4032         nla_nest_end(skb, protoinfo);
4033         return nlmsg_end(skb, nlh);
4034
4035 nla_put_failure:
4036         nlmsg_cancel(skb, nlh);
4037         return -EMSGSIZE;
4038 }
4039
4040 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4041 {
4042         struct net *net = sock_net(skb->sk);
4043         int h, s_h;
4044         int idx = 0, s_idx;
4045         struct net_device *dev;
4046         struct inet6_dev *idev;
4047         struct hlist_head *head;
4048         struct hlist_node *node;
4049
4050         s_h = cb->args[0];
4051         s_idx = cb->args[1];
4052
4053         rcu_read_lock();
4054         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4055                 idx = 0;
4056                 head = &net->dev_index_head[h];
4057                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4058                         if (idx < s_idx)
4059                                 goto cont;
4060                         idev = __in6_dev_get(dev);
4061                         if (!idev)
4062                                 goto cont;
4063                         if (inet6_fill_ifinfo(skb, idev,
4064                                               NETLINK_CB(cb->skb).pid,
4065                                               cb->nlh->nlmsg_seq,
4066                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4067                                 goto out;
4068 cont:
4069                         idx++;
4070                 }
4071         }
4072 out:
4073         rcu_read_unlock();
4074         cb->args[1] = idx;
4075         cb->args[0] = h;
4076
4077         return skb->len;
4078 }
4079
4080 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4081 {
4082         struct sk_buff *skb;
4083         struct net *net = dev_net(idev->dev);
4084         int err = -ENOBUFS;
4085
4086         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4087         if (skb == NULL)
4088                 goto errout;
4089
4090         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4091         if (err < 0) {
4092                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4093                 WARN_ON(err == -EMSGSIZE);
4094                 kfree_skb(skb);
4095                 goto errout;
4096         }
4097         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4098         return;
4099 errout:
4100         if (err < 0)
4101                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4102 }
4103
4104 static inline size_t inet6_prefix_nlmsg_size(void)
4105 {
4106         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4107                + nla_total_size(sizeof(struct in6_addr))
4108                + nla_total_size(sizeof(struct prefix_cacheinfo));
4109 }
4110
4111 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4112                              struct prefix_info *pinfo, u32 pid, u32 seq,
4113                              int event, unsigned int flags)
4114 {
4115         struct prefixmsg *pmsg;
4116         struct nlmsghdr *nlh;
4117         struct prefix_cacheinfo ci;
4118
4119         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4120         if (nlh == NULL)
4121                 return -EMSGSIZE;
4122
4123         pmsg = nlmsg_data(nlh);
4124         pmsg->prefix_family = AF_INET6;
4125         pmsg->prefix_pad1 = 0;
4126         pmsg->prefix_pad2 = 0;
4127         pmsg->prefix_ifindex = idev->dev->ifindex;
4128         pmsg->prefix_len = pinfo->prefix_len;
4129         pmsg->prefix_type = pinfo->type;
4130         pmsg->prefix_pad3 = 0;
4131         pmsg->prefix_flags = 0;
4132         if (pinfo->onlink)
4133                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4134         if (pinfo->autoconf)
4135                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4136
4137         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4138
4139         ci.preferred_time = ntohl(pinfo->prefered);
4140         ci.valid_time = ntohl(pinfo->valid);
4141         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4142
4143         return nlmsg_end(skb, nlh);
4144
4145 nla_put_failure:
4146         nlmsg_cancel(skb, nlh);
4147         return -EMSGSIZE;
4148 }
4149
4150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4151                          struct prefix_info *pinfo)
4152 {
4153         struct sk_buff *skb;
4154         struct net *net = dev_net(idev->dev);
4155         int err = -ENOBUFS;
4156
4157         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4158         if (skb == NULL)
4159                 goto errout;
4160
4161         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4162         if (err < 0) {
4163                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4164                 WARN_ON(err == -EMSGSIZE);
4165                 kfree_skb(skb);
4166                 goto errout;
4167         }
4168         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4169         return;
4170 errout:
4171         if (err < 0)
4172                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4173 }
4174
4175 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4176 {
4177         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4178
4179         switch (event) {
4180         case RTM_NEWADDR:
4181                 /*
4182                  * If the address was optimistic
4183                  * we inserted the route at the start of
4184                  * our DAD process, so we don't need
4185                  * to do it again
4186                  */
4187                 if (!(ifp->rt->rt6i_node))
4188                         ip6_ins_rt(ifp->rt);
4189                 if (ifp->idev->cnf.forwarding)
4190                         addrconf_join_anycast(ifp);
4191                 break;
4192         case RTM_DELADDR:
4193                 if (ifp->idev->cnf.forwarding)
4194                         addrconf_leave_anycast(ifp);
4195                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4196                 dst_hold(&ifp->rt->dst);
4197
4198                 if (ip6_del_rt(ifp->rt))
4199                         dst_free(&ifp->rt->dst);
4200                 break;
4201         }
4202 }
4203
4204 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4205 {
4206         rcu_read_lock_bh();
4207         if (likely(ifp->idev->dead == 0))
4208                 __ipv6_ifa_notify(event, ifp);
4209         rcu_read_unlock_bh();
4210 }
4211
4212 #ifdef CONFIG_SYSCTL
4213
4214 static
4215 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4216                            void __user *buffer, size_t *lenp, loff_t *ppos)
4217 {
4218         int *valp = ctl->data;
4219         int val = *valp;
4220         loff_t pos = *ppos;
4221         int ret;
4222
4223         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4224
4225         if (write)
4226                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4227         if (ret)
4228                 *ppos = pos;
4229         return ret;
4230 }
4231
4232 static void dev_disable_change(struct inet6_dev *idev)
4233 {
4234         if (!idev || !idev->dev)
4235                 return;
4236
4237         if (idev->cnf.disable_ipv6)
4238                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4239         else
4240                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4241 }
4242
4243 static void addrconf_disable_change(struct net *net, __s32 newf)
4244 {
4245         struct net_device *dev;
4246         struct inet6_dev *idev;
4247
4248         rcu_read_lock();
4249         for_each_netdev_rcu(net, dev) {
4250                 idev = __in6_dev_get(dev);
4251                 if (idev) {
4252                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4253                         idev->cnf.disable_ipv6 = newf;
4254                         if (changed)
4255                                 dev_disable_change(idev);
4256                 }
4257         }
4258         rcu_read_unlock();
4259 }
4260
4261 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4262 {
4263         struct net *net;
4264
4265         net = (struct net *)table->extra2;
4266
4267         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4268                 return 0;
4269
4270         if (!rtnl_trylock()) {
4271                 /* Restore the original values before restarting */
4272                 *p = old;
4273                 return restart_syscall();
4274         }
4275
4276         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4277                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4278                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4279                 addrconf_disable_change(net, newf);
4280         } else if ((!*p) ^ (!old))
4281                 dev_disable_change((struct inet6_dev *)table->extra1);
4282
4283         rtnl_unlock();
4284         return 0;
4285 }
4286
4287 static
4288 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4289                             void __user *buffer, size_t *lenp, loff_t *ppos)
4290 {
4291         int *valp = ctl->data;
4292         int val = *valp;
4293         loff_t pos = *ppos;
4294         int ret;
4295
4296         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4297
4298         if (write)
4299                 ret = addrconf_disable_ipv6(ctl, valp, val);
4300         if (ret)
4301                 *ppos = pos;
4302         return ret;
4303 }
4304
4305 static struct addrconf_sysctl_table
4306 {
4307         struct ctl_table_header *sysctl_header;
4308         ctl_table addrconf_vars[DEVCONF_MAX+1];
4309         char *dev_name;
4310 } addrconf_sysctl __read_mostly = {
4311         .sysctl_header = NULL,
4312         .addrconf_vars = {
4313                 {
4314                         .procname       = "forwarding",
4315                         .data           = &ipv6_devconf.forwarding,
4316                         .maxlen         = sizeof(int),
4317                         .mode           = 0644,
4318                         .proc_handler   = addrconf_sysctl_forward,
4319                 },
4320                 {
4321                         .procname       = "hop_limit",
4322                         .data           = &ipv6_devconf.hop_limit,
4323                         .maxlen         = sizeof(int),
4324                         .mode           = 0644,
4325                         .proc_handler   = proc_dointvec,
4326                 },
4327                 {
4328                         .procname       = "mtu",
4329                         .data           = &ipv6_devconf.mtu6,
4330                         .maxlen         = sizeof(int),
4331                         .mode           = 0644,
4332                         .proc_handler   = proc_dointvec,
4333                 },
4334                 {
4335                         .procname       = "accept_ra",
4336                         .data           = &ipv6_devconf.accept_ra,
4337                         .maxlen         = sizeof(int),
4338                         .mode           = 0644,
4339                         .proc_handler   = proc_dointvec,
4340                 },
4341                 {
4342                         .procname       = "accept_redirects",
4343                         .data           = &ipv6_devconf.accept_redirects,
4344                         .maxlen         = sizeof(int),
4345                         .mode           = 0644,
4346                         .proc_handler   = proc_dointvec,
4347                 },
4348                 {
4349                         .procname       = "autoconf",
4350                         .data           = &ipv6_devconf.autoconf,
4351                         .maxlen         = sizeof(int),
4352                         .mode           = 0644,
4353                         .proc_handler   = proc_dointvec,
4354                 },
4355                 {
4356                         .procname       = "dad_transmits",
4357                         .data           = &ipv6_devconf.dad_transmits,
4358                         .maxlen         = sizeof(int),
4359                         .mode           = 0644,
4360                         .proc_handler   = proc_dointvec,
4361                 },
4362                 {
4363                         .procname       = "router_solicitations",
4364                         .data           = &ipv6_devconf.rtr_solicits,
4365                         .maxlen         = sizeof(int),
4366                         .mode           = 0644,
4367                         .proc_handler   = proc_dointvec,
4368                 },
4369                 {
4370                         .procname       = "router_solicitation_interval",
4371                         .data           = &ipv6_devconf.rtr_solicit_interval,
4372                         .maxlen         = sizeof(int),
4373                         .mode           = 0644,
4374                         .proc_handler   = proc_dointvec_jiffies,
4375                 },
4376                 {
4377                         .procname       = "router_solicitation_delay",
4378                         .data           = &ipv6_devconf.rtr_solicit_delay,
4379                         .maxlen         = sizeof(int),
4380                         .mode           = 0644,
4381                         .proc_handler   = proc_dointvec_jiffies,
4382                 },
4383                 {
4384                         .procname       = "force_mld_version",
4385                         .data           = &ipv6_devconf.force_mld_version,
4386                         .maxlen         = sizeof(int),
4387                         .mode           = 0644,
4388                         .proc_handler   = proc_dointvec,
4389                 },
4390 #ifdef CONFIG_IPV6_PRIVACY
4391                 {
4392                         .procname       = "use_tempaddr",
4393                         .data           = &ipv6_devconf.use_tempaddr,
4394                         .maxlen         = sizeof(int),
4395                         .mode           = 0644,
4396                         .proc_handler   = proc_dointvec,
4397                 },
4398                 {
4399                         .procname       = "temp_valid_lft",
4400                         .data           = &ipv6_devconf.temp_valid_lft,
4401                         .maxlen         = sizeof(int),
4402                         .mode           = 0644,
4403                         .proc_handler   = proc_dointvec,
4404                 },
4405                 {
4406                         .procname       = "temp_prefered_lft",
4407                         .data           = &ipv6_devconf.temp_prefered_lft,
4408                         .maxlen         = sizeof(int),
4409                         .mode           = 0644,
4410                         .proc_handler   = proc_dointvec,
4411                 },
4412                 {
4413                         .procname       = "regen_max_retry",
4414                         .data           = &ipv6_devconf.regen_max_retry,
4415                         .maxlen         = sizeof(int),
4416                         .mode           = 0644,
4417                         .proc_handler   = proc_dointvec,
4418                 },
4419                 {
4420                         .procname       = "max_desync_factor",
4421                         .data           = &ipv6_devconf.max_desync_factor,
4422                         .maxlen         = sizeof(int),
4423                         .mode           = 0644,
4424                         .proc_handler   = proc_dointvec,
4425                 },
4426 #endif
4427                 {
4428                         .procname       = "max_addresses",
4429                         .data           = &ipv6_devconf.max_addresses,
4430                         .maxlen         = sizeof(int),
4431                         .mode           = 0644,
4432                         .proc_handler   = proc_dointvec,
4433                 },
4434                 {
4435                         .procname       = "accept_ra_defrtr",
4436                         .data           = &ipv6_devconf.accept_ra_defrtr,
4437                         .maxlen         = sizeof(int),
4438                         .mode           = 0644,
4439                         .proc_handler   = proc_dointvec,
4440                 },
4441                 {
4442                         .procname       = "accept_ra_pinfo",
4443                         .data           = &ipv6_devconf.accept_ra_pinfo,
4444                         .maxlen         = sizeof(int),
4445                         .mode           = 0644,
4446                         .proc_handler   = proc_dointvec,
4447                 },
4448 #ifdef CONFIG_IPV6_ROUTER_PREF
4449                 {
4450                         .procname       = "accept_ra_rtr_pref",
4451                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4452                         .maxlen         = sizeof(int),
4453                         .mode           = 0644,
4454                         .proc_handler   = proc_dointvec,
4455                 },
4456                 {
4457                         .procname       = "router_probe_interval",
4458                         .data           = &ipv6_devconf.rtr_probe_interval,
4459                         .maxlen         = sizeof(int),
4460                         .mode           = 0644,
4461                         .proc_handler   = proc_dointvec_jiffies,
4462                 },
4463 #ifdef CONFIG_IPV6_ROUTE_INFO
4464                 {
4465                         .procname       = "accept_ra_rt_info_max_plen",
4466                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4467                         .maxlen         = sizeof(int),
4468                         .mode           = 0644,
4469                         .proc_handler   = proc_dointvec,
4470                 },
4471 #endif
4472 #endif
4473                 {
4474                         .procname       = "proxy_ndp",
4475                         .data           = &ipv6_devconf.proxy_ndp,
4476                         .maxlen         = sizeof(int),
4477                         .mode           = 0644,
4478                         .proc_handler   = proc_dointvec,
4479                 },
4480                 {
4481                         .procname       = "accept_source_route",
4482                         .data           = &ipv6_devconf.accept_source_route,
4483                         .maxlen         = sizeof(int),
4484                         .mode           = 0644,
4485                         .proc_handler   = proc_dointvec,
4486                 },
4487 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4488                 {
4489                         .procname       = "optimistic_dad",
4490                         .data           = &ipv6_devconf.optimistic_dad,
4491                         .maxlen         = sizeof(int),
4492                         .mode           = 0644,
4493                         .proc_handler   = proc_dointvec,
4494
4495                 },
4496 #endif
4497 #ifdef CONFIG_IPV6_MROUTE
4498                 {
4499                         .procname       = "mc_forwarding",
4500                         .data           = &ipv6_devconf.mc_forwarding,
4501                         .maxlen         = sizeof(int),
4502                         .mode           = 0444,
4503                         .proc_handler   = proc_dointvec,
4504                 },
4505 #endif
4506                 {
4507                         .procname       = "disable_ipv6",
4508                         .data           = &ipv6_devconf.disable_ipv6,
4509                         .maxlen         = sizeof(int),
4510                         .mode           = 0644,
4511                         .proc_handler   = addrconf_sysctl_disable,
4512                 },
4513                 {
4514                         .procname       = "accept_dad",
4515                         .data           = &ipv6_devconf.accept_dad,
4516                         .maxlen         = sizeof(int),
4517                         .mode           = 0644,
4518                         .proc_handler   = proc_dointvec,
4519                 },
4520                 {
4521                         .procname       = "force_tllao",
4522                         .data           = &ipv6_devconf.force_tllao,
4523                         .maxlen         = sizeof(int),
4524                         .mode           = 0644,
4525                         .proc_handler   = proc_dointvec
4526                 },
4527                 {
4528                         /* sentinel */
4529                 }
4530         },
4531 };
4532
4533 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4534                 struct inet6_dev *idev, struct ipv6_devconf *p)
4535 {
4536         int i;
4537         struct addrconf_sysctl_table *t;
4538
4539 #define ADDRCONF_CTL_PATH_DEV   3
4540
4541         struct ctl_path addrconf_ctl_path[] = {
4542                 { .procname = "net", },
4543                 { .procname = "ipv6", },
4544                 { .procname = "conf", },
4545                 { /* to be set */ },
4546                 { },
4547         };
4548
4549
4550         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4551         if (t == NULL)
4552                 goto out;
4553
4554         for (i = 0; t->addrconf_vars[i].data; i++) {
4555                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4556                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4557                 t->addrconf_vars[i].extra2 = net;
4558         }
4559
4560         /*
4561          * Make a copy of dev_name, because '.procname' is regarded as const
4562          * by sysctl and we wouldn't want anyone to change it under our feet
4563          * (see SIOCSIFNAME).
4564          */
4565         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4566         if (!t->dev_name)
4567                 goto free;
4568
4569         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4570
4571         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4572                         t->addrconf_vars);
4573         if (t->sysctl_header == NULL)
4574                 goto free_procname;
4575
4576         p->sysctl = t;
4577         return 0;
4578
4579 free_procname:
4580         kfree(t->dev_name);
4581 free:
4582         kfree(t);
4583 out:
4584         return -ENOBUFS;
4585 }
4586
4587 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4588 {
4589         struct addrconf_sysctl_table *t;
4590
4591         if (p->sysctl == NULL)
4592                 return;
4593
4594         t = p->sysctl;
4595         p->sysctl = NULL;
4596         unregister_net_sysctl_table(t->sysctl_header);
4597         kfree(t->dev_name);
4598         kfree(t);
4599 }
4600
4601 static void addrconf_sysctl_register(struct inet6_dev *idev)
4602 {
4603         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4604                               &ndisc_ifinfo_sysctl_change);
4605         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4606                                         idev, &idev->cnf);
4607 }
4608
4609 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4610 {
4611         __addrconf_sysctl_unregister(&idev->cnf);
4612         neigh_sysctl_unregister(idev->nd_parms);
4613 }
4614
4615
4616 #endif
4617
4618 static int __net_init addrconf_init_net(struct net *net)
4619 {
4620         int err;
4621         struct ipv6_devconf *all, *dflt;
4622
4623         err = -ENOMEM;
4624         all = &ipv6_devconf;
4625         dflt = &ipv6_devconf_dflt;
4626
4627         if (!net_eq(net, &init_net)) {
4628                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4629                 if (all == NULL)
4630                         goto err_alloc_all;
4631
4632                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4633                 if (dflt == NULL)
4634                         goto err_alloc_dflt;
4635         } else {
4636                 /* these will be inherited by all namespaces */
4637                 dflt->autoconf = ipv6_defaults.autoconf;
4638                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4639         }
4640
4641         net->ipv6.devconf_all = all;
4642         net->ipv6.devconf_dflt = dflt;
4643
4644 #ifdef CONFIG_SYSCTL
4645         err = __addrconf_sysctl_register(net, "all", NULL, all);
4646         if (err < 0)
4647                 goto err_reg_all;
4648
4649         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4650         if (err < 0)
4651                 goto err_reg_dflt;
4652 #endif
4653         return 0;
4654
4655 #ifdef CONFIG_SYSCTL
4656 err_reg_dflt:
4657         __addrconf_sysctl_unregister(all);
4658 err_reg_all:
4659         kfree(dflt);
4660 #endif
4661 err_alloc_dflt:
4662         kfree(all);
4663 err_alloc_all:
4664         return err;
4665 }
4666
4667 static void __net_exit addrconf_exit_net(struct net *net)
4668 {
4669 #ifdef CONFIG_SYSCTL
4670         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4671         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4672 #endif
4673         if (!net_eq(net, &init_net)) {
4674                 kfree(net->ipv6.devconf_dflt);
4675                 kfree(net->ipv6.devconf_all);
4676         }
4677 }
4678
4679 static struct pernet_operations addrconf_ops = {
4680         .init = addrconf_init_net,
4681         .exit = addrconf_exit_net,
4682 };
4683
4684 /*
4685  *      Device notifier
4686  */
4687
4688 int register_inet6addr_notifier(struct notifier_block *nb)
4689 {
4690         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4691 }
4692 EXPORT_SYMBOL(register_inet6addr_notifier);
4693
4694 int unregister_inet6addr_notifier(struct notifier_block *nb)
4695 {
4696         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4697 }
4698 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4699
4700 static struct rtnl_af_ops inet6_ops = {
4701         .family           = AF_INET6,
4702         .fill_link_af     = inet6_fill_link_af,
4703         .get_link_af_size = inet6_get_link_af_size,
4704 };
4705
4706 /*
4707  *      Init / cleanup code
4708  */
4709
4710 int __init addrconf_init(void)
4711 {
4712         int i, err;
4713
4714         err = ipv6_addr_label_init();
4715         if (err < 0) {
4716                 printk(KERN_CRIT "IPv6 Addrconf:"
4717                        " cannot initialize default policy table: %d.\n", err);
4718                 goto out;
4719         }
4720
4721         err = register_pernet_subsys(&addrconf_ops);
4722         if (err < 0)
4723                 goto out_addrlabel;
4724
4725         /* The addrconf netdev notifier requires that loopback_dev
4726          * has it's ipv6 private information allocated and setup
4727          * before it can bring up and give link-local addresses
4728          * to other devices which are up.
4729          *
4730          * Unfortunately, loopback_dev is not necessarily the first
4731          * entry in the global dev_base list of net devices.  In fact,
4732          * it is likely to be the very last entry on that list.
4733          * So this causes the notifier registry below to try and
4734          * give link-local addresses to all devices besides loopback_dev
4735          * first, then loopback_dev, which cases all the non-loopback_dev
4736          * devices to fail to get a link-local address.
4737          *
4738          * So, as a temporary fix, allocate the ipv6 structure for
4739          * loopback_dev first by hand.
4740          * Longer term, all of the dependencies ipv6 has upon the loopback
4741          * device and it being up should be removed.
4742          */
4743         rtnl_lock();
4744         if (!ipv6_add_dev(init_net.loopback_dev))
4745                 err = -ENOMEM;
4746         rtnl_unlock();
4747         if (err)
4748                 goto errlo;
4749
4750         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4751                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4752
4753         register_netdevice_notifier(&ipv6_dev_notf);
4754
4755         addrconf_verify(0);
4756
4757         err = rtnl_af_register(&inet6_ops);
4758         if (err < 0)
4759                 goto errout_af;
4760
4761         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4762                               NULL);
4763         if (err < 0)
4764                 goto errout;
4765
4766         /* Only the first call to __rtnl_register can fail */
4767         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4768         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4769         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4770                         inet6_dump_ifaddr, NULL);
4771         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4772                         inet6_dump_ifmcaddr, NULL);
4773         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4774                         inet6_dump_ifacaddr, NULL);
4775
4776         ipv6_addr_label_rtnl_register();
4777
4778         return 0;
4779 errout:
4780         rtnl_af_unregister(&inet6_ops);
4781 errout_af:
4782         unregister_netdevice_notifier(&ipv6_dev_notf);
4783 errlo:
4784         unregister_pernet_subsys(&addrconf_ops);
4785 out_addrlabel:
4786         ipv6_addr_label_cleanup();
4787 out:
4788         return err;
4789 }
4790
4791 void addrconf_cleanup(void)
4792 {
4793         struct net_device *dev;
4794         int i;
4795
4796         unregister_netdevice_notifier(&ipv6_dev_notf);
4797         unregister_pernet_subsys(&addrconf_ops);
4798         ipv6_addr_label_cleanup();
4799
4800         rtnl_lock();
4801
4802         __rtnl_af_unregister(&inet6_ops);
4803
4804         /* clean dev list */
4805         for_each_netdev(&init_net, dev) {
4806                 if (__in6_dev_get(dev) == NULL)
4807                         continue;
4808                 addrconf_ifdown(dev, 1);
4809         }
4810         addrconf_ifdown(init_net.loopback_dev, 2);
4811
4812         /*
4813          *      Check hash table.
4814          */
4815         spin_lock_bh(&addrconf_hash_lock);
4816         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4817                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4818         spin_unlock_bh(&addrconf_hash_lock);
4819
4820         del_timer(&addr_chk_timer);
4821         rtnl_unlock();
4822 }