f8f76d6e21cbe3d212a6d6d830ce3a9cc46c1c03
[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/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/in6.h>
47 #include <linux/netdevice.h>
48 #include <linux/if_addr.h>
49 #include <linux/if_arp.h>
50 #include <linux/if_arcnet.h>
51 #include <linux/if_infiniband.h>
52 #include <linux/route.h>
53 #include <linux/inetdevice.h>
54 #include <linux/init.h>
55 #ifdef CONFIG_SYSCTL
56 #include <linux/sysctl.h>
57 #endif
58 #include <linux/capability.h>
59 #include <linux/delay.h>
60 #include <linux/notifier.h>
61 #include <linux/string.h>
62
63 #include <net/net_namespace.h>
64 #include <net/sock.h>
65 #include <net/snmp.h>
66
67 #include <net/ipv6.h>
68 #include <net/protocol.h>
69 #include <net/ndisc.h>
70 #include <net/ip6_route.h>
71 #include <net/addrconf.h>
72 #include <net/tcp.h>
73 #include <net/ip.h>
74 #include <net/netlink.h>
75 #include <net/pkt_sched.h>
76 #include <linux/if_tunnel.h>
77 #include <linux/rtnetlink.h>
78
79 #ifdef CONFIG_IPV6_PRIVACY
80 #include <linux/random.h>
81 #endif
82
83 #include <asm/uaccess.h>
84 #include <asm/unaligned.h>
85
86 #include <linux/proc_fs.h>
87 #include <linux/seq_file.h>
88
89 /* Set to 3 to get tracing... */
90 #define ACONF_DEBUG 2
91
92 #if ACONF_DEBUG >= 3
93 #define ADBG(x) printk x
94 #else
95 #define ADBG(x)
96 #endif
97
98 #define INFINITY_LIFE_TIME      0xFFFFFFFF
99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
100
101 #ifdef CONFIG_SYSCTL
102 static void addrconf_sysctl_register(struct inet6_dev *idev);
103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
104 #else
105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
106 {
107 }
108
109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
110 {
111 }
112 #endif
113
114 #ifdef CONFIG_IPV6_PRIVACY
115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
117 static void ipv6_regen_rndid(unsigned long data);
118
119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
120 #endif
121
122 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
123 static int ipv6_count_addresses(struct inet6_dev *idev);
124
125 /*
126  *      Configured unicast address hash table
127  */
128 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
129 static DEFINE_RWLOCK(addrconf_hash_lock);
130
131 static void addrconf_verify(unsigned long);
132
133 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
134 static DEFINE_SPINLOCK(addrconf_verify_lock);
135
136 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
137 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
138
139 static int addrconf_ifdown(struct net_device *dev, int how);
140
141 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
142 static void addrconf_dad_timer(unsigned long data);
143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
144 static void addrconf_dad_run(struct inet6_dev *idev);
145 static void addrconf_rs_timer(unsigned long data);
146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
147 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148
149 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
150                                 struct prefix_info *pinfo);
151 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
152                               struct net_device *dev);
153
154 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
155
156 static struct ipv6_devconf ipv6_devconf __read_mostly = {
157         .forwarding             = 0,
158         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
159         .mtu6                   = IPV6_MIN_MTU,
160         .accept_ra              = 1,
161         .accept_redirects       = 1,
162         .autoconf               = 1,
163         .force_mld_version      = 0,
164         .dad_transmits          = 1,
165         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
166         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
167         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
168 #ifdef CONFIG_IPV6_PRIVACY
169         .use_tempaddr           = 0,
170         .temp_valid_lft         = TEMP_VALID_LIFETIME,
171         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
172         .regen_max_retry        = REGEN_MAX_RETRY,
173         .max_desync_factor      = MAX_DESYNC_FACTOR,
174 #endif
175         .max_addresses          = IPV6_MAX_ADDRESSES,
176         .accept_ra_defrtr       = 1,
177         .accept_ra_pinfo        = 1,
178 #ifdef CONFIG_IPV6_ROUTER_PREF
179         .accept_ra_rtr_pref     = 1,
180         .rtr_probe_interval     = 60 * HZ,
181 #ifdef CONFIG_IPV6_ROUTE_INFO
182         .accept_ra_rt_info_max_plen = 0,
183 #endif
184 #endif
185         .proxy_ndp              = 0,
186         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
187         .disable_ipv6           = 0,
188         .accept_dad             = 1,
189 };
190
191 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
192         .forwarding             = 0,
193         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
194         .mtu6                   = IPV6_MIN_MTU,
195         .accept_ra              = 1,
196         .accept_redirects       = 1,
197         .autoconf               = 1,
198         .dad_transmits          = 1,
199         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
200         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
202 #ifdef CONFIG_IPV6_PRIVACY
203         .use_tempaddr           = 0,
204         .temp_valid_lft         = TEMP_VALID_LIFETIME,
205         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
206         .regen_max_retry        = REGEN_MAX_RETRY,
207         .max_desync_factor      = MAX_DESYNC_FACTOR,
208 #endif
209         .max_addresses          = IPV6_MAX_ADDRESSES,
210         .accept_ra_defrtr       = 1,
211         .accept_ra_pinfo        = 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213         .accept_ra_rtr_pref     = 1,
214         .rtr_probe_interval     = 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216         .accept_ra_rt_info_max_plen = 0,
217 #endif
218 #endif
219         .proxy_ndp              = 0,
220         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
221         .disable_ipv6           = 0,
222         .accept_dad             = 1,
223 };
224
225 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
226 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
227 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
228 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
229 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
230
231 /* Check if a valid qdisc is available */
232 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
233 {
234         return !qdisc_tx_is_noop(dev);
235 }
236
237 /* Check if a route is valid prefix route */
238 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
239 {
240         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
241 }
242
243 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
244 {
245         if (del_timer(&ifp->timer))
246                 __in6_ifa_put(ifp);
247 }
248
249 enum addrconf_timer_t
250 {
251         AC_NONE,
252         AC_DAD,
253         AC_RS,
254 };
255
256 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
257                                enum addrconf_timer_t what,
258                                unsigned long when)
259 {
260         if (!del_timer(&ifp->timer))
261                 in6_ifa_hold(ifp);
262
263         switch (what) {
264         case AC_DAD:
265                 ifp->timer.function = addrconf_dad_timer;
266                 break;
267         case AC_RS:
268                 ifp->timer.function = addrconf_rs_timer;
269                 break;
270         default:;
271         }
272         ifp->timer.expires = jiffies + when;
273         add_timer(&ifp->timer);
274 }
275
276 static int snmp6_alloc_dev(struct inet6_dev *idev)
277 {
278         if (snmp_mib_init((void **)idev->stats.ipv6,
279                           sizeof(struct ipstats_mib)) < 0)
280                 goto err_ip;
281         if (snmp_mib_init((void **)idev->stats.icmpv6,
282                           sizeof(struct icmpv6_mib)) < 0)
283                 goto err_icmp;
284         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
285                           sizeof(struct icmpv6msg_mib)) < 0)
286                 goto err_icmpmsg;
287
288         return 0;
289
290 err_icmpmsg:
291         snmp_mib_free((void **)idev->stats.icmpv6);
292 err_icmp:
293         snmp_mib_free((void **)idev->stats.ipv6);
294 err_ip:
295         return -ENOMEM;
296 }
297
298 static void snmp6_free_dev(struct inet6_dev *idev)
299 {
300         snmp_mib_free((void **)idev->stats.icmpv6msg);
301         snmp_mib_free((void **)idev->stats.icmpv6);
302         snmp_mib_free((void **)idev->stats.ipv6);
303 }
304
305 /* Nobody refers to this device, we may destroy it. */
306
307 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
308 {
309         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
310         kfree(idev);
311 }
312
313 void in6_dev_finish_destroy(struct inet6_dev *idev)
314 {
315         struct net_device *dev = idev->dev;
316
317         WARN_ON(idev->addr_list != NULL);
318         WARN_ON(idev->mc_list != NULL);
319
320 #ifdef NET_REFCNT_DEBUG
321         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
322 #endif
323         dev_put(dev);
324         if (!idev->dead) {
325                 printk("Freeing alive inet6 device %p\n", idev);
326                 return;
327         }
328         snmp6_free_dev(idev);
329         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
330 }
331
332 EXPORT_SYMBOL(in6_dev_finish_destroy);
333
334 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
335 {
336         struct inet6_dev *ndev;
337
338         ASSERT_RTNL();
339
340         if (dev->mtu < IPV6_MIN_MTU)
341                 return NULL;
342
343         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
344
345         if (ndev == NULL)
346                 return NULL;
347
348         rwlock_init(&ndev->lock);
349         ndev->dev = dev;
350         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
351         ndev->cnf.mtu6 = dev->mtu;
352         ndev->cnf.sysctl = NULL;
353         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
354         if (ndev->nd_parms == NULL) {
355                 kfree(ndev);
356                 return NULL;
357         }
358         if (ndev->cnf.forwarding)
359                 dev_disable_lro(dev);
360         /* We refer to the device */
361         dev_hold(dev);
362
363         if (snmp6_alloc_dev(ndev) < 0) {
364                 ADBG((KERN_WARNING
365                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
366                         __func__, dev->name));
367                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
368                 ndev->dead = 1;
369                 in6_dev_finish_destroy(ndev);
370                 return NULL;
371         }
372
373         if (snmp6_register_dev(ndev) < 0) {
374                 ADBG((KERN_WARNING
375                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
376                         __func__, dev->name));
377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378                 ndev->dead = 1;
379                 in6_dev_finish_destroy(ndev);
380                 return NULL;
381         }
382
383         /* One reference from device.  We must do this before
384          * we invoke __ipv6_regen_rndid().
385          */
386         in6_dev_hold(ndev);
387
388         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
389                 ndev->cnf.accept_dad = -1;
390
391 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
392         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
393                 printk(KERN_INFO
394                        "%s: Disabled Multicast RS\n",
395                        dev->name);
396                 ndev->cnf.rtr_solicits = 0;
397         }
398 #endif
399
400 #ifdef CONFIG_IPV6_PRIVACY
401         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
402         if ((dev->flags&IFF_LOOPBACK) ||
403             dev->type == ARPHRD_TUNNEL ||
404             dev->type == ARPHRD_TUNNEL6 ||
405             dev->type == ARPHRD_SIT ||
406             dev->type == ARPHRD_NONE) {
407                 printk(KERN_INFO
408                        "%s: Disabled Privacy Extensions\n",
409                        dev->name);
410                 ndev->cnf.use_tempaddr = -1;
411         } else {
412                 in6_dev_hold(ndev);
413                 ipv6_regen_rndid((unsigned long) ndev);
414         }
415 #endif
416
417         if (netif_running(dev) && addrconf_qdisc_ok(dev))
418                 ndev->if_flags |= IF_READY;
419
420         ipv6_mc_init_dev(ndev);
421         ndev->tstamp = jiffies;
422         addrconf_sysctl_register(ndev);
423         /* protected by rtnl_lock */
424         rcu_assign_pointer(dev->ip6_ptr, ndev);
425
426         /* Join all-node multicast group */
427         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
428
429         return ndev;
430 }
431
432 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
433 {
434         struct inet6_dev *idev;
435
436         ASSERT_RTNL();
437
438         if ((idev = __in6_dev_get(dev)) == NULL) {
439                 if ((idev = ipv6_add_dev(dev)) == NULL)
440                         return NULL;
441         }
442
443         if (dev->flags&IFF_UP)
444                 ipv6_mc_up(idev);
445         return idev;
446 }
447
448 #ifdef CONFIG_SYSCTL
449 static void dev_forward_change(struct inet6_dev *idev)
450 {
451         struct net_device *dev;
452         struct inet6_ifaddr *ifa;
453
454         if (!idev)
455                 return;
456         dev = idev->dev;
457         if (idev->cnf.forwarding)
458                 dev_disable_lro(dev);
459         if (dev && (dev->flags & IFF_MULTICAST)) {
460                 if (idev->cnf.forwarding)
461                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
462                 else
463                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
464         }
465         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
466                 if (ifa->flags&IFA_F_TENTATIVE)
467                         continue;
468                 if (idev->cnf.forwarding)
469                         addrconf_join_anycast(ifa);
470                 else
471                         addrconf_leave_anycast(ifa);
472         }
473 }
474
475
476 static void addrconf_forward_change(struct net *net, __s32 newf)
477 {
478         struct net_device *dev;
479         struct inet6_dev *idev;
480
481         read_lock(&dev_base_lock);
482         for_each_netdev(net, dev) {
483                 rcu_read_lock();
484                 idev = __in6_dev_get(dev);
485                 if (idev) {
486                         int changed = (!idev->cnf.forwarding) ^ (!newf);
487                         idev->cnf.forwarding = newf;
488                         if (changed)
489                                 dev_forward_change(idev);
490                 }
491                 rcu_read_unlock();
492         }
493         read_unlock(&dev_base_lock);
494 }
495
496 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
497 {
498         struct net *net;
499
500         net = (struct net *)table->extra2;
501         if (p == &net->ipv6.devconf_dflt->forwarding)
502                 return;
503
504         rtnl_lock();
505         if (p == &net->ipv6.devconf_all->forwarding) {
506                 __s32 newf = net->ipv6.devconf_all->forwarding;
507                 net->ipv6.devconf_dflt->forwarding = newf;
508                 addrconf_forward_change(net, newf);
509         } else if ((!*p) ^ (!old))
510                 dev_forward_change((struct inet6_dev *)table->extra1);
511         rtnl_unlock();
512
513         if (*p)
514                 rt6_purge_dflt_routers(net);
515 }
516 #endif
517
518 /* Nobody refers to this ifaddr, destroy it */
519
520 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
521 {
522         WARN_ON(ifp->if_next != NULL);
523         WARN_ON(ifp->lst_next != NULL);
524
525 #ifdef NET_REFCNT_DEBUG
526         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
527 #endif
528
529         in6_dev_put(ifp->idev);
530
531         if (del_timer(&ifp->timer))
532                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
533
534         if (!ifp->dead) {
535                 printk("Freeing alive inet6 address %p\n", ifp);
536                 return;
537         }
538         dst_release(&ifp->rt->u.dst);
539
540         kfree(ifp);
541 }
542
543 static void
544 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
545 {
546         struct inet6_ifaddr *ifa, **ifap;
547         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
548
549         /*
550          * Each device address list is sorted in order of scope -
551          * global before linklocal.
552          */
553         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
554              ifap = &ifa->if_next) {
555                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
556                         break;
557         }
558
559         ifp->if_next = *ifap;
560         *ifap = ifp;
561 }
562
563 /*
564  *      Hash function taken from net_alias.c
565  */
566 static u8 ipv6_addr_hash(const struct in6_addr *addr)
567 {
568         __u32 word;
569
570         /*
571          * We perform the hash function over the last 64 bits of the address
572          * This will include the IEEE address token on links that support it.
573          */
574
575         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
576         word ^= (word >> 16);
577         word ^= (word >> 8);
578
579         return ((word ^ (word >> 4)) & 0x0f);
580 }
581
582 /* On success it returns ifp with increased reference count */
583
584 static struct inet6_ifaddr *
585 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
586               int scope, u32 flags)
587 {
588         struct inet6_ifaddr *ifa = NULL;
589         struct rt6_info *rt;
590         int hash;
591         int err = 0;
592         int addr_type = ipv6_addr_type(addr);
593
594         if (addr_type == IPV6_ADDR_ANY ||
595             addr_type & IPV6_ADDR_MULTICAST ||
596             (!(idev->dev->flags & IFF_LOOPBACK) &&
597              addr_type & IPV6_ADDR_LOOPBACK))
598                 return ERR_PTR(-EADDRNOTAVAIL);
599
600         rcu_read_lock_bh();
601         if (idev->dead) {
602                 err = -ENODEV;                  /*XXX*/
603                 goto out2;
604         }
605
606         write_lock(&addrconf_hash_lock);
607
608         /* Ignore adding duplicate addresses on an interface */
609         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
610                 ADBG(("ipv6_add_addr: already assigned\n"));
611                 err = -EEXIST;
612                 goto out;
613         }
614
615         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
616
617         if (ifa == NULL) {
618                 ADBG(("ipv6_add_addr: malloc failed\n"));
619                 err = -ENOBUFS;
620                 goto out;
621         }
622
623         rt = addrconf_dst_alloc(idev, addr, 0);
624         if (IS_ERR(rt)) {
625                 err = PTR_ERR(rt);
626                 goto out;
627         }
628
629         ipv6_addr_copy(&ifa->addr, addr);
630
631         spin_lock_init(&ifa->lock);
632         init_timer(&ifa->timer);
633         ifa->timer.data = (unsigned long) ifa;
634         ifa->scope = scope;
635         ifa->prefix_len = pfxlen;
636         ifa->flags = flags | IFA_F_TENTATIVE;
637         ifa->cstamp = ifa->tstamp = jiffies;
638
639         ifa->rt = rt;
640
641         /*
642          * part one of RFC 4429, section 3.3
643          * We should not configure an address as
644          * optimistic if we do not yet know the link
645          * layer address of our nexhop router
646          */
647
648         if (rt->rt6i_nexthop == NULL)
649                 ifa->flags &= ~IFA_F_OPTIMISTIC;
650
651         ifa->idev = idev;
652         in6_dev_hold(idev);
653         /* For caller */
654         in6_ifa_hold(ifa);
655
656         /* Add to big hash table */
657         hash = ipv6_addr_hash(addr);
658
659         ifa->lst_next = inet6_addr_lst[hash];
660         inet6_addr_lst[hash] = ifa;
661         in6_ifa_hold(ifa);
662         write_unlock(&addrconf_hash_lock);
663
664         write_lock(&idev->lock);
665         /* Add to inet6_dev unicast addr list. */
666         ipv6_link_dev_addr(idev, ifa);
667
668 #ifdef CONFIG_IPV6_PRIVACY
669         if (ifa->flags&IFA_F_TEMPORARY) {
670                 ifa->tmp_next = idev->tempaddr_list;
671                 idev->tempaddr_list = ifa;
672                 in6_ifa_hold(ifa);
673         }
674 #endif
675
676         in6_ifa_hold(ifa);
677         write_unlock(&idev->lock);
678 out2:
679         rcu_read_unlock_bh();
680
681         if (likely(err == 0))
682                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
683         else {
684                 kfree(ifa);
685                 ifa = ERR_PTR(err);
686         }
687
688         return ifa;
689 out:
690         write_unlock(&addrconf_hash_lock);
691         goto out2;
692 }
693
694 /* This function wants to get referenced ifp and releases it before return */
695
696 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
697 {
698         struct inet6_ifaddr *ifa, **ifap;
699         struct inet6_dev *idev = ifp->idev;
700         int hash;
701         int deleted = 0, onlink = 0;
702         unsigned long expires = jiffies;
703
704         hash = ipv6_addr_hash(&ifp->addr);
705
706         ifp->dead = 1;
707
708         write_lock_bh(&addrconf_hash_lock);
709         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
710              ifap = &ifa->lst_next) {
711                 if (ifa == ifp) {
712                         *ifap = ifa->lst_next;
713                         __in6_ifa_put(ifp);
714                         ifa->lst_next = NULL;
715                         break;
716                 }
717         }
718         write_unlock_bh(&addrconf_hash_lock);
719
720         write_lock_bh(&idev->lock);
721 #ifdef CONFIG_IPV6_PRIVACY
722         if (ifp->flags&IFA_F_TEMPORARY) {
723                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
724                      ifap = &ifa->tmp_next) {
725                         if (ifa == ifp) {
726                                 *ifap = ifa->tmp_next;
727                                 if (ifp->ifpub) {
728                                         in6_ifa_put(ifp->ifpub);
729                                         ifp->ifpub = NULL;
730                                 }
731                                 __in6_ifa_put(ifp);
732                                 ifa->tmp_next = NULL;
733                                 break;
734                         }
735                 }
736         }
737 #endif
738
739         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
740                 if (ifa == ifp) {
741                         *ifap = ifa->if_next;
742                         __in6_ifa_put(ifp);
743                         ifa->if_next = NULL;
744                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
745                                 break;
746                         deleted = 1;
747                         continue;
748                 } else if (ifp->flags & IFA_F_PERMANENT) {
749                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
750                                               ifp->prefix_len)) {
751                                 if (ifa->flags & IFA_F_PERMANENT) {
752                                         onlink = 1;
753                                         if (deleted)
754                                                 break;
755                                 } else {
756                                         unsigned long lifetime;
757
758                                         if (!onlink)
759                                                 onlink = -1;
760
761                                         spin_lock(&ifa->lock);
762
763                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
764                                         /*
765                                          * Note: Because this address is
766                                          * not permanent, lifetime <
767                                          * LONG_MAX / HZ here.
768                                          */
769                                         if (time_before(expires,
770                                                         ifa->tstamp + lifetime * HZ))
771                                                 expires = ifa->tstamp + lifetime * HZ;
772                                         spin_unlock(&ifa->lock);
773                                 }
774                         }
775                 }
776                 ifap = &ifa->if_next;
777         }
778         write_unlock_bh(&idev->lock);
779
780         addrconf_del_timer(ifp);
781
782         ipv6_ifa_notify(RTM_DELADDR, ifp);
783
784         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
785
786         /*
787          * Purge or update corresponding prefix
788          *
789          * 1) we don't purge prefix here if address was not permanent.
790          *    prefix is managed by its own lifetime.
791          * 2) if there're no addresses, delete prefix.
792          * 3) if there're still other permanent address(es),
793          *    corresponding prefix is still permanent.
794          * 4) otherwise, update prefix lifetime to the
795          *    longest valid lifetime among the corresponding
796          *    addresses on the device.
797          *    Note: subsequent RA will update lifetime.
798          *
799          * --yoshfuji
800          */
801         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
802                 struct in6_addr prefix;
803                 struct rt6_info *rt;
804                 struct net *net = dev_net(ifp->idev->dev);
805                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
806                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
807
808                 if (rt && addrconf_is_prefix_route(rt)) {
809                         if (onlink == 0) {
810                                 ip6_del_rt(rt);
811                                 rt = NULL;
812                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
813                                 rt->rt6i_expires = expires;
814                                 rt->rt6i_flags |= RTF_EXPIRES;
815                         }
816                 }
817                 dst_release(&rt->u.dst);
818         }
819
820         in6_ifa_put(ifp);
821 }
822
823 #ifdef CONFIG_IPV6_PRIVACY
824 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
825 {
826         struct inet6_dev *idev = ifp->idev;
827         struct in6_addr addr, *tmpaddr;
828         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
829         unsigned long regen_advance;
830         int tmp_plen;
831         int ret = 0;
832         int max_addresses;
833         u32 addr_flags;
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         tmp_valid_lft = min_t(__u32,
879                               ifp->valid_lft,
880                               idev->cnf.temp_valid_lft);
881         tmp_prefered_lft = min_t(__u32,
882                                  ifp->prefered_lft,
883                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
884         tmp_plen = ifp->prefix_len;
885         max_addresses = idev->cnf.max_addresses;
886         tmp_cstamp = ifp->cstamp;
887         tmp_tstamp = ifp->tstamp;
888         spin_unlock_bh(&ifp->lock);
889
890         regen_advance = idev->cnf.regen_max_retry *
891                         idev->cnf.dad_transmits *
892                         idev->nd_parms->retrans_time / HZ;
893         write_unlock(&idev->lock);
894
895         /* A temporary address is created only if this calculated Preferred
896          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
897          * an implementation must not create a temporary address with a zero
898          * Preferred Lifetime.
899          */
900         if (tmp_prefered_lft <= regen_advance) {
901                 in6_ifa_put(ifp);
902                 in6_dev_put(idev);
903                 ret = -1;
904                 goto out;
905         }
906
907         addr_flags = IFA_F_TEMPORARY;
908         /* set in addrconf_prefix_rcv() */
909         if (ifp->flags & IFA_F_OPTIMISTIC)
910                 addr_flags |= IFA_F_OPTIMISTIC;
911
912         ift = !max_addresses ||
913               ipv6_count_addresses(idev) < max_addresses ?
914                 ipv6_add_addr(idev, &addr, tmp_plen,
915                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
916                               addr_flags) : NULL;
917         if (!ift || IS_ERR(ift)) {
918                 in6_ifa_put(ifp);
919                 in6_dev_put(idev);
920                 printk(KERN_INFO
921                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
922                 tmpaddr = &addr;
923                 write_lock(&idev->lock);
924                 goto retry;
925         }
926
927         spin_lock_bh(&ift->lock);
928         ift->ifpub = ifp;
929         ift->valid_lft = tmp_valid_lft;
930         ift->prefered_lft = tmp_prefered_lft;
931         ift->cstamp = tmp_cstamp;
932         ift->tstamp = tmp_tstamp;
933         spin_unlock_bh(&ift->lock);
934
935         addrconf_dad_start(ift, 0);
936         in6_ifa_put(ift);
937         in6_dev_put(idev);
938 out:
939         return ret;
940 }
941 #endif
942
943 /*
944  *      Choose an appropriate source address (RFC3484)
945  */
946 enum {
947         IPV6_SADDR_RULE_INIT = 0,
948         IPV6_SADDR_RULE_LOCAL,
949         IPV6_SADDR_RULE_SCOPE,
950         IPV6_SADDR_RULE_PREFERRED,
951 #ifdef CONFIG_IPV6_MIP6
952         IPV6_SADDR_RULE_HOA,
953 #endif
954         IPV6_SADDR_RULE_OIF,
955         IPV6_SADDR_RULE_LABEL,
956 #ifdef CONFIG_IPV6_PRIVACY
957         IPV6_SADDR_RULE_PRIVACY,
958 #endif
959         IPV6_SADDR_RULE_ORCHID,
960         IPV6_SADDR_RULE_PREFIX,
961         IPV6_SADDR_RULE_MAX
962 };
963
964 struct ipv6_saddr_score {
965         int                     rule;
966         int                     addr_type;
967         struct inet6_ifaddr     *ifa;
968         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
969         int                     scopedist;
970         int                     matchlen;
971 };
972
973 struct ipv6_saddr_dst {
974         const struct in6_addr *addr;
975         int ifindex;
976         int scope;
977         int label;
978         unsigned int prefs;
979 };
980
981 static inline int ipv6_saddr_preferred(int type)
982 {
983         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
984                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
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 temprary 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         read_lock(&dev_base_lock);
1130         rcu_read_lock();
1131
1132         for_each_netdev(net, dev) {
1133                 struct inet6_dev *idev;
1134
1135                 /* Candidate Source Address (section 4)
1136                  *  - multicast and link-local destination address,
1137                  *    the set of candidate source address MUST only
1138                  *    include addresses assigned to interfaces
1139                  *    belonging to the same link as the outgoing
1140                  *    interface.
1141                  * (- For site-local destination addresses, the
1142                  *    set of candidate source addresses MUST only
1143                  *    include addresses assigned to interfaces
1144                  *    belonging to the same site as the outgoing
1145                  *    interface.)
1146                  */
1147                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1148                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1149                     dst.ifindex && dev->ifindex != dst.ifindex)
1150                         continue;
1151
1152                 idev = __in6_dev_get(dev);
1153                 if (!idev)
1154                         continue;
1155
1156                 read_lock_bh(&idev->lock);
1157                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1158                         int i;
1159
1160                         /*
1161                          * - Tentative Address (RFC2462 section 5.4)
1162                          *  - A tentative address is not considered
1163                          *    "assigned to an interface" in the traditional
1164                          *    sense, unless it is also flagged as optimistic.
1165                          * - Candidate Source Address (section 4)
1166                          *  - In any case, anycast addresses, multicast
1167                          *    addresses, and the unspecified address MUST
1168                          *    NOT be included in a candidate set.
1169                          */
1170                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1171                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1172                                 continue;
1173
1174                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1175
1176                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1177                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1178                                 LIMIT_NETDEBUG(KERN_DEBUG
1179                                                "ADDRCONF: unspecified / multicast address "
1180                                                "assigned as unicast address on %s",
1181                                                dev->name);
1182                                 continue;
1183                         }
1184
1185                         score->rule = -1;
1186                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1187
1188                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1189                                 int minihiscore, miniscore;
1190
1191                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1192                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1193
1194                                 if (minihiscore > miniscore) {
1195                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1196                                             score->scopedist > 0) {
1197                                                 /*
1198                                                  * special case:
1199                                                  * each remaining entry
1200                                                  * has too small (not enough)
1201                                                  * scope, because ifa entries
1202                                                  * are sorted by their scope
1203                                                  * values.
1204                                                  */
1205                                                 goto try_nextdev;
1206                                         }
1207                                         break;
1208                                 } else if (minihiscore < miniscore) {
1209                                         struct ipv6_saddr_score *tmp;
1210
1211                                         if (hiscore->ifa)
1212                                                 in6_ifa_put(hiscore->ifa);
1213
1214                                         in6_ifa_hold(score->ifa);
1215
1216                                         tmp = hiscore;
1217                                         hiscore = score;
1218                                         score = tmp;
1219
1220                                         /* restore our iterator */
1221                                         score->ifa = hiscore->ifa;
1222
1223                                         break;
1224                                 }
1225                         }
1226                 }
1227 try_nextdev:
1228                 read_unlock_bh(&idev->lock);
1229         }
1230         rcu_read_unlock();
1231         read_unlock(&dev_base_lock);
1232
1233         if (!hiscore->ifa)
1234                 return -EADDRNOTAVAIL;
1235
1236         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1237         in6_ifa_put(hiscore->ifa);
1238         return 0;
1239 }
1240
1241 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1242
1243 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1244                     unsigned char banned_flags)
1245 {
1246         struct inet6_dev *idev;
1247         int err = -EADDRNOTAVAIL;
1248
1249         rcu_read_lock();
1250         if ((idev = __in6_dev_get(dev)) != NULL) {
1251                 struct inet6_ifaddr *ifp;
1252
1253                 read_lock_bh(&idev->lock);
1254                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1255                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1256                                 ipv6_addr_copy(addr, &ifp->addr);
1257                                 err = 0;
1258                                 break;
1259                         }
1260                 }
1261                 read_unlock_bh(&idev->lock);
1262         }
1263         rcu_read_unlock();
1264         return err;
1265 }
1266
1267 static int ipv6_count_addresses(struct inet6_dev *idev)
1268 {
1269         int cnt = 0;
1270         struct inet6_ifaddr *ifp;
1271
1272         read_lock_bh(&idev->lock);
1273         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1274                 cnt++;
1275         read_unlock_bh(&idev->lock);
1276         return cnt;
1277 }
1278
1279 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1280                   struct net_device *dev, int strict)
1281 {
1282         struct inet6_ifaddr * ifp;
1283         u8 hash = ipv6_addr_hash(addr);
1284
1285         read_lock_bh(&addrconf_hash_lock);
1286         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1287                 if (!net_eq(dev_net(ifp->idev->dev), net))
1288                         continue;
1289                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1290                     !(ifp->flags&IFA_F_TENTATIVE)) {
1291                         if (dev == NULL || ifp->idev->dev == dev ||
1292                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1293                                 break;
1294                 }
1295         }
1296         read_unlock_bh(&addrconf_hash_lock);
1297         return ifp != NULL;
1298 }
1299 EXPORT_SYMBOL(ipv6_chk_addr);
1300
1301 static
1302 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1303                        struct net_device *dev)
1304 {
1305         struct inet6_ifaddr * ifp;
1306         u8 hash = ipv6_addr_hash(addr);
1307
1308         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1309                 if (!net_eq(dev_net(ifp->idev->dev), net))
1310                         continue;
1311                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1312                         if (dev == NULL || ifp->idev->dev == dev)
1313                                 break;
1314                 }
1315         }
1316         return ifp != NULL;
1317 }
1318
1319 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1320 {
1321         struct inet6_dev *idev;
1322         struct inet6_ifaddr *ifa;
1323         int     onlink;
1324
1325         onlink = 0;
1326         rcu_read_lock();
1327         idev = __in6_dev_get(dev);
1328         if (idev) {
1329                 read_lock_bh(&idev->lock);
1330                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1331                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1332                                                    ifa->prefix_len);
1333                         if (onlink)
1334                                 break;
1335                 }
1336                 read_unlock_bh(&idev->lock);
1337         }
1338         rcu_read_unlock();
1339         return onlink;
1340 }
1341
1342 EXPORT_SYMBOL(ipv6_chk_prefix);
1343
1344 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1345                                      struct net_device *dev, int strict)
1346 {
1347         struct inet6_ifaddr * ifp;
1348         u8 hash = ipv6_addr_hash(addr);
1349
1350         read_lock_bh(&addrconf_hash_lock);
1351         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1352                 if (!net_eq(dev_net(ifp->idev->dev), net))
1353                         continue;
1354                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1355                         if (dev == NULL || ifp->idev->dev == dev ||
1356                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1357                                 in6_ifa_hold(ifp);
1358                                 break;
1359                         }
1360                 }
1361         }
1362         read_unlock_bh(&addrconf_hash_lock);
1363
1364         return ifp;
1365 }
1366
1367 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1368 {
1369         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1370         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1371         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1372         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1373         int sk_ipv6only = ipv6_only_sock(sk);
1374         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1375         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1376         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1377
1378         if (!sk2_rcv_saddr && !sk_ipv6only)
1379                 return 1;
1380
1381         if (addr_type2 == IPV6_ADDR_ANY &&
1382             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1383                 return 1;
1384
1385         if (addr_type == IPV6_ADDR_ANY &&
1386             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1387                 return 1;
1388
1389         if (sk2_rcv_saddr6 &&
1390             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1391                 return 1;
1392
1393         if (addr_type == IPV6_ADDR_MAPPED &&
1394             !sk2_ipv6only &&
1395             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1396                 return 1;
1397
1398         return 0;
1399 }
1400
1401 /* Gets referenced address, destroys ifaddr */
1402
1403 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1404 {
1405         if (ifp->flags&IFA_F_PERMANENT) {
1406                 spin_lock_bh(&ifp->lock);
1407                 addrconf_del_timer(ifp);
1408                 ifp->flags |= IFA_F_TENTATIVE;
1409                 spin_unlock_bh(&ifp->lock);
1410                 in6_ifa_put(ifp);
1411 #ifdef CONFIG_IPV6_PRIVACY
1412         } else if (ifp->flags&IFA_F_TEMPORARY) {
1413                 struct inet6_ifaddr *ifpub;
1414                 spin_lock_bh(&ifp->lock);
1415                 ifpub = ifp->ifpub;
1416                 if (ifpub) {
1417                         in6_ifa_hold(ifpub);
1418                         spin_unlock_bh(&ifp->lock);
1419                         ipv6_create_tempaddr(ifpub, ifp);
1420                         in6_ifa_put(ifpub);
1421                 } else {
1422                         spin_unlock_bh(&ifp->lock);
1423                 }
1424                 ipv6_del_addr(ifp);
1425 #endif
1426         } else
1427                 ipv6_del_addr(ifp);
1428 }
1429
1430 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1431 {
1432         struct inet6_dev *idev = ifp->idev;
1433         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1434                 struct in6_addr addr;
1435
1436                 addr.s6_addr32[0] = htonl(0xfe800000);
1437                 addr.s6_addr32[1] = 0;
1438
1439                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1440                     ipv6_addr_equal(&ifp->addr, &addr)) {
1441                         /* DAD failed for link-local based on MAC address */
1442                         idev->cnf.disable_ipv6 = 1;
1443                 }
1444         }
1445
1446         if (net_ratelimit())
1447                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1448         addrconf_dad_stop(ifp);
1449 }
1450
1451 /* Join to solicited addr multicast group. */
1452
1453 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1454 {
1455         struct in6_addr maddr;
1456
1457         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1458                 return;
1459
1460         addrconf_addr_solict_mult(addr, &maddr);
1461         ipv6_dev_mc_inc(dev, &maddr);
1462 }
1463
1464 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1465 {
1466         struct in6_addr maddr;
1467
1468         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1469                 return;
1470
1471         addrconf_addr_solict_mult(addr, &maddr);
1472         __ipv6_dev_mc_dec(idev, &maddr);
1473 }
1474
1475 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1476 {
1477         struct in6_addr addr;
1478         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1479         if (ipv6_addr_any(&addr))
1480                 return;
1481         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1482 }
1483
1484 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1485 {
1486         struct in6_addr addr;
1487         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1488         if (ipv6_addr_any(&addr))
1489                 return;
1490         __ipv6_dev_ac_dec(ifp->idev, &addr);
1491 }
1492
1493 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1494 {
1495         if (dev->addr_len != ETH_ALEN)
1496                 return -1;
1497         memcpy(eui, dev->dev_addr, 3);
1498         memcpy(eui + 5, dev->dev_addr + 3, 3);
1499
1500         /*
1501          * The zSeries OSA network cards can be shared among various
1502          * OS instances, but the OSA cards have only one MAC address.
1503          * This leads to duplicate address conflicts in conjunction
1504          * with IPv6 if more than one instance uses the same card.
1505          *
1506          * The driver for these cards can deliver a unique 16-bit
1507          * identifier for each instance sharing the same card.  It is
1508          * placed instead of 0xFFFE in the interface identifier.  The
1509          * "u" bit of the interface identifier is not inverted in this
1510          * case.  Hence the resulting interface identifier has local
1511          * scope according to RFC2373.
1512          */
1513         if (dev->dev_id) {
1514                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1515                 eui[4] = dev->dev_id & 0xFF;
1516         } else {
1517                 eui[3] = 0xFF;
1518                 eui[4] = 0xFE;
1519                 eui[0] ^= 2;
1520         }
1521         return 0;
1522 }
1523
1524 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1525 {
1526         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1527         if (dev->addr_len != ARCNET_ALEN)
1528                 return -1;
1529         memset(eui, 0, 7);
1530         eui[7] = *(u8*)dev->dev_addr;
1531         return 0;
1532 }
1533
1534 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1535 {
1536         if (dev->addr_len != INFINIBAND_ALEN)
1537                 return -1;
1538         memcpy(eui, dev->dev_addr + 12, 8);
1539         eui[0] |= 2;
1540         return 0;
1541 }
1542
1543 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1544 {
1545         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1546                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1547                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1548                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1549                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1550                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1551         eui[1] = 0;
1552         eui[2] = 0x5E;
1553         eui[3] = 0xFE;
1554         memcpy(eui + 4, &addr, 4);
1555         return 0;
1556 }
1557 EXPORT_SYMBOL(__ipv6_isatap_ifid);
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 ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1567 {
1568         switch (dev->type) {
1569         case ARPHRD_ETHER:
1570         case ARPHRD_FDDI:
1571         case ARPHRD_IEEE802_TR:
1572                 return addrconf_ifid_eui48(eui, dev);
1573         case ARPHRD_ARCNET:
1574                 return addrconf_ifid_arcnet(eui, dev);
1575         case ARPHRD_INFINIBAND:
1576                 return addrconf_ifid_infiniband(eui, dev);
1577         case ARPHRD_SIT:
1578                 return addrconf_ifid_sit(eui, dev);
1579         }
1580         return -1;
1581 }
1582
1583 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1584 {
1585         int err = -1;
1586         struct inet6_ifaddr *ifp;
1587
1588         read_lock_bh(&idev->lock);
1589         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1590                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1591                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1592                         err = 0;
1593                         break;
1594                 }
1595         }
1596         read_unlock_bh(&idev->lock);
1597         return err;
1598 }
1599
1600 #ifdef CONFIG_IPV6_PRIVACY
1601 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1602 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1603 {
1604 regen:
1605         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1606         idev->rndid[0] &= ~0x02;
1607
1608         /*
1609          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1610          * check if generated address is not inappropriate
1611          *
1612          *  - Reserved subnet anycast (RFC 2526)
1613          *      11111101 11....11 1xxxxxxx
1614          *  - ISATAP (RFC4214) 6.1
1615          *      00-00-5E-FE-xx-xx-xx-xx
1616          *  - value 0
1617          *  - XXX: already assigned to an address on the device
1618          */
1619         if (idev->rndid[0] == 0xfd &&
1620             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1621             (idev->rndid[7]&0x80))
1622                 goto regen;
1623         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1624                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1625                         goto regen;
1626                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627                         goto regen;
1628         }
1629
1630         return 0;
1631 }
1632
1633 static void ipv6_regen_rndid(unsigned long data)
1634 {
1635         struct inet6_dev *idev = (struct inet6_dev *) data;
1636         unsigned long expires;
1637
1638         rcu_read_lock_bh();
1639         write_lock_bh(&idev->lock);
1640
1641         if (idev->dead)
1642                 goto out;
1643
1644         if (__ipv6_regen_rndid(idev) < 0)
1645                 goto out;
1646
1647         expires = jiffies +
1648                 idev->cnf.temp_prefered_lft * HZ -
1649                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1650         if (time_before(expires, jiffies)) {
1651                 printk(KERN_WARNING
1652                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1653                         idev->dev->name);
1654                 goto out;
1655         }
1656
1657         if (!mod_timer(&idev->regen_timer, expires))
1658                 in6_dev_hold(idev);
1659
1660 out:
1661         write_unlock_bh(&idev->lock);
1662         rcu_read_unlock_bh();
1663         in6_dev_put(idev);
1664 }
1665
1666 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1667         int ret = 0;
1668
1669         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1670                 ret = __ipv6_regen_rndid(idev);
1671         return ret;
1672 }
1673 #endif
1674
1675 /*
1676  *      Add prefix route.
1677  */
1678
1679 static void
1680 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1681                       unsigned long expires, u32 flags)
1682 {
1683         struct fib6_config cfg = {
1684                 .fc_table = RT6_TABLE_PREFIX,
1685                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1686                 .fc_ifindex = dev->ifindex,
1687                 .fc_expires = expires,
1688                 .fc_dst_len = plen,
1689                 .fc_flags = RTF_UP | flags,
1690                 .fc_nlinfo.nl_net = dev_net(dev),
1691                 .fc_protocol = RTPROT_KERNEL,
1692         };
1693
1694         ipv6_addr_copy(&cfg.fc_dst, pfx);
1695
1696         /* Prevent useless cloning on PtP SIT.
1697            This thing is done here expecting that the whole
1698            class of non-broadcast devices need not cloning.
1699          */
1700 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1701         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1702                 cfg.fc_flags |= RTF_NONEXTHOP;
1703 #endif
1704
1705         ip6_route_add(&cfg);
1706 }
1707
1708 /* Create "default" multicast route to the interface */
1709
1710 static void addrconf_add_mroute(struct net_device *dev)
1711 {
1712         struct fib6_config cfg = {
1713                 .fc_table = RT6_TABLE_LOCAL,
1714                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1715                 .fc_ifindex = dev->ifindex,
1716                 .fc_dst_len = 8,
1717                 .fc_flags = RTF_UP,
1718                 .fc_nlinfo.nl_net = dev_net(dev),
1719         };
1720
1721         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1722
1723         ip6_route_add(&cfg);
1724 }
1725
1726 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1727 static void sit_route_add(struct net_device *dev)
1728 {
1729         struct fib6_config cfg = {
1730                 .fc_table = RT6_TABLE_MAIN,
1731                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1732                 .fc_ifindex = dev->ifindex,
1733                 .fc_dst_len = 96,
1734                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1735                 .fc_nlinfo.nl_net = dev_net(dev),
1736         };
1737
1738         /* prefix length - 96 bits "::d.d.d.d" */
1739         ip6_route_add(&cfg);
1740 }
1741 #endif
1742
1743 static void addrconf_add_lroute(struct net_device *dev)
1744 {
1745         struct in6_addr addr;
1746
1747         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1748         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1749 }
1750
1751 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1752 {
1753         struct inet6_dev *idev;
1754
1755         ASSERT_RTNL();
1756
1757         if ((idev = ipv6_find_idev(dev)) == NULL)
1758                 return NULL;
1759
1760         /* Add default multicast route */
1761         addrconf_add_mroute(dev);
1762
1763         /* Add link local route */
1764         addrconf_add_lroute(dev);
1765         return idev;
1766 }
1767
1768 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1769 {
1770         struct prefix_info *pinfo;
1771         __u32 valid_lft;
1772         __u32 prefered_lft;
1773         int addr_type;
1774         struct inet6_dev *in6_dev;
1775
1776         pinfo = (struct prefix_info *) opt;
1777
1778         if (len < sizeof(struct prefix_info)) {
1779                 ADBG(("addrconf: prefix option too short\n"));
1780                 return;
1781         }
1782
1783         /*
1784          *      Validation checks ([ADDRCONF], page 19)
1785          */
1786
1787         addr_type = ipv6_addr_type(&pinfo->prefix);
1788
1789         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1790                 return;
1791
1792         valid_lft = ntohl(pinfo->valid);
1793         prefered_lft = ntohl(pinfo->prefered);
1794
1795         if (prefered_lft > valid_lft) {
1796                 if (net_ratelimit())
1797                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1798                 return;
1799         }
1800
1801         in6_dev = in6_dev_get(dev);
1802
1803         if (in6_dev == NULL) {
1804                 if (net_ratelimit())
1805                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1806                 return;
1807         }
1808
1809         /*
1810          *      Two things going on here:
1811          *      1) Add routes for on-link prefixes
1812          *      2) Configure prefixes with the auto flag set
1813          */
1814
1815         if (pinfo->onlink) {
1816                 struct rt6_info *rt;
1817                 unsigned long rt_expires;
1818
1819                 /* Avoid arithmetic overflow. Really, we could
1820                  * save rt_expires in seconds, likely valid_lft,
1821                  * but it would require division in fib gc, that it
1822                  * not good.
1823                  */
1824                 if (HZ > USER_HZ)
1825                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1826                 else
1827                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1828
1829                 if (addrconf_finite_timeout(rt_expires))
1830                         rt_expires *= HZ;
1831
1832                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1833                                 dev->ifindex, 1);
1834
1835                 if (rt && addrconf_is_prefix_route(rt)) {
1836                         /* Autoconf prefix route */
1837                         if (valid_lft == 0) {
1838                                 ip6_del_rt(rt);
1839                                 rt = NULL;
1840                         } else if (addrconf_finite_timeout(rt_expires)) {
1841                                 /* not infinity */
1842                                 rt->rt6i_expires = jiffies + rt_expires;
1843                                 rt->rt6i_flags |= RTF_EXPIRES;
1844                         } else {
1845                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1846                                 rt->rt6i_expires = 0;
1847                         }
1848                 } else if (valid_lft) {
1849                         clock_t expires = 0;
1850                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1851                         if (addrconf_finite_timeout(rt_expires)) {
1852                                 /* not infinity */
1853                                 flags |= RTF_EXPIRES;
1854                                 expires = jiffies_to_clock_t(rt_expires);
1855                         }
1856                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1857                                               dev, expires, flags);
1858                 }
1859                 if (rt)
1860                         dst_release(&rt->u.dst);
1861         }
1862
1863         /* Try to figure out our local address for this prefix */
1864
1865         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1866                 struct inet6_ifaddr * ifp;
1867                 struct in6_addr addr;
1868                 int create = 0, update_lft = 0;
1869                 struct net *net = dev_net(dev);
1870
1871                 if (pinfo->prefix_len == 64) {
1872                         memcpy(&addr, &pinfo->prefix, 8);
1873                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1874                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1875                                 in6_dev_put(in6_dev);
1876                                 return;
1877                         }
1878                         goto ok;
1879                 }
1880                 if (net_ratelimit())
1881                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1882                                pinfo->prefix_len);
1883                 in6_dev_put(in6_dev);
1884                 return;
1885
1886 ok:
1887
1888                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1889
1890                 if (ifp == NULL && valid_lft) {
1891                         int max_addresses = in6_dev->cnf.max_addresses;
1892                         u32 addr_flags = 0;
1893
1894 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1895                         if (in6_dev->cnf.optimistic_dad &&
1896                             !net->ipv6.devconf_all->forwarding)
1897                                 addr_flags = IFA_F_OPTIMISTIC;
1898 #endif
1899
1900                         /* Do not allow to create too much of autoconfigured
1901                          * addresses; this would be too easy way to crash kernel.
1902                          */
1903                         if (!max_addresses ||
1904                             ipv6_count_addresses(in6_dev) < max_addresses)
1905                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1906                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1907                                                     addr_flags);
1908
1909                         if (!ifp || IS_ERR(ifp)) {
1910                                 in6_dev_put(in6_dev);
1911                                 return;
1912                         }
1913
1914                         update_lft = create = 1;
1915                         ifp->cstamp = jiffies;
1916                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1917                 }
1918
1919                 if (ifp) {
1920                         int flags;
1921                         unsigned long now;
1922 #ifdef CONFIG_IPV6_PRIVACY
1923                         struct inet6_ifaddr *ift;
1924 #endif
1925                         u32 stored_lft;
1926
1927                         /* update lifetime (RFC2462 5.5.3 e) */
1928                         spin_lock(&ifp->lock);
1929                         now = jiffies;
1930                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1931                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1932                         else
1933                                 stored_lft = 0;
1934                         if (!update_lft && stored_lft) {
1935                                 if (valid_lft > MIN_VALID_LIFETIME ||
1936                                     valid_lft > stored_lft)
1937                                         update_lft = 1;
1938                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1939                                         /* valid_lft <= stored_lft is always true */
1940                                         /* XXX: IPsec */
1941                                         update_lft = 0;
1942                                 } else {
1943                                         valid_lft = MIN_VALID_LIFETIME;
1944                                         if (valid_lft < prefered_lft)
1945                                                 prefered_lft = valid_lft;
1946                                         update_lft = 1;
1947                                 }
1948                         }
1949
1950                         if (update_lft) {
1951                                 ifp->valid_lft = valid_lft;
1952                                 ifp->prefered_lft = prefered_lft;
1953                                 ifp->tstamp = now;
1954                                 flags = ifp->flags;
1955                                 ifp->flags &= ~IFA_F_DEPRECATED;
1956                                 spin_unlock(&ifp->lock);
1957
1958                                 if (!(flags&IFA_F_TENTATIVE))
1959                                         ipv6_ifa_notify(0, ifp);
1960                         } else
1961                                 spin_unlock(&ifp->lock);
1962
1963 #ifdef CONFIG_IPV6_PRIVACY
1964                         read_lock_bh(&in6_dev->lock);
1965                         /* update all temporary addresses in the list */
1966                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1967                                 /*
1968                                  * When adjusting the lifetimes of an existing
1969                                  * temporary address, only lower the lifetimes.
1970                                  * Implementations must not increase the
1971                                  * lifetimes of an existing temporary address
1972                                  * when processing a Prefix Information Option.
1973                                  */
1974                                 if (ifp != ift->ifpub)
1975                                         continue;
1976
1977                                 spin_lock(&ift->lock);
1978                                 flags = ift->flags;
1979                                 if (ift->valid_lft > valid_lft &&
1980                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1981                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1982                                 if (ift->prefered_lft > prefered_lft &&
1983                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1984                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1985                                 spin_unlock(&ift->lock);
1986                                 if (!(flags&IFA_F_TENTATIVE))
1987                                         ipv6_ifa_notify(0, ift);
1988                         }
1989
1990                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1991                                 /*
1992                                  * When a new public address is created as described in [ADDRCONF],
1993                                  * also create a new temporary address.
1994                                  */
1995                                 read_unlock_bh(&in6_dev->lock);
1996                                 ipv6_create_tempaddr(ifp, NULL);
1997                         } else {
1998                                 read_unlock_bh(&in6_dev->lock);
1999                         }
2000 #endif
2001                         in6_ifa_put(ifp);
2002                         addrconf_verify(0);
2003                 }
2004         }
2005         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2006         in6_dev_put(in6_dev);
2007 }
2008
2009 /*
2010  *      Set destination address.
2011  *      Special case for SIT interfaces where we create a new "virtual"
2012  *      device.
2013  */
2014 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2015 {
2016         struct in6_ifreq ireq;
2017         struct net_device *dev;
2018         int err = -EINVAL;
2019
2020         rtnl_lock();
2021
2022         err = -EFAULT;
2023         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2024                 goto err_exit;
2025
2026         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2027
2028         err = -ENODEV;
2029         if (dev == NULL)
2030                 goto err_exit;
2031
2032 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2033         if (dev->type == ARPHRD_SIT) {
2034                 const struct net_device_ops *ops = dev->netdev_ops;
2035                 struct ifreq ifr;
2036                 struct ip_tunnel_parm p;
2037
2038                 err = -EADDRNOTAVAIL;
2039                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2040                         goto err_exit;
2041
2042                 memset(&p, 0, sizeof(p));
2043                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2044                 p.iph.saddr = 0;
2045                 p.iph.version = 4;
2046                 p.iph.ihl = 5;
2047                 p.iph.protocol = IPPROTO_IPV6;
2048                 p.iph.ttl = 64;
2049                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2050
2051                 if (ops->ndo_do_ioctl) {
2052                         mm_segment_t oldfs = get_fs();
2053
2054                         set_fs(KERNEL_DS);
2055                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2056                         set_fs(oldfs);
2057                 } else
2058                         err = -EOPNOTSUPP;
2059
2060                 if (err == 0) {
2061                         err = -ENOBUFS;
2062                         dev = __dev_get_by_name(net, p.name);
2063                         if (!dev)
2064                                 goto err_exit;
2065                         err = dev_open(dev);
2066                 }
2067         }
2068 #endif
2069
2070 err_exit:
2071         rtnl_unlock();
2072         return err;
2073 }
2074
2075 /*
2076  *      Manual configuration of address on an interface
2077  */
2078 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2079                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2080                           __u32 valid_lft)
2081 {
2082         struct inet6_ifaddr *ifp;
2083         struct inet6_dev *idev;
2084         struct net_device *dev;
2085         int scope;
2086         u32 flags;
2087         clock_t expires;
2088         unsigned long timeout;
2089
2090         ASSERT_RTNL();
2091
2092         if (plen > 128)
2093                 return -EINVAL;
2094
2095         /* check the lifetime */
2096         if (!valid_lft || prefered_lft > valid_lft)
2097                 return -EINVAL;
2098
2099         dev = __dev_get_by_index(net, ifindex);
2100         if (!dev)
2101                 return -ENODEV;
2102
2103         if ((idev = addrconf_add_dev(dev)) == NULL)
2104                 return -ENOBUFS;
2105
2106         scope = ipv6_addr_scope(pfx);
2107
2108         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2109         if (addrconf_finite_timeout(timeout)) {
2110                 expires = jiffies_to_clock_t(timeout * HZ);
2111                 valid_lft = timeout;
2112                 flags = RTF_EXPIRES;
2113         } else {
2114                 expires = 0;
2115                 flags = 0;
2116                 ifa_flags |= IFA_F_PERMANENT;
2117         }
2118
2119         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2120         if (addrconf_finite_timeout(timeout)) {
2121                 if (timeout == 0)
2122                         ifa_flags |= IFA_F_DEPRECATED;
2123                 prefered_lft = timeout;
2124         }
2125
2126         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2127
2128         if (!IS_ERR(ifp)) {
2129                 spin_lock_bh(&ifp->lock);
2130                 ifp->valid_lft = valid_lft;
2131                 ifp->prefered_lft = prefered_lft;
2132                 ifp->tstamp = jiffies;
2133                 spin_unlock_bh(&ifp->lock);
2134
2135                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2136                                       expires, flags);
2137                 /*
2138                  * Note that section 3.1 of RFC 4429 indicates
2139                  * that the Optimistic flag should not be set for
2140                  * manually configured addresses
2141                  */
2142                 addrconf_dad_start(ifp, 0);
2143                 in6_ifa_put(ifp);
2144                 addrconf_verify(0);
2145                 return 0;
2146         }
2147
2148         return PTR_ERR(ifp);
2149 }
2150
2151 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2152                           unsigned int plen)
2153 {
2154         struct inet6_ifaddr *ifp;
2155         struct inet6_dev *idev;
2156         struct net_device *dev;
2157
2158         if (plen > 128)
2159                 return -EINVAL;
2160
2161         dev = __dev_get_by_index(net, ifindex);
2162         if (!dev)
2163                 return -ENODEV;
2164
2165         if ((idev = __in6_dev_get(dev)) == NULL)
2166                 return -ENXIO;
2167
2168         read_lock_bh(&idev->lock);
2169         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2170                 if (ifp->prefix_len == plen &&
2171                     ipv6_addr_equal(pfx, &ifp->addr)) {
2172                         in6_ifa_hold(ifp);
2173                         read_unlock_bh(&idev->lock);
2174
2175                         ipv6_del_addr(ifp);
2176
2177                         /* If the last address is deleted administratively,
2178                            disable IPv6 on this interface.
2179                          */
2180                         if (idev->addr_list == NULL)
2181                                 addrconf_ifdown(idev->dev, 1);
2182                         return 0;
2183                 }
2184         }
2185         read_unlock_bh(&idev->lock);
2186         return -EADDRNOTAVAIL;
2187 }
2188
2189
2190 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2191 {
2192         struct in6_ifreq ireq;
2193         int err;
2194
2195         if (!capable(CAP_NET_ADMIN))
2196                 return -EPERM;
2197
2198         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2199                 return -EFAULT;
2200
2201         rtnl_lock();
2202         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2203                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2204                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2205         rtnl_unlock();
2206         return err;
2207 }
2208
2209 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2210 {
2211         struct in6_ifreq ireq;
2212         int err;
2213
2214         if (!capable(CAP_NET_ADMIN))
2215                 return -EPERM;
2216
2217         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2218                 return -EFAULT;
2219
2220         rtnl_lock();
2221         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2222                              ireq.ifr6_prefixlen);
2223         rtnl_unlock();
2224         return err;
2225 }
2226
2227 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2228                      int plen, int scope)
2229 {
2230         struct inet6_ifaddr *ifp;
2231
2232         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2233         if (!IS_ERR(ifp)) {
2234                 spin_lock_bh(&ifp->lock);
2235                 ifp->flags &= ~IFA_F_TENTATIVE;
2236                 spin_unlock_bh(&ifp->lock);
2237                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2238                 in6_ifa_put(ifp);
2239         }
2240 }
2241
2242 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2243 static void sit_add_v4_addrs(struct inet6_dev *idev)
2244 {
2245         struct in6_addr addr;
2246         struct net_device *dev;
2247         struct net *net = dev_net(idev->dev);
2248         int scope;
2249
2250         ASSERT_RTNL();
2251
2252         memset(&addr, 0, sizeof(struct in6_addr));
2253         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2254
2255         if (idev->dev->flags&IFF_POINTOPOINT) {
2256                 addr.s6_addr32[0] = htonl(0xfe800000);
2257                 scope = IFA_LINK;
2258         } else {
2259                 scope = IPV6_ADDR_COMPATv4;
2260         }
2261
2262         if (addr.s6_addr32[3]) {
2263                 add_addr(idev, &addr, 128, scope);
2264                 return;
2265         }
2266
2267         for_each_netdev(net, dev) {
2268                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2269                 if (in_dev && (dev->flags & IFF_UP)) {
2270                         struct in_ifaddr * ifa;
2271
2272                         int flag = scope;
2273
2274                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2275                                 int plen;
2276
2277                                 addr.s6_addr32[3] = ifa->ifa_local;
2278
2279                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2280                                         continue;
2281                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2282                                         if (idev->dev->flags&IFF_POINTOPOINT)
2283                                                 continue;
2284                                         flag |= IFA_HOST;
2285                                 }
2286                                 if (idev->dev->flags&IFF_POINTOPOINT)
2287                                         plen = 64;
2288                                 else
2289                                         plen = 96;
2290
2291                                 add_addr(idev, &addr, plen, flag);
2292                         }
2293                 }
2294         }
2295 }
2296 #endif
2297
2298 static void init_loopback(struct net_device *dev)
2299 {
2300         struct inet6_dev  *idev;
2301
2302         /* ::1 */
2303
2304         ASSERT_RTNL();
2305
2306         if ((idev = ipv6_find_idev(dev)) == NULL) {
2307                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2308                 return;
2309         }
2310
2311         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2312 }
2313
2314 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2315 {
2316         struct inet6_ifaddr * ifp;
2317         u32 addr_flags = IFA_F_PERMANENT;
2318
2319 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2320         if (idev->cnf.optimistic_dad &&
2321             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2322                 addr_flags |= IFA_F_OPTIMISTIC;
2323 #endif
2324
2325
2326         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2327         if (!IS_ERR(ifp)) {
2328                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2329                 addrconf_dad_start(ifp, 0);
2330                 in6_ifa_put(ifp);
2331         }
2332 }
2333
2334 static void addrconf_dev_config(struct net_device *dev)
2335 {
2336         struct in6_addr addr;
2337         struct inet6_dev    * idev;
2338
2339         ASSERT_RTNL();
2340
2341         if ((dev->type != ARPHRD_ETHER) &&
2342             (dev->type != ARPHRD_FDDI) &&
2343             (dev->type != ARPHRD_IEEE802_TR) &&
2344             (dev->type != ARPHRD_ARCNET) &&
2345             (dev->type != ARPHRD_INFINIBAND)) {
2346                 /* Alas, we support only Ethernet autoconfiguration. */
2347                 return;
2348         }
2349
2350         idev = addrconf_add_dev(dev);
2351         if (idev == NULL)
2352                 return;
2353
2354         memset(&addr, 0, sizeof(struct in6_addr));
2355         addr.s6_addr32[0] = htonl(0xFE800000);
2356
2357         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2358                 addrconf_add_linklocal(idev, &addr);
2359 }
2360
2361 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2362 static void addrconf_sit_config(struct net_device *dev)
2363 {
2364         struct inet6_dev *idev;
2365
2366         ASSERT_RTNL();
2367
2368         /*
2369          * Configure the tunnel with one of our IPv4
2370          * addresses... we should configure all of
2371          * our v4 addrs in the tunnel
2372          */
2373
2374         if ((idev = ipv6_find_idev(dev)) == NULL) {
2375                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2376                 return;
2377         }
2378
2379         if (dev->priv_flags & IFF_ISATAP) {
2380                 struct in6_addr addr;
2381
2382                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2383                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2384                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2385                         addrconf_add_linklocal(idev, &addr);
2386                 return;
2387         }
2388
2389         sit_add_v4_addrs(idev);
2390
2391         if (dev->flags&IFF_POINTOPOINT) {
2392                 addrconf_add_mroute(dev);
2393                 addrconf_add_lroute(dev);
2394         } else
2395                 sit_route_add(dev);
2396 }
2397 #endif
2398
2399 static inline int
2400 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2401 {
2402         struct in6_addr lladdr;
2403
2404         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2405                 addrconf_add_linklocal(idev, &lladdr);
2406                 return 0;
2407         }
2408         return -1;
2409 }
2410
2411 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2412 {
2413         struct net_device *link_dev;
2414         struct net *net = dev_net(idev->dev);
2415
2416         /* first try to inherit the link-local address from the link device */
2417         if (idev->dev->iflink &&
2418             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2419                 if (!ipv6_inherit_linklocal(idev, link_dev))
2420                         return;
2421         }
2422         /* then try to inherit it from any device */
2423         for_each_netdev(net, link_dev) {
2424                 if (!ipv6_inherit_linklocal(idev, link_dev))
2425                         return;
2426         }
2427         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2428 }
2429
2430 /*
2431  * Autoconfigure tunnel with a link-local address so routing protocols,
2432  * DHCPv6, MLD etc. can be run over the virtual link
2433  */
2434
2435 static void addrconf_ip6_tnl_config(struct net_device *dev)
2436 {
2437         struct inet6_dev *idev;
2438
2439         ASSERT_RTNL();
2440
2441         if ((idev = addrconf_add_dev(dev)) == NULL) {
2442                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2443                 return;
2444         }
2445         ip6_tnl_add_linklocal(idev);
2446 }
2447
2448 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2449                            void * data)
2450 {
2451         struct net_device *dev = (struct net_device *) data;
2452         struct inet6_dev *idev = __in6_dev_get(dev);
2453         int run_pending = 0;
2454         int err;
2455
2456         switch(event) {
2457         case NETDEV_REGISTER:
2458                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2459                         idev = ipv6_add_dev(dev);
2460                         if (!idev)
2461                                 return notifier_from_errno(-ENOMEM);
2462                 }
2463                 break;
2464         case NETDEV_UP:
2465         case NETDEV_CHANGE:
2466                 if (dev->flags & IFF_SLAVE)
2467                         break;
2468
2469                 if (event == NETDEV_UP) {
2470                         if (!addrconf_qdisc_ok(dev)) {
2471                                 /* device is not ready yet. */
2472                                 printk(KERN_INFO
2473                                         "ADDRCONF(NETDEV_UP): %s: "
2474                                         "link is not ready\n",
2475                                         dev->name);
2476                                 break;
2477                         }
2478
2479                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2480                                 idev = ipv6_add_dev(dev);
2481
2482                         if (idev) {
2483                                 idev->if_flags |= IF_READY;
2484                                 run_pending = 1;
2485                         }
2486                 } else {
2487                         if (!addrconf_qdisc_ok(dev)) {
2488                                 /* device is still not ready. */
2489                                 break;
2490                         }
2491
2492                         if (idev) {
2493                                 if (idev->if_flags & IF_READY) {
2494                                         /* device is already configured. */
2495                                         break;
2496                                 }
2497                                 idev->if_flags |= IF_READY;
2498                         }
2499
2500                         printk(KERN_INFO
2501                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2502                                         "link becomes ready\n",
2503                                         dev->name);
2504
2505                         run_pending = 1;
2506                 }
2507
2508                 switch(dev->type) {
2509 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2510                 case ARPHRD_SIT:
2511                         addrconf_sit_config(dev);
2512                         break;
2513 #endif
2514                 case ARPHRD_TUNNEL6:
2515                         addrconf_ip6_tnl_config(dev);
2516                         break;
2517                 case ARPHRD_LOOPBACK:
2518                         init_loopback(dev);
2519                         break;
2520
2521                 default:
2522                         addrconf_dev_config(dev);
2523                         break;
2524                 }
2525                 if (idev) {
2526                         if (run_pending)
2527                                 addrconf_dad_run(idev);
2528
2529                         /* If the MTU changed during the interface down, when the
2530                            interface up, the changed MTU must be reflected in the
2531                            idev as well as routers.
2532                          */
2533                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2534                                 rt6_mtu_change(dev, dev->mtu);
2535                                 idev->cnf.mtu6 = dev->mtu;
2536                         }
2537                         idev->tstamp = jiffies;
2538                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2539                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2540                            stop IPv6 on this interface.
2541                          */
2542                         if (dev->mtu < IPV6_MIN_MTU)
2543                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2544                 }
2545                 break;
2546
2547         case NETDEV_CHANGEMTU:
2548                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2549                         rt6_mtu_change(dev, dev->mtu);
2550                         idev->cnf.mtu6 = dev->mtu;
2551                         break;
2552                 }
2553
2554                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2555                         idev = ipv6_add_dev(dev);
2556                         if (idev)
2557                                 break;
2558                 }
2559
2560                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2561
2562         case NETDEV_DOWN:
2563         case NETDEV_UNREGISTER:
2564                 /*
2565                  *      Remove all addresses from this interface.
2566                  */
2567                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2568                 break;
2569
2570         case NETDEV_CHANGENAME:
2571                 if (idev) {
2572                         snmp6_unregister_dev(idev);
2573                         addrconf_sysctl_unregister(idev);
2574                         addrconf_sysctl_register(idev);
2575                         err = snmp6_register_dev(idev);
2576                         if (err)
2577                                 return notifier_from_errno(err);
2578                 }
2579                 break;
2580         }
2581
2582         return NOTIFY_OK;
2583 }
2584
2585 /*
2586  *      addrconf module should be notified of a device going up
2587  */
2588 static struct notifier_block ipv6_dev_notf = {
2589         .notifier_call = addrconf_notify,
2590         .priority = 0
2591 };
2592
2593 static int addrconf_ifdown(struct net_device *dev, int how)
2594 {
2595         struct inet6_dev *idev;
2596         struct inet6_ifaddr *ifa, **bifa;
2597         struct net *net = dev_net(dev);
2598         int i;
2599
2600         ASSERT_RTNL();
2601
2602         if ((dev->flags & IFF_LOOPBACK) && how == 1)
2603                 how = 0;
2604
2605         rt6_ifdown(net, dev);
2606         neigh_ifdown(&nd_tbl, dev);
2607
2608         idev = __in6_dev_get(dev);
2609         if (idev == NULL)
2610                 return -ENODEV;
2611
2612         /* Step 1: remove reference to ipv6 device from parent device.
2613                    Do not dev_put!
2614          */
2615         if (how) {
2616                 idev->dead = 1;
2617
2618                 /* protected by rtnl_lock */
2619                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2620
2621                 /* Step 1.5: remove snmp6 entry */
2622                 snmp6_unregister_dev(idev);
2623
2624         }
2625
2626         /* Step 2: clear hash table */
2627         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2628                 bifa = &inet6_addr_lst[i];
2629
2630                 write_lock_bh(&addrconf_hash_lock);
2631                 while ((ifa = *bifa) != NULL) {
2632                         if (ifa->idev == idev) {
2633                                 *bifa = ifa->lst_next;
2634                                 ifa->lst_next = NULL;
2635                                 addrconf_del_timer(ifa);
2636                                 in6_ifa_put(ifa);
2637                                 continue;
2638                         }
2639                         bifa = &ifa->lst_next;
2640                 }
2641                 write_unlock_bh(&addrconf_hash_lock);
2642         }
2643
2644         write_lock_bh(&idev->lock);
2645
2646         /* Step 3: clear flags for stateless addrconf */
2647         if (!how)
2648                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2649
2650         /* Step 4: clear address list */
2651 #ifdef CONFIG_IPV6_PRIVACY
2652         if (how && del_timer(&idev->regen_timer))
2653                 in6_dev_put(idev);
2654
2655         /* clear tempaddr list */
2656         while ((ifa = idev->tempaddr_list) != NULL) {
2657                 idev->tempaddr_list = ifa->tmp_next;
2658                 ifa->tmp_next = NULL;
2659                 ifa->dead = 1;
2660                 write_unlock_bh(&idev->lock);
2661                 spin_lock_bh(&ifa->lock);
2662
2663                 if (ifa->ifpub) {
2664                         in6_ifa_put(ifa->ifpub);
2665                         ifa->ifpub = NULL;
2666                 }
2667                 spin_unlock_bh(&ifa->lock);
2668                 in6_ifa_put(ifa);
2669                 write_lock_bh(&idev->lock);
2670         }
2671 #endif
2672         while ((ifa = idev->addr_list) != NULL) {
2673                 idev->addr_list = ifa->if_next;
2674                 ifa->if_next = NULL;
2675                 ifa->dead = 1;
2676                 addrconf_del_timer(ifa);
2677                 write_unlock_bh(&idev->lock);
2678
2679                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2680                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2681                 in6_ifa_put(ifa);
2682
2683                 write_lock_bh(&idev->lock);
2684         }
2685         write_unlock_bh(&idev->lock);
2686
2687         /* Step 5: Discard multicast list */
2688
2689         if (how)
2690                 ipv6_mc_destroy_dev(idev);
2691         else
2692                 ipv6_mc_down(idev);
2693
2694         idev->tstamp = jiffies;
2695
2696         /* Shot the device (if unregistered) */
2697
2698         if (how) {
2699                 addrconf_sysctl_unregister(idev);
2700                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2701                 neigh_ifdown(&nd_tbl, dev);
2702                 in6_dev_put(idev);
2703         }
2704         return 0;
2705 }
2706
2707 static void addrconf_rs_timer(unsigned long data)
2708 {
2709         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2710
2711         if (ifp->idev->cnf.forwarding)
2712                 goto out;
2713
2714         if (ifp->idev->if_flags & IF_RA_RCVD) {
2715                 /*
2716                  *      Announcement received after solicitation
2717                  *      was sent
2718                  */
2719                 goto out;
2720         }
2721
2722         spin_lock(&ifp->lock);
2723         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2724                 /* The wait after the last probe can be shorter */
2725                 addrconf_mod_timer(ifp, AC_RS,
2726                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2727                                    ifp->idev->cnf.rtr_solicit_delay :
2728                                    ifp->idev->cnf.rtr_solicit_interval);
2729                 spin_unlock(&ifp->lock);
2730
2731                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2732         } else {
2733                 spin_unlock(&ifp->lock);
2734                 /*
2735                  * Note: we do not support deprecated "all on-link"
2736                  * assumption any longer.
2737                  */
2738                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2739                        ifp->idev->dev->name);
2740         }
2741
2742 out:
2743         in6_ifa_put(ifp);
2744 }
2745
2746 /*
2747  *      Duplicate Address Detection
2748  */
2749 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2750 {
2751         unsigned long rand_num;
2752         struct inet6_dev *idev = ifp->idev;
2753
2754         if (ifp->flags & IFA_F_OPTIMISTIC)
2755                 rand_num = 0;
2756         else
2757                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2758
2759         ifp->probes = idev->cnf.dad_transmits;
2760         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2761 }
2762
2763 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2764 {
2765         struct inet6_dev *idev = ifp->idev;
2766         struct net_device *dev = idev->dev;
2767
2768         addrconf_join_solict(dev, &ifp->addr);
2769
2770         net_srandom(ifp->addr.s6_addr32[3]);
2771
2772         read_lock_bh(&idev->lock);
2773         if (ifp->dead)
2774                 goto out;
2775         spin_lock_bh(&ifp->lock);
2776
2777         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2778             idev->cnf.accept_dad < 1 ||
2779             !(ifp->flags&IFA_F_TENTATIVE) ||
2780             ifp->flags & IFA_F_NODAD) {
2781                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2782                 spin_unlock_bh(&ifp->lock);
2783                 read_unlock_bh(&idev->lock);
2784
2785                 addrconf_dad_completed(ifp);
2786                 return;
2787         }
2788
2789         if (!(idev->if_flags & IF_READY)) {
2790                 spin_unlock_bh(&ifp->lock);
2791                 read_unlock_bh(&idev->lock);
2792                 /*
2793                  * If the defice is not ready:
2794                  * - keep it tentative if it is a permanent address.
2795                  * - otherwise, kill it.
2796                  */
2797                 in6_ifa_hold(ifp);
2798                 addrconf_dad_stop(ifp);
2799                 return;
2800         }
2801
2802         /*
2803          * Optimistic nodes can start receiving
2804          * Frames right away
2805          */
2806         if(ifp->flags & IFA_F_OPTIMISTIC)
2807                 ip6_ins_rt(ifp->rt);
2808
2809         addrconf_dad_kick(ifp);
2810         spin_unlock_bh(&ifp->lock);
2811 out:
2812         read_unlock_bh(&idev->lock);
2813 }
2814
2815 static void addrconf_dad_timer(unsigned long data)
2816 {
2817         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2818         struct inet6_dev *idev = ifp->idev;
2819         struct in6_addr mcaddr;
2820
2821         read_lock_bh(&idev->lock);
2822         if (idev->dead) {
2823                 read_unlock_bh(&idev->lock);
2824                 goto out;
2825         }
2826         if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2827                 read_unlock_bh(&idev->lock);
2828                 addrconf_dad_failure(ifp);
2829                 return;
2830         }
2831         spin_lock_bh(&ifp->lock);
2832         if (ifp->probes == 0) {
2833                 /*
2834                  * DAD was successful
2835                  */
2836
2837                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2838                 spin_unlock_bh(&ifp->lock);
2839                 read_unlock_bh(&idev->lock);
2840
2841                 addrconf_dad_completed(ifp);
2842
2843                 goto out;
2844         }
2845
2846         ifp->probes--;
2847         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2848         spin_unlock_bh(&ifp->lock);
2849         read_unlock_bh(&idev->lock);
2850
2851         /* send a neighbour solicitation for our addr */
2852         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2853         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2854 out:
2855         in6_ifa_put(ifp);
2856 }
2857
2858 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2859 {
2860         struct net_device *     dev = ifp->idev->dev;
2861
2862         /*
2863          *      Configure the address for reception. Now it is valid.
2864          */
2865
2866         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2867
2868         /* If added prefix is link local and forwarding is off,
2869            start sending router solicitations.
2870          */
2871
2872         if (ifp->idev->cnf.forwarding == 0 &&
2873             ifp->idev->cnf.rtr_solicits > 0 &&
2874             (dev->flags&IFF_LOOPBACK) == 0 &&
2875             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2876                 /*
2877                  *      If a host as already performed a random delay
2878                  *      [...] as part of DAD [...] there is no need
2879                  *      to delay again before sending the first RS
2880                  */
2881                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2882
2883                 spin_lock_bh(&ifp->lock);
2884                 ifp->probes = 1;
2885                 ifp->idev->if_flags |= IF_RS_SENT;
2886                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2887                 spin_unlock_bh(&ifp->lock);
2888         }
2889 }
2890
2891 static void addrconf_dad_run(struct inet6_dev *idev) {
2892         struct inet6_ifaddr *ifp;
2893
2894         read_lock_bh(&idev->lock);
2895         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2896                 spin_lock_bh(&ifp->lock);
2897                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2898                         spin_unlock_bh(&ifp->lock);
2899                         continue;
2900                 }
2901                 spin_unlock_bh(&ifp->lock);
2902                 addrconf_dad_kick(ifp);
2903         }
2904         read_unlock_bh(&idev->lock);
2905 }
2906
2907 #ifdef CONFIG_PROC_FS
2908 struct if6_iter_state {
2909         struct seq_net_private p;
2910         int bucket;
2911 };
2912
2913 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2914 {
2915         struct inet6_ifaddr *ifa = NULL;
2916         struct if6_iter_state *state = seq->private;
2917         struct net *net = seq_file_net(seq);
2918
2919         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2920                 ifa = inet6_addr_lst[state->bucket];
2921
2922                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2923                         ifa = ifa->lst_next;
2924                 if (ifa)
2925                         break;
2926         }
2927         return ifa;
2928 }
2929
2930 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2931 {
2932         struct if6_iter_state *state = seq->private;
2933         struct net *net = seq_file_net(seq);
2934
2935         ifa = ifa->lst_next;
2936 try_again:
2937         if (ifa) {
2938                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2939                         ifa = ifa->lst_next;
2940                         goto try_again;
2941                 }
2942         }
2943
2944         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2945                 ifa = inet6_addr_lst[state->bucket];
2946                 goto try_again;
2947         }
2948
2949         return ifa;
2950 }
2951
2952 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2953 {
2954         struct inet6_ifaddr *ifa = if6_get_first(seq);
2955
2956         if (ifa)
2957                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2958                         --pos;
2959         return pos ? NULL : ifa;
2960 }
2961
2962 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2963         __acquires(addrconf_hash_lock)
2964 {
2965         read_lock_bh(&addrconf_hash_lock);
2966         return if6_get_idx(seq, *pos);
2967 }
2968
2969 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2970 {
2971         struct inet6_ifaddr *ifa;
2972
2973         ifa = if6_get_next(seq, v);
2974         ++*pos;
2975         return ifa;
2976 }
2977
2978 static void if6_seq_stop(struct seq_file *seq, void *v)
2979         __releases(addrconf_hash_lock)
2980 {
2981         read_unlock_bh(&addrconf_hash_lock);
2982 }
2983
2984 static int if6_seq_show(struct seq_file *seq, void *v)
2985 {
2986         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2987         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2988                    &ifp->addr,
2989                    ifp->idev->dev->ifindex,
2990                    ifp->prefix_len,
2991                    ifp->scope,
2992                    ifp->flags,
2993                    ifp->idev->dev->name);
2994         return 0;
2995 }
2996
2997 static const struct seq_operations if6_seq_ops = {
2998         .start  = if6_seq_start,
2999         .next   = if6_seq_next,
3000         .show   = if6_seq_show,
3001         .stop   = if6_seq_stop,
3002 };
3003
3004 static int if6_seq_open(struct inode *inode, struct file *file)
3005 {
3006         return seq_open_net(inode, file, &if6_seq_ops,
3007                             sizeof(struct if6_iter_state));
3008 }
3009
3010 static const struct file_operations if6_fops = {
3011         .owner          = THIS_MODULE,
3012         .open           = if6_seq_open,
3013         .read           = seq_read,
3014         .llseek         = seq_lseek,
3015         .release        = seq_release_net,
3016 };
3017
3018 static int if6_proc_net_init(struct net *net)
3019 {
3020         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3021                 return -ENOMEM;
3022         return 0;
3023 }
3024
3025 static void if6_proc_net_exit(struct net *net)
3026 {
3027        proc_net_remove(net, "if_inet6");
3028 }
3029
3030 static struct pernet_operations if6_proc_net_ops = {
3031        .init = if6_proc_net_init,
3032        .exit = if6_proc_net_exit,
3033 };
3034
3035 int __init if6_proc_init(void)
3036 {
3037         return register_pernet_subsys(&if6_proc_net_ops);
3038 }
3039
3040 void if6_proc_exit(void)
3041 {
3042         unregister_pernet_subsys(&if6_proc_net_ops);
3043 }
3044 #endif  /* CONFIG_PROC_FS */
3045
3046 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3047 /* Check if address is a home address configured on any interface. */
3048 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3049 {
3050         int ret = 0;
3051         struct inet6_ifaddr * ifp;
3052         u8 hash = ipv6_addr_hash(addr);
3053         read_lock_bh(&addrconf_hash_lock);
3054         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3055                 if (!net_eq(dev_net(ifp->idev->dev), net))
3056                         continue;
3057                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3058                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3059                         ret = 1;
3060                         break;
3061                 }
3062         }
3063         read_unlock_bh(&addrconf_hash_lock);
3064         return ret;
3065 }
3066 #endif
3067
3068 /*
3069  *      Periodic address status verification
3070  */
3071
3072 static void addrconf_verify(unsigned long foo)
3073 {
3074         struct inet6_ifaddr *ifp;
3075         unsigned long now, next;
3076         int i;
3077
3078         spin_lock_bh(&addrconf_verify_lock);
3079         now = jiffies;
3080         next = now + ADDR_CHECK_FREQUENCY;
3081
3082         del_timer(&addr_chk_timer);
3083
3084         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3085
3086 restart:
3087                 read_lock(&addrconf_hash_lock);
3088                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3089                         unsigned long age;
3090 #ifdef CONFIG_IPV6_PRIVACY
3091                         unsigned long regen_advance;
3092 #endif
3093
3094                         if (ifp->flags & IFA_F_PERMANENT)
3095                                 continue;
3096
3097                         spin_lock(&ifp->lock);
3098                         age = (now - ifp->tstamp) / HZ;
3099
3100 #ifdef CONFIG_IPV6_PRIVACY
3101                         regen_advance = ifp->idev->cnf.regen_max_retry *
3102                                         ifp->idev->cnf.dad_transmits *
3103                                         ifp->idev->nd_parms->retrans_time / HZ;
3104 #endif
3105
3106                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3107                             age >= ifp->valid_lft) {
3108                                 spin_unlock(&ifp->lock);
3109                                 in6_ifa_hold(ifp);
3110                                 read_unlock(&addrconf_hash_lock);
3111                                 ipv6_del_addr(ifp);
3112                                 goto restart;
3113                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3114                                 spin_unlock(&ifp->lock);
3115                                 continue;
3116                         } else if (age >= ifp->prefered_lft) {
3117                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3118                                 int deprecate = 0;
3119
3120                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3121                                         deprecate = 1;
3122                                         ifp->flags |= IFA_F_DEPRECATED;
3123                                 }
3124
3125                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3126                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3127
3128                                 spin_unlock(&ifp->lock);
3129
3130                                 if (deprecate) {
3131                                         in6_ifa_hold(ifp);
3132                                         read_unlock(&addrconf_hash_lock);
3133
3134                                         ipv6_ifa_notify(0, ifp);
3135                                         in6_ifa_put(ifp);
3136                                         goto restart;
3137                                 }
3138 #ifdef CONFIG_IPV6_PRIVACY
3139                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3140                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3141                                 if (age >= ifp->prefered_lft - regen_advance) {
3142                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3143                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3144                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3145                                         if (!ifp->regen_count && ifpub) {
3146                                                 ifp->regen_count++;
3147                                                 in6_ifa_hold(ifp);
3148                                                 in6_ifa_hold(ifpub);
3149                                                 spin_unlock(&ifp->lock);
3150                                                 read_unlock(&addrconf_hash_lock);
3151                                                 spin_lock(&ifpub->lock);
3152                                                 ifpub->regen_count = 0;
3153                                                 spin_unlock(&ifpub->lock);
3154                                                 ipv6_create_tempaddr(ifpub, ifp);
3155                                                 in6_ifa_put(ifpub);
3156                                                 in6_ifa_put(ifp);
3157                                                 goto restart;
3158                                         }
3159                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3160                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3161                                 spin_unlock(&ifp->lock);
3162 #endif
3163                         } else {
3164                                 /* ifp->prefered_lft <= ifp->valid_lft */
3165                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3166                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3167                                 spin_unlock(&ifp->lock);
3168                         }
3169                 }
3170                 read_unlock(&addrconf_hash_lock);
3171         }
3172
3173         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3174         add_timer(&addr_chk_timer);
3175         spin_unlock_bh(&addrconf_verify_lock);
3176 }
3177
3178 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3179 {
3180         struct in6_addr *pfx = NULL;
3181
3182         if (addr)
3183                 pfx = nla_data(addr);
3184
3185         if (local) {
3186                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3187                         pfx = NULL;
3188                 else
3189                         pfx = nla_data(local);
3190         }
3191
3192         return pfx;
3193 }
3194
3195 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3196         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3197         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3198         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3199 };
3200
3201 static int
3202 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3203 {
3204         struct net *net = sock_net(skb->sk);
3205         struct ifaddrmsg *ifm;
3206         struct nlattr *tb[IFA_MAX+1];
3207         struct in6_addr *pfx;
3208         int err;
3209
3210         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3211         if (err < 0)
3212                 return err;
3213
3214         ifm = nlmsg_data(nlh);
3215         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3216         if (pfx == NULL)
3217                 return -EINVAL;
3218
3219         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3220 }
3221
3222 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3223                              u32 prefered_lft, u32 valid_lft)
3224 {
3225         u32 flags;
3226         clock_t expires;
3227         unsigned long timeout;
3228
3229         if (!valid_lft || (prefered_lft > valid_lft))
3230                 return -EINVAL;
3231
3232         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3233         if (addrconf_finite_timeout(timeout)) {
3234                 expires = jiffies_to_clock_t(timeout * HZ);
3235                 valid_lft = timeout;
3236                 flags = RTF_EXPIRES;
3237         } else {
3238                 expires = 0;
3239                 flags = 0;
3240                 ifa_flags |= IFA_F_PERMANENT;
3241         }
3242
3243         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3244         if (addrconf_finite_timeout(timeout)) {
3245                 if (timeout == 0)
3246                         ifa_flags |= IFA_F_DEPRECATED;
3247                 prefered_lft = timeout;
3248         }
3249
3250         spin_lock_bh(&ifp->lock);
3251         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3252         ifp->tstamp = jiffies;
3253         ifp->valid_lft = valid_lft;
3254         ifp->prefered_lft = prefered_lft;
3255
3256         spin_unlock_bh(&ifp->lock);
3257         if (!(ifp->flags&IFA_F_TENTATIVE))
3258                 ipv6_ifa_notify(0, ifp);
3259
3260         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3261                               expires, flags);
3262         addrconf_verify(0);
3263
3264         return 0;
3265 }
3266
3267 static int
3268 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3269 {
3270         struct net *net = sock_net(skb->sk);
3271         struct ifaddrmsg *ifm;
3272         struct nlattr *tb[IFA_MAX+1];
3273         struct in6_addr *pfx;
3274         struct inet6_ifaddr *ifa;
3275         struct net_device *dev;
3276         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3277         u8 ifa_flags;
3278         int err;
3279
3280         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3281         if (err < 0)
3282                 return err;
3283
3284         ifm = nlmsg_data(nlh);
3285         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3286         if (pfx == NULL)
3287                 return -EINVAL;
3288
3289         if (tb[IFA_CACHEINFO]) {
3290                 struct ifa_cacheinfo *ci;
3291
3292                 ci = nla_data(tb[IFA_CACHEINFO]);
3293                 valid_lft = ci->ifa_valid;
3294                 preferred_lft = ci->ifa_prefered;
3295         } else {
3296                 preferred_lft = INFINITY_LIFE_TIME;
3297                 valid_lft = INFINITY_LIFE_TIME;
3298         }
3299
3300         dev =  __dev_get_by_index(net, ifm->ifa_index);
3301         if (dev == NULL)
3302                 return -ENODEV;
3303
3304         /* We ignore other flags so far. */
3305         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3306
3307         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3308         if (ifa == NULL) {
3309                 /*
3310                  * It would be best to check for !NLM_F_CREATE here but
3311                  * userspace alreay relies on not having to provide this.
3312                  */
3313                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3314                                       ifm->ifa_prefixlen, ifa_flags,
3315                                       preferred_lft, valid_lft);
3316         }
3317
3318         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3319             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3320                 err = -EEXIST;
3321         else
3322                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3323
3324         in6_ifa_put(ifa);
3325
3326         return err;
3327 }
3328
3329 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3330                           u8 scope, int ifindex)
3331 {
3332         struct ifaddrmsg *ifm;
3333
3334         ifm = nlmsg_data(nlh);
3335         ifm->ifa_family = AF_INET6;
3336         ifm->ifa_prefixlen = prefixlen;
3337         ifm->ifa_flags = flags;
3338         ifm->ifa_scope = scope;
3339         ifm->ifa_index = ifindex;
3340 }
3341
3342 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3343                          unsigned long tstamp, u32 preferred, u32 valid)
3344 {
3345         struct ifa_cacheinfo ci;
3346
3347         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3348                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3349         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3350                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3351         ci.ifa_prefered = preferred;
3352         ci.ifa_valid = valid;
3353
3354         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3355 }
3356
3357 static inline int rt_scope(int ifa_scope)
3358 {
3359         if (ifa_scope & IFA_HOST)
3360                 return RT_SCOPE_HOST;
3361         else if (ifa_scope & IFA_LINK)
3362                 return RT_SCOPE_LINK;
3363         else if (ifa_scope & IFA_SITE)
3364                 return RT_SCOPE_SITE;
3365         else
3366                 return RT_SCOPE_UNIVERSE;
3367 }
3368
3369 static inline int inet6_ifaddr_msgsize(void)
3370 {
3371         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3372                + nla_total_size(16) /* IFA_ADDRESS */
3373                + nla_total_size(sizeof(struct ifa_cacheinfo));
3374 }
3375
3376 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3377                              u32 pid, u32 seq, int event, unsigned int flags)
3378 {
3379         struct nlmsghdr  *nlh;
3380         u32 preferred, valid;
3381
3382         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3383         if (nlh == NULL)
3384                 return -EMSGSIZE;
3385
3386         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3387                       ifa->idev->dev->ifindex);
3388
3389         if (!(ifa->flags&IFA_F_PERMANENT)) {
3390                 preferred = ifa->prefered_lft;
3391                 valid = ifa->valid_lft;
3392                 if (preferred != INFINITY_LIFE_TIME) {
3393                         long tval = (jiffies - ifa->tstamp)/HZ;
3394                         preferred -= tval;
3395                         if (valid != INFINITY_LIFE_TIME)
3396                                 valid -= tval;
3397                 }
3398         } else {
3399                 preferred = INFINITY_LIFE_TIME;
3400                 valid = INFINITY_LIFE_TIME;
3401         }
3402
3403         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3404             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3405                 nlmsg_cancel(skb, nlh);
3406                 return -EMSGSIZE;
3407         }
3408
3409         return nlmsg_end(skb, nlh);
3410 }
3411
3412 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3413                                 u32 pid, u32 seq, int event, u16 flags)
3414 {
3415         struct nlmsghdr  *nlh;
3416         u8 scope = RT_SCOPE_UNIVERSE;
3417         int ifindex = ifmca->idev->dev->ifindex;
3418
3419         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3420                 scope = RT_SCOPE_SITE;
3421
3422         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3423         if (nlh == NULL)
3424                 return -EMSGSIZE;
3425
3426         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3427         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3428             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3429                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3430                 nlmsg_cancel(skb, nlh);
3431                 return -EMSGSIZE;
3432         }
3433
3434         return nlmsg_end(skb, nlh);
3435 }
3436
3437 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3438                                 u32 pid, u32 seq, int event, unsigned int flags)
3439 {
3440         struct nlmsghdr  *nlh;
3441         u8 scope = RT_SCOPE_UNIVERSE;
3442         int ifindex = ifaca->aca_idev->dev->ifindex;
3443
3444         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3445                 scope = RT_SCOPE_SITE;
3446
3447         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3448         if (nlh == NULL)
3449                 return -EMSGSIZE;
3450
3451         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3452         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3453             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3454                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3455                 nlmsg_cancel(skb, nlh);
3456                 return -EMSGSIZE;
3457         }
3458
3459         return nlmsg_end(skb, nlh);
3460 }
3461
3462 enum addr_type_t
3463 {
3464         UNICAST_ADDR,
3465         MULTICAST_ADDR,
3466         ANYCAST_ADDR,
3467 };
3468
3469 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3470                            enum addr_type_t type)
3471 {
3472         int idx, ip_idx;
3473         int s_idx, s_ip_idx;
3474         int err = 1;
3475         struct net_device *dev;
3476         struct inet6_dev *idev = NULL;
3477         struct inet6_ifaddr *ifa;
3478         struct ifmcaddr6 *ifmca;
3479         struct ifacaddr6 *ifaca;
3480         struct net *net = sock_net(skb->sk);
3481
3482         s_idx = cb->args[0];
3483         s_ip_idx = ip_idx = cb->args[1];
3484
3485         idx = 0;
3486         for_each_netdev(net, dev) {
3487                 if (idx < s_idx)
3488                         goto cont;
3489                 if (idx > s_idx)
3490                         s_ip_idx = 0;
3491                 ip_idx = 0;
3492                 if ((idev = in6_dev_get(dev)) == NULL)
3493                         goto cont;
3494                 read_lock_bh(&idev->lock);
3495                 switch (type) {
3496                 case UNICAST_ADDR:
3497                         /* unicast address incl. temp addr */
3498                         for (ifa = idev->addr_list; ifa;
3499                              ifa = ifa->if_next, ip_idx++) {
3500                                 if (ip_idx < s_ip_idx)
3501                                         continue;
3502                                 err = inet6_fill_ifaddr(skb, ifa,
3503                                                         NETLINK_CB(cb->skb).pid,
3504                                                         cb->nlh->nlmsg_seq,
3505                                                         RTM_NEWADDR,
3506                                                         NLM_F_MULTI);
3507                         }
3508                         break;
3509                 case MULTICAST_ADDR:
3510                         /* multicast address */
3511                         for (ifmca = idev->mc_list; ifmca;
3512                              ifmca = ifmca->next, ip_idx++) {
3513                                 if (ip_idx < s_ip_idx)
3514                                         continue;
3515                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3516                                                           NETLINK_CB(cb->skb).pid,
3517                                                           cb->nlh->nlmsg_seq,
3518                                                           RTM_GETMULTICAST,
3519                                                           NLM_F_MULTI);
3520                         }
3521                         break;
3522                 case ANYCAST_ADDR:
3523                         /* anycast address */
3524                         for (ifaca = idev->ac_list; ifaca;
3525                              ifaca = ifaca->aca_next, ip_idx++) {
3526                                 if (ip_idx < s_ip_idx)
3527                                         continue;
3528                                 err = inet6_fill_ifacaddr(skb, ifaca,
3529                                                           NETLINK_CB(cb->skb).pid,
3530                                                           cb->nlh->nlmsg_seq,
3531                                                           RTM_GETANYCAST,
3532                                                           NLM_F_MULTI);
3533                         }
3534                         break;
3535                 default:
3536                         break;
3537                 }
3538                 read_unlock_bh(&idev->lock);
3539                 in6_dev_put(idev);
3540
3541                 if (err <= 0)
3542                         break;
3543 cont:
3544                 idx++;
3545         }
3546         cb->args[0] = idx;
3547         cb->args[1] = ip_idx;
3548         return skb->len;
3549 }
3550
3551 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3552 {
3553         enum addr_type_t type = UNICAST_ADDR;
3554
3555         return inet6_dump_addr(skb, cb, type);
3556 }
3557
3558 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3559 {
3560         enum addr_type_t type = MULTICAST_ADDR;
3561
3562         return inet6_dump_addr(skb, cb, type);
3563 }
3564
3565
3566 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3567 {
3568         enum addr_type_t type = ANYCAST_ADDR;
3569
3570         return inet6_dump_addr(skb, cb, type);
3571 }
3572
3573 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3574                              void *arg)
3575 {
3576         struct net *net = sock_net(in_skb->sk);
3577         struct ifaddrmsg *ifm;
3578         struct nlattr *tb[IFA_MAX+1];
3579         struct in6_addr *addr = NULL;
3580         struct net_device *dev = NULL;
3581         struct inet6_ifaddr *ifa;
3582         struct sk_buff *skb;
3583         int err;
3584
3585         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3586         if (err < 0)
3587                 goto errout;
3588
3589         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3590         if (addr == NULL) {
3591                 err = -EINVAL;
3592                 goto errout;
3593         }
3594
3595         ifm = nlmsg_data(nlh);
3596         if (ifm->ifa_index)
3597                 dev = __dev_get_by_index(net, ifm->ifa_index);
3598
3599         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3600                 err = -EADDRNOTAVAIL;
3601                 goto errout;
3602         }
3603
3604         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3605                 err = -ENOBUFS;
3606                 goto errout_ifa;
3607         }
3608
3609         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3610                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3611         if (err < 0) {
3612                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3613                 WARN_ON(err == -EMSGSIZE);
3614                 kfree_skb(skb);
3615                 goto errout_ifa;
3616         }
3617         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3618 errout_ifa:
3619         in6_ifa_put(ifa);
3620 errout:
3621         return err;
3622 }
3623
3624 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3625 {
3626         struct sk_buff *skb;
3627         struct net *net = dev_net(ifa->idev->dev);
3628         int err = -ENOBUFS;
3629
3630         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3631         if (skb == NULL)
3632                 goto errout;
3633
3634         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3635         if (err < 0) {
3636                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3637                 WARN_ON(err == -EMSGSIZE);
3638                 kfree_skb(skb);
3639                 goto errout;
3640         }
3641         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3642         return;
3643 errout:
3644         if (err < 0)
3645                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3646 }
3647
3648 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3649                                 __s32 *array, int bytes)
3650 {
3651         BUG_ON(bytes < (DEVCONF_MAX * 4));
3652
3653         memset(array, 0, bytes);
3654         array[DEVCONF_FORWARDING] = cnf->forwarding;
3655         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3656         array[DEVCONF_MTU6] = cnf->mtu6;
3657         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3658         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3659         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3660         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3661         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3662         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3663         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3664         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3665 #ifdef CONFIG_IPV6_PRIVACY
3666         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3667         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3668         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3669         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3670         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3671 #endif
3672         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3673         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3674         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3675 #ifdef CONFIG_IPV6_ROUTER_PREF
3676         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3677         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3678 #ifdef CONFIG_IPV6_ROUTE_INFO
3679         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3680 #endif
3681 #endif
3682         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3683         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3684 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3685         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3686 #endif
3687 #ifdef CONFIG_IPV6_MROUTE
3688         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3689 #endif
3690         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3691         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3692 }
3693
3694 static inline size_t inet6_if_nlmsg_size(void)
3695 {
3696         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3697                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3698                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3699                + nla_total_size(4) /* IFLA_MTU */
3700                + nla_total_size(4) /* IFLA_LINK */
3701                + nla_total_size( /* IFLA_PROTINFO */
3702                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3703                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3704                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3705                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3706                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3707                  );
3708 }
3709
3710 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3711                                       int bytes)
3712 {
3713         int i;
3714         int pad = bytes - sizeof(u64) * items;
3715         BUG_ON(pad < 0);
3716
3717         /* Use put_unaligned() because stats may not be aligned for u64. */
3718         put_unaligned(items, &stats[0]);
3719         for (i = 1; i < items; i++)
3720                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3721
3722         memset(&stats[items], 0, pad);
3723 }
3724
3725 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3726                              int bytes)
3727 {
3728         switch(attrtype) {
3729         case IFLA_INET6_STATS:
3730                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3731                 break;
3732         case IFLA_INET6_ICMP6STATS:
3733                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3734                 break;
3735         }
3736 }
3737
3738 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3739                              u32 pid, u32 seq, int event, unsigned int flags)
3740 {
3741         struct net_device *dev = idev->dev;
3742         struct nlattr *nla;
3743         struct ifinfomsg *hdr;
3744         struct nlmsghdr *nlh;
3745         void *protoinfo;
3746         struct ifla_cacheinfo ci;
3747
3748         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3749         if (nlh == NULL)
3750                 return -EMSGSIZE;
3751
3752         hdr = nlmsg_data(nlh);
3753         hdr->ifi_family = AF_INET6;
3754         hdr->__ifi_pad = 0;
3755         hdr->ifi_type = dev->type;
3756         hdr->ifi_index = dev->ifindex;
3757         hdr->ifi_flags = dev_get_flags(dev);
3758         hdr->ifi_change = 0;
3759
3760         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3761
3762         if (dev->addr_len)
3763                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3764
3765         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3766         if (dev->ifindex != dev->iflink)
3767                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3768
3769         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3770         if (protoinfo == NULL)
3771                 goto nla_put_failure;
3772
3773         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3774
3775         ci.max_reasm_len = IPV6_MAXPLEN;
3776         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3777                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3778         ci.reachable_time = idev->nd_parms->reachable_time;
3779         ci.retrans_time = idev->nd_parms->retrans_time;
3780         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3781
3782         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3783         if (nla == NULL)
3784                 goto nla_put_failure;
3785         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3786
3787         /* XXX - MC not implemented */
3788
3789         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3790         if (nla == NULL)
3791                 goto nla_put_failure;
3792         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3793
3794         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3795         if (nla == NULL)
3796                 goto nla_put_failure;
3797         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3798
3799         nla_nest_end(skb, protoinfo);
3800         return nlmsg_end(skb, nlh);
3801
3802 nla_put_failure:
3803         nlmsg_cancel(skb, nlh);
3804         return -EMSGSIZE;
3805 }
3806
3807 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3808 {
3809         struct net *net = sock_net(skb->sk);
3810         int idx, err;
3811         int s_idx = cb->args[0];
3812         struct net_device *dev;
3813         struct inet6_dev *idev;
3814
3815         read_lock(&dev_base_lock);
3816         idx = 0;
3817         for_each_netdev(net, dev) {
3818                 if (idx < s_idx)
3819                         goto cont;
3820                 if ((idev = in6_dev_get(dev)) == NULL)
3821                         goto cont;
3822                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3823                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3824                 in6_dev_put(idev);
3825                 if (err <= 0)
3826                         break;
3827 cont:
3828                 idx++;
3829         }
3830         read_unlock(&dev_base_lock);
3831         cb->args[0] = idx;
3832
3833         return skb->len;
3834 }
3835
3836 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3837 {
3838         struct sk_buff *skb;
3839         struct net *net = dev_net(idev->dev);
3840         int err = -ENOBUFS;
3841
3842         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3843         if (skb == NULL)
3844                 goto errout;
3845
3846         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3847         if (err < 0) {
3848                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3849                 WARN_ON(err == -EMSGSIZE);
3850                 kfree_skb(skb);
3851                 goto errout;
3852         }
3853         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3854         return;
3855 errout:
3856         if (err < 0)
3857                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3858 }
3859
3860 static inline size_t inet6_prefix_nlmsg_size(void)
3861 {
3862         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3863                + nla_total_size(sizeof(struct in6_addr))
3864                + nla_total_size(sizeof(struct prefix_cacheinfo));
3865 }
3866
3867 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3868                              struct prefix_info *pinfo, u32 pid, u32 seq,
3869                              int event, unsigned int flags)
3870 {
3871         struct prefixmsg *pmsg;
3872         struct nlmsghdr *nlh;
3873         struct prefix_cacheinfo ci;
3874
3875         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3876         if (nlh == NULL)
3877                 return -EMSGSIZE;
3878
3879         pmsg = nlmsg_data(nlh);
3880         pmsg->prefix_family = AF_INET6;
3881         pmsg->prefix_pad1 = 0;
3882         pmsg->prefix_pad2 = 0;
3883         pmsg->prefix_ifindex = idev->dev->ifindex;
3884         pmsg->prefix_len = pinfo->prefix_len;
3885         pmsg->prefix_type = pinfo->type;
3886         pmsg->prefix_pad3 = 0;
3887         pmsg->prefix_flags = 0;
3888         if (pinfo->onlink)
3889                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3890         if (pinfo->autoconf)
3891                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3892
3893         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3894
3895         ci.preferred_time = ntohl(pinfo->prefered);
3896         ci.valid_time = ntohl(pinfo->valid);
3897         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3898
3899         return nlmsg_end(skb, nlh);
3900
3901 nla_put_failure:
3902         nlmsg_cancel(skb, nlh);
3903         return -EMSGSIZE;
3904 }
3905
3906 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3907                          struct prefix_info *pinfo)
3908 {
3909         struct sk_buff *skb;
3910         struct net *net = dev_net(idev->dev);
3911         int err = -ENOBUFS;
3912
3913         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3914         if (skb == NULL)
3915                 goto errout;
3916
3917         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3918         if (err < 0) {
3919                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3920                 WARN_ON(err == -EMSGSIZE);
3921                 kfree_skb(skb);
3922                 goto errout;
3923         }
3924         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3925         return;
3926 errout:
3927         if (err < 0)
3928                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3929 }
3930
3931 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3932 {
3933         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3934
3935         switch (event) {
3936         case RTM_NEWADDR:
3937                 /*
3938                  * If the address was optimistic
3939                  * we inserted the route at the start of
3940                  * our DAD process, so we don't need
3941                  * to do it again
3942                  */
3943                 if (!(ifp->rt->rt6i_node))
3944                         ip6_ins_rt(ifp->rt);
3945                 if (ifp->idev->cnf.forwarding)
3946                         addrconf_join_anycast(ifp);
3947                 break;
3948         case RTM_DELADDR:
3949                 if (ifp->idev->cnf.forwarding)
3950                         addrconf_leave_anycast(ifp);
3951                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3952                 dst_hold(&ifp->rt->u.dst);
3953                 if (ip6_del_rt(ifp->rt))
3954                         dst_free(&ifp->rt->u.dst);
3955                 break;
3956         }
3957 }
3958
3959 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3960 {
3961         rcu_read_lock_bh();
3962         if (likely(ifp->idev->dead == 0))
3963                 __ipv6_ifa_notify(event, ifp);
3964         rcu_read_unlock_bh();
3965 }
3966
3967 #ifdef CONFIG_SYSCTL
3968
3969 static
3970 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3971                            void __user *buffer, size_t *lenp, loff_t *ppos)
3972 {
3973         int *valp = ctl->data;
3974         int val = *valp;
3975         int ret;
3976
3977         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3978
3979         if (write)
3980                 addrconf_fixup_forwarding(ctl, valp, val);
3981         return ret;
3982 }
3983
3984 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3985                                             void __user *oldval,
3986                                             size_t __user *oldlenp,
3987                                             void __user *newval, size_t newlen)
3988 {
3989         int *valp = table->data;
3990         int val = *valp;
3991         int new;
3992
3993         if (!newval || !newlen)
3994                 return 0;
3995         if (newlen != sizeof(int))
3996                 return -EINVAL;
3997         if (get_user(new, (int __user *)newval))
3998                 return -EFAULT;
3999         if (new == *valp)
4000                 return 0;
4001         if (oldval && oldlenp) {
4002                 size_t len;
4003                 if (get_user(len, oldlenp))
4004                         return -EFAULT;
4005                 if (len) {
4006                         if (len > table->maxlen)
4007                                 len = table->maxlen;
4008                         if (copy_to_user(oldval, valp, len))
4009                                 return -EFAULT;
4010                         if (put_user(len, oldlenp))
4011                                 return -EFAULT;
4012                 }
4013         }
4014
4015         *valp = new;
4016         addrconf_fixup_forwarding(table, valp, val);
4017         return 1;
4018 }
4019
4020 static struct addrconf_sysctl_table
4021 {
4022         struct ctl_table_header *sysctl_header;
4023         ctl_table addrconf_vars[DEVCONF_MAX+1];
4024         char *dev_name;
4025 } addrconf_sysctl __read_mostly = {
4026         .sysctl_header = NULL,
4027         .addrconf_vars = {
4028                 {
4029                         .ctl_name       =       NET_IPV6_FORWARDING,
4030                         .procname       =       "forwarding",
4031                         .data           =       &ipv6_devconf.forwarding,
4032                         .maxlen         =       sizeof(int),
4033                         .mode           =       0644,
4034                         .proc_handler   =       addrconf_sysctl_forward,
4035                         .strategy       =       addrconf_sysctl_forward_strategy,
4036                 },
4037                 {
4038                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4039                         .procname       =       "hop_limit",
4040                         .data           =       &ipv6_devconf.hop_limit,
4041                         .maxlen         =       sizeof(int),
4042                         .mode           =       0644,
4043                         .proc_handler   =       proc_dointvec,
4044                 },
4045                 {
4046                         .ctl_name       =       NET_IPV6_MTU,
4047                         .procname       =       "mtu",
4048                         .data           =       &ipv6_devconf.mtu6,
4049                         .maxlen         =       sizeof(int),
4050                         .mode           =       0644,
4051                         .proc_handler   =       proc_dointvec,
4052                 },
4053                 {
4054                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4055                         .procname       =       "accept_ra",
4056                         .data           =       &ipv6_devconf.accept_ra,
4057                         .maxlen         =       sizeof(int),
4058                         .mode           =       0644,
4059                         .proc_handler   =       proc_dointvec,
4060                 },
4061                 {
4062                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4063                         .procname       =       "accept_redirects",
4064                         .data           =       &ipv6_devconf.accept_redirects,
4065                         .maxlen         =       sizeof(int),
4066                         .mode           =       0644,
4067                         .proc_handler   =       proc_dointvec,
4068                 },
4069                 {
4070                         .ctl_name       =       NET_IPV6_AUTOCONF,
4071                         .procname       =       "autoconf",
4072                         .data           =       &ipv6_devconf.autoconf,
4073                         .maxlen         =       sizeof(int),
4074                         .mode           =       0644,
4075                         .proc_handler   =       proc_dointvec,
4076                 },
4077                 {
4078                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4079                         .procname       =       "dad_transmits",
4080                         .data           =       &ipv6_devconf.dad_transmits,
4081                         .maxlen         =       sizeof(int),
4082                         .mode           =       0644,
4083                         .proc_handler   =       proc_dointvec,
4084                 },
4085                 {
4086                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4087                         .procname       =       "router_solicitations",
4088                         .data           =       &ipv6_devconf.rtr_solicits,
4089                         .maxlen         =       sizeof(int),
4090                         .mode           =       0644,
4091                         .proc_handler   =       proc_dointvec,
4092                 },
4093                 {
4094                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4095                         .procname       =       "router_solicitation_interval",
4096                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4097                         .maxlen         =       sizeof(int),
4098                         .mode           =       0644,
4099                         .proc_handler   =       proc_dointvec_jiffies,
4100                         .strategy       =       sysctl_jiffies,
4101                 },
4102                 {
4103                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4104                         .procname       =       "router_solicitation_delay",
4105                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4106                         .maxlen         =       sizeof(int),
4107                         .mode           =       0644,
4108                         .proc_handler   =       proc_dointvec_jiffies,
4109                         .strategy       =       sysctl_jiffies,
4110                 },
4111                 {
4112                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4113                         .procname       =       "force_mld_version",
4114                         .data           =       &ipv6_devconf.force_mld_version,
4115                         .maxlen         =       sizeof(int),
4116                         .mode           =       0644,
4117                         .proc_handler   =       proc_dointvec,
4118                 },
4119 #ifdef CONFIG_IPV6_PRIVACY
4120                 {
4121                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4122                         .procname       =       "use_tempaddr",
4123                         .data           =       &ipv6_devconf.use_tempaddr,
4124                         .maxlen         =       sizeof(int),
4125                         .mode           =       0644,
4126                         .proc_handler   =       proc_dointvec,
4127                 },
4128                 {
4129                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4130                         .procname       =       "temp_valid_lft",
4131                         .data           =       &ipv6_devconf.temp_valid_lft,
4132                         .maxlen         =       sizeof(int),
4133                         .mode           =       0644,
4134                         .proc_handler   =       proc_dointvec,
4135                 },
4136                 {
4137                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4138                         .procname       =       "temp_prefered_lft",
4139                         .data           =       &ipv6_devconf.temp_prefered_lft,
4140                         .maxlen         =       sizeof(int),
4141                         .mode           =       0644,
4142                         .proc_handler   =       proc_dointvec,
4143                 },
4144                 {
4145                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4146                         .procname       =       "regen_max_retry",
4147                         .data           =       &ipv6_devconf.regen_max_retry,
4148                         .maxlen         =       sizeof(int),
4149                         .mode           =       0644,
4150                         .proc_handler   =       proc_dointvec,
4151                 },
4152                 {
4153                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4154                         .procname       =       "max_desync_factor",
4155                         .data           =       &ipv6_devconf.max_desync_factor,
4156                         .maxlen         =       sizeof(int),
4157                         .mode           =       0644,
4158                         .proc_handler   =       proc_dointvec,
4159                 },
4160 #endif
4161                 {
4162                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4163                         .procname       =       "max_addresses",
4164                         .data           =       &ipv6_devconf.max_addresses,
4165                         .maxlen         =       sizeof(int),
4166                         .mode           =       0644,
4167                         .proc_handler   =       proc_dointvec,
4168                 },
4169                 {
4170                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4171                         .procname       =       "accept_ra_defrtr",
4172                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4173                         .maxlen         =       sizeof(int),
4174                         .mode           =       0644,
4175                         .proc_handler   =       proc_dointvec,
4176                 },
4177                 {
4178                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4179                         .procname       =       "accept_ra_pinfo",
4180                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4181                         .maxlen         =       sizeof(int),
4182                         .mode           =       0644,
4183                         .proc_handler   =       proc_dointvec,
4184                 },
4185 #ifdef CONFIG_IPV6_ROUTER_PREF
4186                 {
4187                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4188                         .procname       =       "accept_ra_rtr_pref",
4189                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4190                         .maxlen         =       sizeof(int),
4191                         .mode           =       0644,
4192                         .proc_handler   =       proc_dointvec,
4193                 },
4194                 {
4195                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4196                         .procname       =       "router_probe_interval",
4197                         .data           =       &ipv6_devconf.rtr_probe_interval,
4198                         .maxlen         =       sizeof(int),
4199                         .mode           =       0644,
4200                         .proc_handler   =       proc_dointvec_jiffies,
4201                         .strategy       =       sysctl_jiffies,
4202                 },
4203 #ifdef CONFIG_IPV6_ROUTE_INFO
4204                 {
4205                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4206                         .procname       =       "accept_ra_rt_info_max_plen",
4207                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4208                         .maxlen         =       sizeof(int),
4209                         .mode           =       0644,
4210                         .proc_handler   =       proc_dointvec,
4211                 },
4212 #endif
4213 #endif
4214                 {
4215                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4216                         .procname       =       "proxy_ndp",
4217                         .data           =       &ipv6_devconf.proxy_ndp,
4218                         .maxlen         =       sizeof(int),
4219                         .mode           =       0644,
4220                         .proc_handler   =       proc_dointvec,
4221                 },
4222                 {
4223                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4224                         .procname       =       "accept_source_route",
4225                         .data           =       &ipv6_devconf.accept_source_route,
4226                         .maxlen         =       sizeof(int),
4227                         .mode           =       0644,
4228                         .proc_handler   =       proc_dointvec,
4229                 },
4230 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4231                 {
4232                         .ctl_name       =       CTL_UNNUMBERED,
4233                         .procname       =       "optimistic_dad",
4234                         .data           =       &ipv6_devconf.optimistic_dad,
4235                         .maxlen         =       sizeof(int),
4236                         .mode           =       0644,
4237                         .proc_handler   =       proc_dointvec,
4238
4239                 },
4240 #endif
4241 #ifdef CONFIG_IPV6_MROUTE
4242                 {
4243                         .ctl_name       =       CTL_UNNUMBERED,
4244                         .procname       =       "mc_forwarding",
4245                         .data           =       &ipv6_devconf.mc_forwarding,
4246                         .maxlen         =       sizeof(int),
4247                         .mode           =       0444,
4248                         .proc_handler   =       proc_dointvec,
4249                 },
4250 #endif
4251                 {
4252                         .ctl_name       =       CTL_UNNUMBERED,
4253                         .procname       =       "disable_ipv6",
4254                         .data           =       &ipv6_devconf.disable_ipv6,
4255                         .maxlen         =       sizeof(int),
4256                         .mode           =       0644,
4257                         .proc_handler   =       proc_dointvec,
4258                 },
4259                 {
4260                         .ctl_name       =       CTL_UNNUMBERED,
4261                         .procname       =       "accept_dad",
4262                         .data           =       &ipv6_devconf.accept_dad,
4263                         .maxlen         =       sizeof(int),
4264                         .mode           =       0644,
4265                         .proc_handler   =       proc_dointvec,
4266                 },
4267                 {
4268                         .ctl_name       =       0,      /* sentinel */
4269                 }
4270         },
4271 };
4272
4273 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4274                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4275 {
4276         int i;
4277         struct addrconf_sysctl_table *t;
4278
4279 #define ADDRCONF_CTL_PATH_DEV   3
4280
4281         struct ctl_path addrconf_ctl_path[] = {
4282                 { .procname = "net", .ctl_name = CTL_NET, },
4283                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4284                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4285                 { /* to be set */ },
4286                 { },
4287         };
4288
4289
4290         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4291         if (t == NULL)
4292                 goto out;
4293
4294         for (i=0; t->addrconf_vars[i].data; i++) {
4295                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4296                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4297                 t->addrconf_vars[i].extra2 = net;
4298         }
4299
4300         /*
4301          * Make a copy of dev_name, because '.procname' is regarded as const
4302          * by sysctl and we wouldn't want anyone to change it under our feet
4303          * (see SIOCSIFNAME).
4304          */
4305         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4306         if (!t->dev_name)
4307                 goto free;
4308
4309         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4310         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4311
4312         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4313                         t->addrconf_vars);
4314         if (t->sysctl_header == NULL)
4315                 goto free_procname;
4316
4317         p->sysctl = t;
4318         return 0;
4319
4320 free_procname:
4321         kfree(t->dev_name);
4322 free:
4323         kfree(t);
4324 out:
4325         return -ENOBUFS;
4326 }
4327
4328 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4329 {
4330         struct addrconf_sysctl_table *t;
4331
4332         if (p->sysctl == NULL)
4333                 return;
4334
4335         t = p->sysctl;
4336         p->sysctl = NULL;
4337         unregister_sysctl_table(t->sysctl_header);
4338         kfree(t->dev_name);
4339         kfree(t);
4340 }
4341
4342 static void addrconf_sysctl_register(struct inet6_dev *idev)
4343 {
4344         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4345                               NET_IPV6_NEIGH, "ipv6",
4346                               &ndisc_ifinfo_sysctl_change,
4347                               ndisc_ifinfo_sysctl_strategy);
4348         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4349                         idev->dev->ifindex, idev, &idev->cnf);
4350 }
4351
4352 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4353 {
4354         __addrconf_sysctl_unregister(&idev->cnf);
4355         neigh_sysctl_unregister(idev->nd_parms);
4356 }
4357
4358
4359 #endif
4360
4361 static int addrconf_init_net(struct net *net)
4362 {
4363         int err;
4364         struct ipv6_devconf *all, *dflt;
4365
4366         err = -ENOMEM;
4367         all = &ipv6_devconf;
4368         dflt = &ipv6_devconf_dflt;
4369
4370         if (net != &init_net) {
4371                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4372                 if (all == NULL)
4373                         goto err_alloc_all;
4374
4375                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4376                 if (dflt == NULL)
4377                         goto err_alloc_dflt;
4378         }
4379
4380         net->ipv6.devconf_all = all;
4381         net->ipv6.devconf_dflt = dflt;
4382
4383 #ifdef CONFIG_SYSCTL
4384         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4385                         NULL, all);
4386         if (err < 0)
4387                 goto err_reg_all;
4388
4389         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4390                         NULL, dflt);
4391         if (err < 0)
4392                 goto err_reg_dflt;
4393 #endif
4394         return 0;
4395
4396 #ifdef CONFIG_SYSCTL
4397 err_reg_dflt:
4398         __addrconf_sysctl_unregister(all);
4399 err_reg_all:
4400         kfree(dflt);
4401 #endif
4402 err_alloc_dflt:
4403         kfree(all);
4404 err_alloc_all:
4405         return err;
4406 }
4407
4408 static void addrconf_exit_net(struct net *net)
4409 {
4410 #ifdef CONFIG_SYSCTL
4411         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4412         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4413 #endif
4414         if (net != &init_net) {
4415                 kfree(net->ipv6.devconf_dflt);
4416                 kfree(net->ipv6.devconf_all);
4417         }
4418 }
4419
4420 static struct pernet_operations addrconf_ops = {
4421         .init = addrconf_init_net,
4422         .exit = addrconf_exit_net,
4423 };
4424
4425 /*
4426  *      Device notifier
4427  */
4428
4429 int register_inet6addr_notifier(struct notifier_block *nb)
4430 {
4431         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4432 }
4433
4434 EXPORT_SYMBOL(register_inet6addr_notifier);
4435
4436 int unregister_inet6addr_notifier(struct notifier_block *nb)
4437 {
4438         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4439 }
4440
4441 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4442
4443 static void addrconf_net_exit(struct net *net)
4444 {
4445         struct net_device *dev;
4446
4447         rtnl_lock();
4448         /* clean dev list */
4449         for_each_netdev(net, dev) {
4450                 if (__in6_dev_get(dev) == NULL)
4451                         continue;
4452                 addrconf_ifdown(dev, 1);
4453         }
4454         addrconf_ifdown(net->loopback_dev, 2);
4455         rtnl_unlock();
4456 }
4457
4458 static struct pernet_operations addrconf_net_ops = {
4459         .exit = addrconf_net_exit,
4460 };
4461
4462 /*
4463  *      Init / cleanup code
4464  */
4465
4466 int __init addrconf_init(void)
4467 {
4468         int err;
4469
4470         if ((err = ipv6_addr_label_init()) < 0) {
4471                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4472                         err);
4473                 return err;
4474         }
4475
4476         register_pernet_subsys(&addrconf_ops);
4477
4478         /* The addrconf netdev notifier requires that loopback_dev
4479          * has it's ipv6 private information allocated and setup
4480          * before it can bring up and give link-local addresses
4481          * to other devices which are up.
4482          *
4483          * Unfortunately, loopback_dev is not necessarily the first
4484          * entry in the global dev_base list of net devices.  In fact,
4485          * it is likely to be the very last entry on that list.
4486          * So this causes the notifier registry below to try and
4487          * give link-local addresses to all devices besides loopback_dev
4488          * first, then loopback_dev, which cases all the non-loopback_dev
4489          * devices to fail to get a link-local address.
4490          *
4491          * So, as a temporary fix, allocate the ipv6 structure for
4492          * loopback_dev first by hand.
4493          * Longer term, all of the dependencies ipv6 has upon the loopback
4494          * device and it being up should be removed.
4495          */
4496         rtnl_lock();
4497         if (!ipv6_add_dev(init_net.loopback_dev))
4498                 err = -ENOMEM;
4499         rtnl_unlock();
4500         if (err)
4501                 goto errlo;
4502
4503         err = register_pernet_device(&addrconf_net_ops);
4504         if (err)
4505                 return err;
4506
4507         register_netdevice_notifier(&ipv6_dev_notf);
4508
4509         addrconf_verify(0);
4510
4511         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4512         if (err < 0)
4513                 goto errout;
4514
4515         /* Only the first call to __rtnl_register can fail */
4516         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4517         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4518         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4519         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4520         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4521
4522         ipv6_addr_label_rtnl_register();
4523
4524         return 0;
4525 errout:
4526         unregister_netdevice_notifier(&ipv6_dev_notf);
4527 errlo:
4528         unregister_pernet_subsys(&addrconf_ops);
4529
4530         return err;
4531 }
4532
4533 void addrconf_cleanup(void)
4534 {
4535         struct inet6_ifaddr *ifa;
4536         int i;
4537
4538         unregister_netdevice_notifier(&ipv6_dev_notf);
4539         unregister_pernet_device(&addrconf_net_ops);
4540
4541         unregister_pernet_subsys(&addrconf_ops);
4542
4543         rtnl_lock();
4544
4545         /*
4546          *      Check hash table.
4547          */
4548         write_lock_bh(&addrconf_hash_lock);
4549         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4550                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4551                         struct inet6_ifaddr *bifa;
4552
4553                         bifa = ifa;
4554                         ifa = ifa->lst_next;
4555                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4556                         /* Do not free it; something is wrong.
4557                            Now we can investigate it with debugger.
4558                          */
4559                 }
4560         }
4561         write_unlock_bh(&addrconf_hash_lock);
4562
4563         del_timer(&addr_chk_timer);
4564         rtnl_unlock();
4565
4566         unregister_pernet_subsys(&addrconf_net_ops);
4567 }