5c84c798331d3e59b84455073ea50cb8c7458300
[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 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 int addrconf_qdisc_ok(struct net_device *dev)
233 {
234         struct netdev_queue *txq = &dev->tx_queue;
235         return (txq->qdisc != &noop_qdisc);
236 }
237
238 /* Check if a route is valid prefix route */
239 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
240 {
241         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
242 }
243
244 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
245 {
246         if (del_timer(&ifp->timer))
247                 __in6_ifa_put(ifp);
248 }
249
250 enum addrconf_timer_t
251 {
252         AC_NONE,
253         AC_DAD,
254         AC_RS,
255 };
256
257 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
258                                enum addrconf_timer_t what,
259                                unsigned long when)
260 {
261         if (!del_timer(&ifp->timer))
262                 in6_ifa_hold(ifp);
263
264         switch (what) {
265         case AC_DAD:
266                 ifp->timer.function = addrconf_dad_timer;
267                 break;
268         case AC_RS:
269                 ifp->timer.function = addrconf_rs_timer;
270                 break;
271         default:;
272         }
273         ifp->timer.expires = jiffies + when;
274         add_timer(&ifp->timer);
275 }
276
277 static int snmp6_alloc_dev(struct inet6_dev *idev)
278 {
279         if (snmp_mib_init((void **)idev->stats.ipv6,
280                           sizeof(struct ipstats_mib)) < 0)
281                 goto err_ip;
282         if (snmp_mib_init((void **)idev->stats.icmpv6,
283                           sizeof(struct icmpv6_mib)) < 0)
284                 goto err_icmp;
285         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
286                           sizeof(struct icmpv6msg_mib)) < 0)
287                 goto err_icmpmsg;
288
289         return 0;
290
291 err_icmpmsg:
292         snmp_mib_free((void **)idev->stats.icmpv6);
293 err_icmp:
294         snmp_mib_free((void **)idev->stats.ipv6);
295 err_ip:
296         return -ENOMEM;
297 }
298
299 static void snmp6_free_dev(struct inet6_dev *idev)
300 {
301         snmp_mib_free((void **)idev->stats.icmpv6msg);
302         snmp_mib_free((void **)idev->stats.icmpv6);
303         snmp_mib_free((void **)idev->stats.ipv6);
304 }
305
306 /* Nobody refers to this device, we may destroy it. */
307
308 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
309 {
310         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
311         kfree(idev);
312 }
313
314 void in6_dev_finish_destroy(struct inet6_dev *idev)
315 {
316         struct net_device *dev = idev->dev;
317         BUG_TRAP(idev->addr_list==NULL);
318         BUG_TRAP(idev->mc_list==NULL);
319 #ifdef NET_REFCNT_DEBUG
320         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
321 #endif
322         dev_put(dev);
323         if (!idev->dead) {
324                 printk("Freeing alive inet6 device %p\n", idev);
325                 return;
326         }
327         snmp6_free_dev(idev);
328         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
329 }
330
331 EXPORT_SYMBOL(in6_dev_finish_destroy);
332
333 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
334 {
335         struct inet6_dev *ndev;
336
337         ASSERT_RTNL();
338
339         if (dev->mtu < IPV6_MIN_MTU)
340                 return NULL;
341
342         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
343
344         if (ndev == NULL)
345                 return NULL;
346
347         rwlock_init(&ndev->lock);
348         ndev->dev = dev;
349         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
350         ndev->cnf.mtu6 = dev->mtu;
351         ndev->cnf.sysctl = NULL;
352         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
353         if (ndev->nd_parms == NULL) {
354                 kfree(ndev);
355                 return NULL;
356         }
357         if (ndev->cnf.forwarding)
358                 dev_disable_lro(dev);
359         /* We refer to the device */
360         dev_hold(dev);
361
362         if (snmp6_alloc_dev(ndev) < 0) {
363                 ADBG((KERN_WARNING
364                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
365                         __func__, dev->name));
366                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
367                 ndev->dead = 1;
368                 in6_dev_finish_destroy(ndev);
369                 return NULL;
370         }
371
372         if (snmp6_register_dev(ndev) < 0) {
373                 ADBG((KERN_WARNING
374                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
375                         __func__, dev->name));
376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
377                 ndev->dead = 1;
378                 in6_dev_finish_destroy(ndev);
379                 return NULL;
380         }
381
382         /* One reference from device.  We must do this before
383          * we invoke __ipv6_regen_rndid().
384          */
385         in6_dev_hold(ndev);
386
387         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
388                 ndev->cnf.accept_dad = -1;
389
390 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
391         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
392                 printk(KERN_INFO
393                        "%s: Disabled Multicast RS\n",
394                        dev->name);
395                 ndev->cnf.rtr_solicits = 0;
396         }
397 #endif
398
399 #ifdef CONFIG_IPV6_PRIVACY
400         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
401         if ((dev->flags&IFF_LOOPBACK) ||
402             dev->type == ARPHRD_TUNNEL ||
403             dev->type == ARPHRD_TUNNEL6 ||
404             dev->type == ARPHRD_SIT ||
405             dev->type == ARPHRD_NONE) {
406                 printk(KERN_INFO
407                        "%s: Disabled Privacy Extensions\n",
408                        dev->name);
409                 ndev->cnf.use_tempaddr = -1;
410         } else {
411                 in6_dev_hold(ndev);
412                 ipv6_regen_rndid((unsigned long) ndev);
413         }
414 #endif
415
416         if (netif_running(dev) && addrconf_qdisc_ok(dev))
417                 ndev->if_flags |= IF_READY;
418
419         ipv6_mc_init_dev(ndev);
420         ndev->tstamp = jiffies;
421         addrconf_sysctl_register(ndev);
422         /* protected by rtnl_lock */
423         rcu_assign_pointer(dev->ip6_ptr, ndev);
424
425         /* Join all-node multicast group */
426         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
427
428         return ndev;
429 }
430
431 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
432 {
433         struct inet6_dev *idev;
434
435         ASSERT_RTNL();
436
437         if ((idev = __in6_dev_get(dev)) == NULL) {
438                 if ((idev = ipv6_add_dev(dev)) == NULL)
439                         return NULL;
440         }
441
442         if (dev->flags&IFF_UP)
443                 ipv6_mc_up(idev);
444         return idev;
445 }
446
447 #ifdef CONFIG_SYSCTL
448 static void dev_forward_change(struct inet6_dev *idev)
449 {
450         struct net_device *dev;
451         struct inet6_ifaddr *ifa;
452
453         if (!idev)
454                 return;
455         dev = idev->dev;
456         if (idev->cnf.forwarding)
457                 dev_disable_lro(dev);
458         if (dev && (dev->flags & IFF_MULTICAST)) {
459                 if (idev->cnf.forwarding)
460                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
461                 else
462                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
463         }
464         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
465                 if (ifa->flags&IFA_F_TENTATIVE)
466                         continue;
467                 if (idev->cnf.forwarding)
468                         addrconf_join_anycast(ifa);
469                 else
470                         addrconf_leave_anycast(ifa);
471         }
472 }
473
474
475 static void addrconf_forward_change(struct net *net, __s32 newf)
476 {
477         struct net_device *dev;
478         struct inet6_dev *idev;
479
480         read_lock(&dev_base_lock);
481         for_each_netdev(net, dev) {
482                 rcu_read_lock();
483                 idev = __in6_dev_get(dev);
484                 if (idev) {
485                         int changed = (!idev->cnf.forwarding) ^ (!newf);
486                         idev->cnf.forwarding = newf;
487                         if (changed)
488                                 dev_forward_change(idev);
489                 }
490                 rcu_read_unlock();
491         }
492         read_unlock(&dev_base_lock);
493 }
494
495 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
496 {
497         struct net *net;
498
499         net = (struct net *)table->extra2;
500         if (p == &net->ipv6.devconf_dflt->forwarding)
501                 return;
502
503         rtnl_lock();
504         if (p == &net->ipv6.devconf_all->forwarding) {
505                 __s32 newf = net->ipv6.devconf_all->forwarding;
506                 net->ipv6.devconf_dflt->forwarding = newf;
507                 addrconf_forward_change(net, newf);
508         } else if ((!*p) ^ (!old))
509                 dev_forward_change((struct inet6_dev *)table->extra1);
510         rtnl_unlock();
511
512         if (*p)
513                 rt6_purge_dflt_routers(net);
514 }
515 #endif
516
517 /* Nobody refers to this ifaddr, destroy it */
518
519 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
520 {
521         BUG_TRAP(ifp->if_next==NULL);
522         BUG_TRAP(ifp->lst_next==NULL);
523 #ifdef NET_REFCNT_DEBUG
524         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
525 #endif
526
527         in6_dev_put(ifp->idev);
528
529         if (del_timer(&ifp->timer))
530                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
531
532         if (!ifp->dead) {
533                 printk("Freeing alive inet6 address %p\n", ifp);
534                 return;
535         }
536         dst_release(&ifp->rt->u.dst);
537
538         kfree(ifp);
539 }
540
541 static void
542 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
543 {
544         struct inet6_ifaddr *ifa, **ifap;
545         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
546
547         /*
548          * Each device address list is sorted in order of scope -
549          * global before linklocal.
550          */
551         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
552              ifap = &ifa->if_next) {
553                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
554                         break;
555         }
556
557         ifp->if_next = *ifap;
558         *ifap = ifp;
559 }
560
561 /*
562  *      Hash function taken from net_alias.c
563  */
564 static u8 ipv6_addr_hash(const struct in6_addr *addr)
565 {
566         __u32 word;
567
568         /*
569          * We perform the hash function over the last 64 bits of the address
570          * This will include the IEEE address token on links that support it.
571          */
572
573         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
574         word ^= (word >> 16);
575         word ^= (word >> 8);
576
577         return ((word ^ (word >> 4)) & 0x0f);
578 }
579
580 /* On success it returns ifp with increased reference count */
581
582 static struct inet6_ifaddr *
583 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
584               int scope, u32 flags)
585 {
586         struct inet6_ifaddr *ifa = NULL;
587         struct rt6_info *rt;
588         int hash;
589         int err = 0;
590         int addr_type = ipv6_addr_type(addr);
591
592         if (addr_type == IPV6_ADDR_ANY ||
593             addr_type & IPV6_ADDR_MULTICAST ||
594             (!(idev->dev->flags & IFF_LOOPBACK) &&
595              addr_type & IPV6_ADDR_LOOPBACK))
596                 return ERR_PTR(-EADDRNOTAVAIL);
597
598         rcu_read_lock_bh();
599         if (idev->dead) {
600                 err = -ENODEV;                  /*XXX*/
601                 goto out2;
602         }
603
604         write_lock(&addrconf_hash_lock);
605
606         /* Ignore adding duplicate addresses on an interface */
607         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
608                 ADBG(("ipv6_add_addr: already assigned\n"));
609                 err = -EEXIST;
610                 goto out;
611         }
612
613         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
614
615         if (ifa == NULL) {
616                 ADBG(("ipv6_add_addr: malloc failed\n"));
617                 err = -ENOBUFS;
618                 goto out;
619         }
620
621         rt = addrconf_dst_alloc(idev, addr, 0);
622         if (IS_ERR(rt)) {
623                 err = PTR_ERR(rt);
624                 goto out;
625         }
626
627         ipv6_addr_copy(&ifa->addr, addr);
628
629         spin_lock_init(&ifa->lock);
630         init_timer(&ifa->timer);
631         ifa->timer.data = (unsigned long) ifa;
632         ifa->scope = scope;
633         ifa->prefix_len = pfxlen;
634         ifa->flags = flags | IFA_F_TENTATIVE;
635         ifa->cstamp = ifa->tstamp = jiffies;
636
637         ifa->rt = rt;
638
639         /*
640          * part one of RFC 4429, section 3.3
641          * We should not configure an address as
642          * optimistic if we do not yet know the link
643          * layer address of our nexhop router
644          */
645
646         if (rt->rt6i_nexthop == NULL)
647                 ifa->flags &= ~IFA_F_OPTIMISTIC;
648
649         ifa->idev = idev;
650         in6_dev_hold(idev);
651         /* For caller */
652         in6_ifa_hold(ifa);
653
654         /* Add to big hash table */
655         hash = ipv6_addr_hash(addr);
656
657         ifa->lst_next = inet6_addr_lst[hash];
658         inet6_addr_lst[hash] = ifa;
659         in6_ifa_hold(ifa);
660         write_unlock(&addrconf_hash_lock);
661
662         write_lock(&idev->lock);
663         /* Add to inet6_dev unicast addr list. */
664         ipv6_link_dev_addr(idev, ifa);
665
666 #ifdef CONFIG_IPV6_PRIVACY
667         if (ifa->flags&IFA_F_TEMPORARY) {
668                 ifa->tmp_next = idev->tempaddr_list;
669                 idev->tempaddr_list = ifa;
670                 in6_ifa_hold(ifa);
671         }
672 #endif
673
674         in6_ifa_hold(ifa);
675         write_unlock(&idev->lock);
676 out2:
677         rcu_read_unlock_bh();
678
679         if (likely(err == 0))
680                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
681         else {
682                 kfree(ifa);
683                 ifa = ERR_PTR(err);
684         }
685
686         return ifa;
687 out:
688         write_unlock(&addrconf_hash_lock);
689         goto out2;
690 }
691
692 /* This function wants to get referenced ifp and releases it before return */
693
694 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
695 {
696         struct inet6_ifaddr *ifa, **ifap;
697         struct inet6_dev *idev = ifp->idev;
698         int hash;
699         int deleted = 0, onlink = 0;
700         unsigned long expires = jiffies;
701
702         hash = ipv6_addr_hash(&ifp->addr);
703
704         ifp->dead = 1;
705
706         write_lock_bh(&addrconf_hash_lock);
707         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
708              ifap = &ifa->lst_next) {
709                 if (ifa == ifp) {
710                         *ifap = ifa->lst_next;
711                         __in6_ifa_put(ifp);
712                         ifa->lst_next = NULL;
713                         break;
714                 }
715         }
716         write_unlock_bh(&addrconf_hash_lock);
717
718         write_lock_bh(&idev->lock);
719 #ifdef CONFIG_IPV6_PRIVACY
720         if (ifp->flags&IFA_F_TEMPORARY) {
721                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
722                      ifap = &ifa->tmp_next) {
723                         if (ifa == ifp) {
724                                 *ifap = ifa->tmp_next;
725                                 if (ifp->ifpub) {
726                                         in6_ifa_put(ifp->ifpub);
727                                         ifp->ifpub = NULL;
728                                 }
729                                 __in6_ifa_put(ifp);
730                                 ifa->tmp_next = NULL;
731                                 break;
732                         }
733                 }
734         }
735 #endif
736
737         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
738                 if (ifa == ifp) {
739                         *ifap = ifa->if_next;
740                         __in6_ifa_put(ifp);
741                         ifa->if_next = NULL;
742                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
743                                 break;
744                         deleted = 1;
745                         continue;
746                 } else if (ifp->flags & IFA_F_PERMANENT) {
747                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
748                                               ifp->prefix_len)) {
749                                 if (ifa->flags & IFA_F_PERMANENT) {
750                                         onlink = 1;
751                                         if (deleted)
752                                                 break;
753                                 } else {
754                                         unsigned long lifetime;
755
756                                         if (!onlink)
757                                                 onlink = -1;
758
759                                         spin_lock(&ifa->lock);
760
761                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
762                                         /*
763                                          * Note: Because this address is
764                                          * not permanent, lifetime <
765                                          * LONG_MAX / HZ here.
766                                          */
767                                         if (time_before(expires,
768                                                         ifa->tstamp + lifetime * HZ))
769                                                 expires = ifa->tstamp + lifetime * HZ;
770                                         spin_unlock(&ifa->lock);
771                                 }
772                         }
773                 }
774                 ifap = &ifa->if_next;
775         }
776         write_unlock_bh(&idev->lock);
777
778         addrconf_del_timer(ifp);
779
780         ipv6_ifa_notify(RTM_DELADDR, ifp);
781
782         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
783
784         /*
785          * Purge or update corresponding prefix
786          *
787          * 1) we don't purge prefix here if address was not permanent.
788          *    prefix is managed by its own lifetime.
789          * 2) if there're no addresses, delete prefix.
790          * 3) if there're still other permanent address(es),
791          *    corresponding prefix is still permanent.
792          * 4) otherwise, update prefix lifetime to the
793          *    longest valid lifetime among the corresponding
794          *    addresses on the device.
795          *    Note: subsequent RA will update lifetime.
796          *
797          * --yoshfuji
798          */
799         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
800                 struct in6_addr prefix;
801                 struct rt6_info *rt;
802                 struct net *net = dev_net(ifp->idev->dev);
803                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
804                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
805
806                 if (rt && addrconf_is_prefix_route(rt)) {
807                         if (onlink == 0) {
808                                 ip6_del_rt(rt);
809                                 rt = NULL;
810                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
811                                 rt->rt6i_expires = expires;
812                                 rt->rt6i_flags |= RTF_EXPIRES;
813                         }
814                 }
815                 dst_release(&rt->u.dst);
816         }
817
818         in6_ifa_put(ifp);
819 }
820
821 #ifdef CONFIG_IPV6_PRIVACY
822 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
823 {
824         struct inet6_dev *idev = ifp->idev;
825         struct in6_addr addr, *tmpaddr;
826         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
827         unsigned long regen_advance;
828         int tmp_plen;
829         int ret = 0;
830         int max_addresses;
831         u32 addr_flags;
832
833         write_lock(&idev->lock);
834         if (ift) {
835                 spin_lock_bh(&ift->lock);
836                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
837                 spin_unlock_bh(&ift->lock);
838                 tmpaddr = &addr;
839         } else {
840                 tmpaddr = NULL;
841         }
842 retry:
843         in6_dev_hold(idev);
844         if (idev->cnf.use_tempaddr <= 0) {
845                 write_unlock(&idev->lock);
846                 printk(KERN_INFO
847                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
848                 in6_dev_put(idev);
849                 ret = -1;
850                 goto out;
851         }
852         spin_lock_bh(&ifp->lock);
853         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
854                 idev->cnf.use_tempaddr = -1;    /*XXX*/
855                 spin_unlock_bh(&ifp->lock);
856                 write_unlock(&idev->lock);
857                 printk(KERN_WARNING
858                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
859                 in6_dev_put(idev);
860                 ret = -1;
861                 goto out;
862         }
863         in6_ifa_hold(ifp);
864         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
865         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
866                 spin_unlock_bh(&ifp->lock);
867                 write_unlock(&idev->lock);
868                 printk(KERN_WARNING
869                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
870                 in6_ifa_put(ifp);
871                 in6_dev_put(idev);
872                 ret = -1;
873                 goto out;
874         }
875         memcpy(&addr.s6_addr[8], idev->rndid, 8);
876         tmp_valid_lft = min_t(__u32,
877                               ifp->valid_lft,
878                               idev->cnf.temp_valid_lft);
879         tmp_prefered_lft = min_t(__u32,
880                                  ifp->prefered_lft,
881                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
882         tmp_plen = ifp->prefix_len;
883         max_addresses = idev->cnf.max_addresses;
884         tmp_cstamp = ifp->cstamp;
885         tmp_tstamp = ifp->tstamp;
886         spin_unlock_bh(&ifp->lock);
887
888         regen_advance = idev->cnf.regen_max_retry *
889                         idev->cnf.dad_transmits *
890                         idev->nd_parms->retrans_time / HZ;
891         write_unlock(&idev->lock);
892
893         /* A temporary address is created only if this calculated Preferred
894          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
895          * an implementation must not create a temporary address with a zero
896          * Preferred Lifetime.
897          */
898         if (tmp_prefered_lft <= regen_advance) {
899                 in6_ifa_put(ifp);
900                 in6_dev_put(idev);
901                 ret = -1;
902                 goto out;
903         }
904
905         addr_flags = IFA_F_TEMPORARY;
906         /* set in addrconf_prefix_rcv() */
907         if (ifp->flags & IFA_F_OPTIMISTIC)
908                 addr_flags |= IFA_F_OPTIMISTIC;
909
910         ift = !max_addresses ||
911               ipv6_count_addresses(idev) < max_addresses ?
912                 ipv6_add_addr(idev, &addr, tmp_plen,
913                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
914                               addr_flags) : NULL;
915         if (!ift || IS_ERR(ift)) {
916                 in6_ifa_put(ifp);
917                 in6_dev_put(idev);
918                 printk(KERN_INFO
919                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
920                 tmpaddr = &addr;
921                 write_lock(&idev->lock);
922                 goto retry;
923         }
924
925         spin_lock_bh(&ift->lock);
926         ift->ifpub = ifp;
927         ift->valid_lft = tmp_valid_lft;
928         ift->prefered_lft = tmp_prefered_lft;
929         ift->cstamp = tmp_cstamp;
930         ift->tstamp = tmp_tstamp;
931         spin_unlock_bh(&ift->lock);
932
933         addrconf_dad_start(ift, 0);
934         in6_ifa_put(ift);
935         in6_dev_put(idev);
936 out:
937         return ret;
938 }
939 #endif
940
941 /*
942  *      Choose an appropriate source address (RFC3484)
943  */
944 enum {
945         IPV6_SADDR_RULE_INIT = 0,
946         IPV6_SADDR_RULE_LOCAL,
947         IPV6_SADDR_RULE_SCOPE,
948         IPV6_SADDR_RULE_PREFERRED,
949 #ifdef CONFIG_IPV6_MIP6
950         IPV6_SADDR_RULE_HOA,
951 #endif
952         IPV6_SADDR_RULE_OIF,
953         IPV6_SADDR_RULE_LABEL,
954 #ifdef CONFIG_IPV6_PRIVACY
955         IPV6_SADDR_RULE_PRIVACY,
956 #endif
957         IPV6_SADDR_RULE_ORCHID,
958         IPV6_SADDR_RULE_PREFIX,
959         IPV6_SADDR_RULE_MAX
960 };
961
962 struct ipv6_saddr_score {
963         int                     rule;
964         int                     addr_type;
965         struct inet6_ifaddr     *ifa;
966         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
967         int                     scopedist;
968         int                     matchlen;
969 };
970
971 struct ipv6_saddr_dst {
972         const struct in6_addr *addr;
973         int ifindex;
974         int scope;
975         int label;
976         unsigned int prefs;
977 };
978
979 static inline int ipv6_saddr_preferred(int type)
980 {
981         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
982                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
983                 return 1;
984         return 0;
985 }
986
987 static int ipv6_get_saddr_eval(struct net *net,
988                                struct ipv6_saddr_score *score,
989                                struct ipv6_saddr_dst *dst,
990                                int i)
991 {
992         int ret;
993
994         if (i <= score->rule) {
995                 switch (i) {
996                 case IPV6_SADDR_RULE_SCOPE:
997                         ret = score->scopedist;
998                         break;
999                 case IPV6_SADDR_RULE_PREFIX:
1000                         ret = score->matchlen;
1001                         break;
1002                 default:
1003                         ret = !!test_bit(i, score->scorebits);
1004                 }
1005                 goto out;
1006         }
1007
1008         switch (i) {
1009         case IPV6_SADDR_RULE_INIT:
1010                 /* Rule 0: remember if hiscore is not ready yet */
1011                 ret = !!score->ifa;
1012                 break;
1013         case IPV6_SADDR_RULE_LOCAL:
1014                 /* Rule 1: Prefer same address */
1015                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1016                 break;
1017         case IPV6_SADDR_RULE_SCOPE:
1018                 /* Rule 2: Prefer appropriate scope
1019                  *
1020                  *      ret
1021                  *       ^
1022                  *    -1 |  d 15
1023                  *    ---+--+-+---> scope
1024                  *       |
1025                  *       |             d is scope of the destination.
1026                  *  B-d  |  \
1027                  *       |   \      <- smaller scope is better if
1028                  *  B-15 |    \        if scope is enough for destinaion.
1029                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1030                  * d-C-1 | /
1031                  *       |/         <- greater is better
1032                  *   -C  /             if scope is not enough for destination.
1033                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1034                  *
1035                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1036                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1037                  * Assume B = 0 and we get C > 29.
1038                  */
1039                 ret = __ipv6_addr_src_scope(score->addr_type);
1040                 if (ret >= dst->scope)
1041                         ret = -ret;
1042                 else
1043                         ret -= 128;     /* 30 is enough */
1044                 score->scopedist = ret;
1045                 break;
1046         case IPV6_SADDR_RULE_PREFERRED:
1047                 /* Rule 3: Avoid deprecated and optimistic addresses */
1048                 ret = ipv6_saddr_preferred(score->addr_type) ||
1049                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1050                 break;
1051 #ifdef CONFIG_IPV6_MIP6
1052         case IPV6_SADDR_RULE_HOA:
1053             {
1054                 /* Rule 4: Prefer home address */
1055                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1056                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1057                 break;
1058             }
1059 #endif
1060         case IPV6_SADDR_RULE_OIF:
1061                 /* Rule 5: Prefer outgoing interface */
1062                 ret = (!dst->ifindex ||
1063                        dst->ifindex == score->ifa->idev->dev->ifindex);
1064                 break;
1065         case IPV6_SADDR_RULE_LABEL:
1066                 /* Rule 6: Prefer matching label */
1067                 ret = ipv6_addr_label(net,
1068                                       &score->ifa->addr, score->addr_type,
1069                                       score->ifa->idev->dev->ifindex) == dst->label;
1070                 break;
1071 #ifdef CONFIG_IPV6_PRIVACY
1072         case IPV6_SADDR_RULE_PRIVACY:
1073             {
1074                 /* Rule 7: Prefer public address
1075                  * Note: prefer temprary address if use_tempaddr >= 2
1076                  */
1077                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1078                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1079                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1080                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1081                 break;
1082             }
1083 #endif
1084         case IPV6_SADDR_RULE_ORCHID:
1085                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1086                  *          non-ORCHID vs non-ORCHID
1087                  */
1088                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1089                         ipv6_addr_orchid(dst->addr));
1090                 break;
1091         case IPV6_SADDR_RULE_PREFIX:
1092                 /* Rule 8: Use longest matching prefix */
1093                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1094                                                        dst->addr);
1095                 break;
1096         default:
1097                 ret = 0;
1098         }
1099
1100         if (ret)
1101                 __set_bit(i, score->scorebits);
1102         score->rule = i;
1103 out:
1104         return ret;
1105 }
1106
1107 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1108                        const struct in6_addr *daddr, unsigned int prefs,
1109                        struct in6_addr *saddr)
1110 {
1111         struct ipv6_saddr_score scores[2],
1112                                 *score = &scores[0], *hiscore = &scores[1];
1113         struct net *net = dev_net(dst_dev);
1114         struct ipv6_saddr_dst dst;
1115         struct net_device *dev;
1116         int dst_type;
1117
1118         dst_type = __ipv6_addr_type(daddr);
1119         dst.addr = daddr;
1120         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1121         dst.scope = __ipv6_addr_src_scope(dst_type);
1122         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1123         dst.prefs = prefs;
1124
1125         hiscore->rule = -1;
1126         hiscore->ifa = NULL;
1127
1128         read_lock(&dev_base_lock);
1129         rcu_read_lock();
1130
1131         for_each_netdev(net, dev) {
1132                 struct inet6_dev *idev;
1133
1134                 /* Candidate Source Address (section 4)
1135                  *  - multicast and link-local destination address,
1136                  *    the set of candidate source address MUST only
1137                  *    include addresses assigned to interfaces
1138                  *    belonging to the same link as the outgoing
1139                  *    interface.
1140                  * (- For site-local destination addresses, the
1141                  *    set of candidate source addresses MUST only
1142                  *    include addresses assigned to interfaces
1143                  *    belonging to the same site as the outgoing
1144                  *    interface.)
1145                  */
1146                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1147                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1148                     dst.ifindex && dev->ifindex != dst.ifindex)
1149                         continue;
1150
1151                 idev = __in6_dev_get(dev);
1152                 if (!idev)
1153                         continue;
1154
1155                 read_lock_bh(&idev->lock);
1156                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1157                         int i;
1158
1159                         /*
1160                          * - Tentative Address (RFC2462 section 5.4)
1161                          *  - A tentative address is not considered
1162                          *    "assigned to an interface" in the traditional
1163                          *    sense, unless it is also flagged as optimistic.
1164                          * - Candidate Source Address (section 4)
1165                          *  - In any case, anycast addresses, multicast
1166                          *    addresses, and the unspecified address MUST
1167                          *    NOT be included in a candidate set.
1168                          */
1169                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1170                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171                                 continue;
1172
1173                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174
1175                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1176                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1177                                 LIMIT_NETDEBUG(KERN_DEBUG
1178                                                "ADDRCONF: unspecified / multicast address "
1179                                                "assigned as unicast address on %s",
1180                                                dev->name);
1181                                 continue;
1182                         }
1183
1184                         score->rule = -1;
1185                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186
1187                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1188                                 int minihiscore, miniscore;
1189
1190                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1191                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192
1193                                 if (minihiscore > miniscore) {
1194                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1195                                             score->scopedist > 0) {
1196                                                 /*
1197                                                  * special case:
1198                                                  * each remaining entry
1199                                                  * has too small (not enough)
1200                                                  * scope, because ifa entries
1201                                                  * are sorted by their scope
1202                                                  * values.
1203                                                  */
1204                                                 goto try_nextdev;
1205                                         }
1206                                         break;
1207                                 } else if (minihiscore < miniscore) {
1208                                         struct ipv6_saddr_score *tmp;
1209
1210                                         if (hiscore->ifa)
1211                                                 in6_ifa_put(hiscore->ifa);
1212
1213                                         in6_ifa_hold(score->ifa);
1214
1215                                         tmp = hiscore;
1216                                         hiscore = score;
1217                                         score = tmp;
1218
1219                                         /* restore our iterator */
1220                                         score->ifa = hiscore->ifa;
1221
1222                                         break;
1223                                 }
1224                         }
1225                 }
1226 try_nextdev:
1227                 read_unlock_bh(&idev->lock);
1228         }
1229         rcu_read_unlock();
1230         read_unlock(&dev_base_lock);
1231
1232         if (!hiscore->ifa)
1233                 return -EADDRNOTAVAIL;
1234
1235         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1236         in6_ifa_put(hiscore->ifa);
1237         return 0;
1238 }
1239
1240 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1241
1242 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1243                     unsigned char banned_flags)
1244 {
1245         struct inet6_dev *idev;
1246         int err = -EADDRNOTAVAIL;
1247
1248         rcu_read_lock();
1249         if ((idev = __in6_dev_get(dev)) != NULL) {
1250                 struct inet6_ifaddr *ifp;
1251
1252                 read_lock_bh(&idev->lock);
1253                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1254                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1255                                 ipv6_addr_copy(addr, &ifp->addr);
1256                                 err = 0;
1257                                 break;
1258                         }
1259                 }
1260                 read_unlock_bh(&idev->lock);
1261         }
1262         rcu_read_unlock();
1263         return err;
1264 }
1265
1266 static int ipv6_count_addresses(struct inet6_dev *idev)
1267 {
1268         int cnt = 0;
1269         struct inet6_ifaddr *ifp;
1270
1271         read_lock_bh(&idev->lock);
1272         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1273                 cnt++;
1274         read_unlock_bh(&idev->lock);
1275         return cnt;
1276 }
1277
1278 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1279                   struct net_device *dev, int strict)
1280 {
1281         struct inet6_ifaddr * ifp;
1282         u8 hash = ipv6_addr_hash(addr);
1283
1284         read_lock_bh(&addrconf_hash_lock);
1285         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1286                 if (!net_eq(dev_net(ifp->idev->dev), net))
1287                         continue;
1288                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1289                     !(ifp->flags&IFA_F_TENTATIVE)) {
1290                         if (dev == NULL || ifp->idev->dev == dev ||
1291                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1292                                 break;
1293                 }
1294         }
1295         read_unlock_bh(&addrconf_hash_lock);
1296         return ifp != NULL;
1297 }
1298 EXPORT_SYMBOL(ipv6_chk_addr);
1299
1300 static
1301 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1302                        struct net_device *dev)
1303 {
1304         struct inet6_ifaddr * ifp;
1305         u8 hash = ipv6_addr_hash(addr);
1306
1307         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1308                 if (!net_eq(dev_net(ifp->idev->dev), net))
1309                         continue;
1310                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1311                         if (dev == NULL || ifp->idev->dev == dev)
1312                                 break;
1313                 }
1314         }
1315         return ifp != NULL;
1316 }
1317
1318 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1319 {
1320         struct inet6_dev *idev;
1321         struct inet6_ifaddr *ifa;
1322         int     onlink;
1323
1324         onlink = 0;
1325         rcu_read_lock();
1326         idev = __in6_dev_get(dev);
1327         if (idev) {
1328                 read_lock_bh(&idev->lock);
1329                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1330                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1331                                                    ifa->prefix_len);
1332                         if (onlink)
1333                                 break;
1334                 }
1335                 read_unlock_bh(&idev->lock);
1336         }
1337         rcu_read_unlock();
1338         return onlink;
1339 }
1340
1341 EXPORT_SYMBOL(ipv6_chk_prefix);
1342
1343 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1344                                      struct net_device *dev, int strict)
1345 {
1346         struct inet6_ifaddr * ifp;
1347         u8 hash = ipv6_addr_hash(addr);
1348
1349         read_lock_bh(&addrconf_hash_lock);
1350         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1351                 if (!net_eq(dev_net(ifp->idev->dev), net))
1352                         continue;
1353                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354                         if (dev == NULL || ifp->idev->dev == dev ||
1355                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356                                 in6_ifa_hold(ifp);
1357                                 break;
1358                         }
1359                 }
1360         }
1361         read_unlock_bh(&addrconf_hash_lock);
1362
1363         return ifp;
1364 }
1365
1366 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1367 {
1368         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1369         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1370         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1371         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1372         int sk_ipv6only = ipv6_only_sock(sk);
1373         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1374         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1375         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1376
1377         if (!sk2_rcv_saddr && !sk_ipv6only)
1378                 return 1;
1379
1380         if (addr_type2 == IPV6_ADDR_ANY &&
1381             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1382                 return 1;
1383
1384         if (addr_type == IPV6_ADDR_ANY &&
1385             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1386                 return 1;
1387
1388         if (sk2_rcv_saddr6 &&
1389             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1390                 return 1;
1391
1392         if (addr_type == IPV6_ADDR_MAPPED &&
1393             !sk2_ipv6only &&
1394             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1395                 return 1;
1396
1397         return 0;
1398 }
1399
1400 /* Gets referenced address, destroys ifaddr */
1401
1402 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1403 {
1404         if (ifp->flags&IFA_F_PERMANENT) {
1405                 spin_lock_bh(&ifp->lock);
1406                 addrconf_del_timer(ifp);
1407                 ifp->flags |= IFA_F_TENTATIVE;
1408                 spin_unlock_bh(&ifp->lock);
1409                 in6_ifa_put(ifp);
1410 #ifdef CONFIG_IPV6_PRIVACY
1411         } else if (ifp->flags&IFA_F_TEMPORARY) {
1412                 struct inet6_ifaddr *ifpub;
1413                 spin_lock_bh(&ifp->lock);
1414                 ifpub = ifp->ifpub;
1415                 if (ifpub) {
1416                         in6_ifa_hold(ifpub);
1417                         spin_unlock_bh(&ifp->lock);
1418                         ipv6_create_tempaddr(ifpub, ifp);
1419                         in6_ifa_put(ifpub);
1420                 } else {
1421                         spin_unlock_bh(&ifp->lock);
1422                 }
1423                 ipv6_del_addr(ifp);
1424 #endif
1425         } else
1426                 ipv6_del_addr(ifp);
1427 }
1428
1429 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1430 {
1431         struct inet6_dev *idev = ifp->idev;
1432         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1433                 struct in6_addr addr;
1434
1435                 addr.s6_addr32[0] = htonl(0xfe800000);
1436                 addr.s6_addr32[1] = 0;
1437
1438                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1439                     ipv6_addr_equal(&ifp->addr, &addr)) {
1440                         /* DAD failed for link-local based on MAC address */
1441                         idev->cnf.disable_ipv6 = 1;
1442                 }
1443         }
1444
1445         if (net_ratelimit())
1446                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1447         addrconf_dad_stop(ifp);
1448 }
1449
1450 /* Join to solicited addr multicast group. */
1451
1452 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1453 {
1454         struct in6_addr maddr;
1455
1456         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1457                 return;
1458
1459         addrconf_addr_solict_mult(addr, &maddr);
1460         ipv6_dev_mc_inc(dev, &maddr);
1461 }
1462
1463 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1464 {
1465         struct in6_addr maddr;
1466
1467         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1468                 return;
1469
1470         addrconf_addr_solict_mult(addr, &maddr);
1471         __ipv6_dev_mc_dec(idev, &maddr);
1472 }
1473
1474 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1475 {
1476         struct in6_addr addr;
1477         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1478         if (ipv6_addr_any(&addr))
1479                 return;
1480         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1481 }
1482
1483 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1484 {
1485         struct in6_addr addr;
1486         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1487         if (ipv6_addr_any(&addr))
1488                 return;
1489         __ipv6_dev_ac_dec(ifp->idev, &addr);
1490 }
1491
1492 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1493 {
1494         if (dev->addr_len != ETH_ALEN)
1495                 return -1;
1496         memcpy(eui, dev->dev_addr, 3);
1497         memcpy(eui + 5, dev->dev_addr + 3, 3);
1498
1499         /*
1500          * The zSeries OSA network cards can be shared among various
1501          * OS instances, but the OSA cards have only one MAC address.
1502          * This leads to duplicate address conflicts in conjunction
1503          * with IPv6 if more than one instance uses the same card.
1504          *
1505          * The driver for these cards can deliver a unique 16-bit
1506          * identifier for each instance sharing the same card.  It is
1507          * placed instead of 0xFFFE in the interface identifier.  The
1508          * "u" bit of the interface identifier is not inverted in this
1509          * case.  Hence the resulting interface identifier has local
1510          * scope according to RFC2373.
1511          */
1512         if (dev->dev_id) {
1513                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1514                 eui[4] = dev->dev_id & 0xFF;
1515         } else {
1516                 eui[3] = 0xFF;
1517                 eui[4] = 0xFE;
1518                 eui[0] ^= 2;
1519         }
1520         return 0;
1521 }
1522
1523 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1524 {
1525         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1526         if (dev->addr_len != ARCNET_ALEN)
1527                 return -1;
1528         memset(eui, 0, 7);
1529         eui[7] = *(u8*)dev->dev_addr;
1530         return 0;
1531 }
1532
1533 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1534 {
1535         if (dev->addr_len != INFINIBAND_ALEN)
1536                 return -1;
1537         memcpy(eui, dev->dev_addr + 12, 8);
1538         eui[0] |= 2;
1539         return 0;
1540 }
1541
1542 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1543 {
1544         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1545                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1546                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1547                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1548                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1549                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1550         eui[1] = 0;
1551         eui[2] = 0x5E;
1552         eui[3] = 0xFE;
1553         memcpy(eui + 4, &addr, 4);
1554         return 0;
1555 }
1556 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1557
1558 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1559 {
1560         if (dev->priv_flags & IFF_ISATAP)
1561                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1562         return -1;
1563 }
1564
1565 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1566 {
1567         switch (dev->type) {
1568         case ARPHRD_ETHER:
1569         case ARPHRD_FDDI:
1570         case ARPHRD_IEEE802_TR:
1571                 return addrconf_ifid_eui48(eui, dev);
1572         case ARPHRD_ARCNET:
1573                 return addrconf_ifid_arcnet(eui, dev);
1574         case ARPHRD_INFINIBAND:
1575                 return addrconf_ifid_infiniband(eui, dev);
1576         case ARPHRD_SIT:
1577                 return addrconf_ifid_sit(eui, dev);
1578         }
1579         return -1;
1580 }
1581
1582 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1583 {
1584         int err = -1;
1585         struct inet6_ifaddr *ifp;
1586
1587         read_lock_bh(&idev->lock);
1588         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1589                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1590                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1591                         err = 0;
1592                         break;
1593                 }
1594         }
1595         read_unlock_bh(&idev->lock);
1596         return err;
1597 }
1598
1599 #ifdef CONFIG_IPV6_PRIVACY
1600 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1601 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1602 {
1603 regen:
1604         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1605         idev->rndid[0] &= ~0x02;
1606
1607         /*
1608          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1609          * check if generated address is not inappropriate
1610          *
1611          *  - Reserved subnet anycast (RFC 2526)
1612          *      11111101 11....11 1xxxxxxx
1613          *  - ISATAP (RFC4214) 6.1
1614          *      00-00-5E-FE-xx-xx-xx-xx
1615          *  - value 0
1616          *  - XXX: already assigned to an address on the device
1617          */
1618         if (idev->rndid[0] == 0xfd &&
1619             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1620             (idev->rndid[7]&0x80))
1621                 goto regen;
1622         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1623                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1624                         goto regen;
1625                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1626                         goto regen;
1627         }
1628
1629         return 0;
1630 }
1631
1632 static void ipv6_regen_rndid(unsigned long data)
1633 {
1634         struct inet6_dev *idev = (struct inet6_dev *) data;
1635         unsigned long expires;
1636
1637         rcu_read_lock_bh();
1638         write_lock_bh(&idev->lock);
1639
1640         if (idev->dead)
1641                 goto out;
1642
1643         if (__ipv6_regen_rndid(idev) < 0)
1644                 goto out;
1645
1646         expires = jiffies +
1647                 idev->cnf.temp_prefered_lft * HZ -
1648                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1649         if (time_before(expires, jiffies)) {
1650                 printk(KERN_WARNING
1651                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1652                         idev->dev->name);
1653                 goto out;
1654         }
1655
1656         if (!mod_timer(&idev->regen_timer, expires))
1657                 in6_dev_hold(idev);
1658
1659 out:
1660         write_unlock_bh(&idev->lock);
1661         rcu_read_unlock_bh();
1662         in6_dev_put(idev);
1663 }
1664
1665 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1666         int ret = 0;
1667
1668         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1669                 ret = __ipv6_regen_rndid(idev);
1670         return ret;
1671 }
1672 #endif
1673
1674 /*
1675  *      Add prefix route.
1676  */
1677
1678 static void
1679 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1680                       unsigned long expires, u32 flags)
1681 {
1682         struct fib6_config cfg = {
1683                 .fc_table = RT6_TABLE_PREFIX,
1684                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1685                 .fc_ifindex = dev->ifindex,
1686                 .fc_expires = expires,
1687                 .fc_dst_len = plen,
1688                 .fc_flags = RTF_UP | flags,
1689                 .fc_nlinfo.nl_net = dev_net(dev),
1690         };
1691
1692         ipv6_addr_copy(&cfg.fc_dst, pfx);
1693
1694         /* Prevent useless cloning on PtP SIT.
1695            This thing is done here expecting that the whole
1696            class of non-broadcast devices need not cloning.
1697          */
1698 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1699         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1700                 cfg.fc_flags |= RTF_NONEXTHOP;
1701 #endif
1702
1703         ip6_route_add(&cfg);
1704 }
1705
1706 /* Create "default" multicast route to the interface */
1707
1708 static void addrconf_add_mroute(struct net_device *dev)
1709 {
1710         struct fib6_config cfg = {
1711                 .fc_table = RT6_TABLE_LOCAL,
1712                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1713                 .fc_ifindex = dev->ifindex,
1714                 .fc_dst_len = 8,
1715                 .fc_flags = RTF_UP,
1716                 .fc_nlinfo.nl_net = dev_net(dev),
1717         };
1718
1719         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1720
1721         ip6_route_add(&cfg);
1722 }
1723
1724 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1725 static void sit_route_add(struct net_device *dev)
1726 {
1727         struct fib6_config cfg = {
1728                 .fc_table = RT6_TABLE_MAIN,
1729                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1730                 .fc_ifindex = dev->ifindex,
1731                 .fc_dst_len = 96,
1732                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1733                 .fc_nlinfo.nl_net = dev_net(dev),
1734         };
1735
1736         /* prefix length - 96 bits "::d.d.d.d" */
1737         ip6_route_add(&cfg);
1738 }
1739 #endif
1740
1741 static void addrconf_add_lroute(struct net_device *dev)
1742 {
1743         struct in6_addr addr;
1744
1745         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1746         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1747 }
1748
1749 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1750 {
1751         struct inet6_dev *idev;
1752
1753         ASSERT_RTNL();
1754
1755         if ((idev = ipv6_find_idev(dev)) == NULL)
1756                 return NULL;
1757
1758         /* Add default multicast route */
1759         addrconf_add_mroute(dev);
1760
1761         /* Add link local route */
1762         addrconf_add_lroute(dev);
1763         return idev;
1764 }
1765
1766 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1767 {
1768         struct prefix_info *pinfo;
1769         __u32 valid_lft;
1770         __u32 prefered_lft;
1771         int addr_type;
1772         struct inet6_dev *in6_dev;
1773
1774         pinfo = (struct prefix_info *) opt;
1775
1776         if (len < sizeof(struct prefix_info)) {
1777                 ADBG(("addrconf: prefix option too short\n"));
1778                 return;
1779         }
1780
1781         /*
1782          *      Validation checks ([ADDRCONF], page 19)
1783          */
1784
1785         addr_type = ipv6_addr_type(&pinfo->prefix);
1786
1787         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1788                 return;
1789
1790         valid_lft = ntohl(pinfo->valid);
1791         prefered_lft = ntohl(pinfo->prefered);
1792
1793         if (prefered_lft > valid_lft) {
1794                 if (net_ratelimit())
1795                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1796                 return;
1797         }
1798
1799         in6_dev = in6_dev_get(dev);
1800
1801         if (in6_dev == NULL) {
1802                 if (net_ratelimit())
1803                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1804                 return;
1805         }
1806
1807         /*
1808          *      Two things going on here:
1809          *      1) Add routes for on-link prefixes
1810          *      2) Configure prefixes with the auto flag set
1811          */
1812
1813         if (pinfo->onlink) {
1814                 struct rt6_info *rt;
1815                 unsigned long rt_expires;
1816
1817                 /* Avoid arithmetic overflow. Really, we could
1818                  * save rt_expires in seconds, likely valid_lft,
1819                  * but it would require division in fib gc, that it
1820                  * not good.
1821                  */
1822                 if (HZ > USER_HZ)
1823                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1824                 else
1825                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1826
1827                 if (addrconf_finite_timeout(rt_expires))
1828                         rt_expires *= HZ;
1829
1830                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1831                                 dev->ifindex, 1);
1832
1833                 if (rt && addrconf_is_prefix_route(rt)) {
1834                         /* Autoconf prefix route */
1835                         if (valid_lft == 0) {
1836                                 ip6_del_rt(rt);
1837                                 rt = NULL;
1838                         } else if (addrconf_finite_timeout(rt_expires)) {
1839                                 /* not infinity */
1840                                 rt->rt6i_expires = jiffies + rt_expires;
1841                                 rt->rt6i_flags |= RTF_EXPIRES;
1842                         } else {
1843                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1844                                 rt->rt6i_expires = 0;
1845                         }
1846                 } else if (valid_lft) {
1847                         clock_t expires = 0;
1848                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1849                         if (addrconf_finite_timeout(rt_expires)) {
1850                                 /* not infinity */
1851                                 flags |= RTF_EXPIRES;
1852                                 expires = jiffies_to_clock_t(rt_expires);
1853                         }
1854                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1855                                               dev, expires, flags);
1856                 }
1857                 if (rt)
1858                         dst_release(&rt->u.dst);
1859         }
1860
1861         /* Try to figure out our local address for this prefix */
1862
1863         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1864                 struct inet6_ifaddr * ifp;
1865                 struct in6_addr addr;
1866                 int create = 0, update_lft = 0;
1867
1868                 if (pinfo->prefix_len == 64) {
1869                         memcpy(&addr, &pinfo->prefix, 8);
1870                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1871                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1872                                 in6_dev_put(in6_dev);
1873                                 return;
1874                         }
1875                         goto ok;
1876                 }
1877                 if (net_ratelimit())
1878                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1879                                pinfo->prefix_len);
1880                 in6_dev_put(in6_dev);
1881                 return;
1882
1883 ok:
1884
1885                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1886
1887                 if (ifp == NULL && valid_lft) {
1888                         int max_addresses = in6_dev->cnf.max_addresses;
1889                         u32 addr_flags = 0;
1890
1891 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1892                         if (in6_dev->cnf.optimistic_dad &&
1893                             !ipv6_devconf.forwarding)
1894                                 addr_flags = IFA_F_OPTIMISTIC;
1895 #endif
1896
1897                         /* Do not allow to create too much of autoconfigured
1898                          * addresses; this would be too easy way to crash kernel.
1899                          */
1900                         if (!max_addresses ||
1901                             ipv6_count_addresses(in6_dev) < max_addresses)
1902                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1903                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1904                                                     addr_flags);
1905
1906                         if (!ifp || IS_ERR(ifp)) {
1907                                 in6_dev_put(in6_dev);
1908                                 return;
1909                         }
1910
1911                         update_lft = create = 1;
1912                         ifp->cstamp = jiffies;
1913                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1914                 }
1915
1916                 if (ifp) {
1917                         int flags;
1918                         unsigned long now;
1919 #ifdef CONFIG_IPV6_PRIVACY
1920                         struct inet6_ifaddr *ift;
1921 #endif
1922                         u32 stored_lft;
1923
1924                         /* update lifetime (RFC2462 5.5.3 e) */
1925                         spin_lock(&ifp->lock);
1926                         now = jiffies;
1927                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1928                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1929                         else
1930                                 stored_lft = 0;
1931                         if (!update_lft && stored_lft) {
1932                                 if (valid_lft > MIN_VALID_LIFETIME ||
1933                                     valid_lft > stored_lft)
1934                                         update_lft = 1;
1935                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1936                                         /* valid_lft <= stored_lft is always true */
1937                                         /* XXX: IPsec */
1938                                         update_lft = 0;
1939                                 } else {
1940                                         valid_lft = MIN_VALID_LIFETIME;
1941                                         if (valid_lft < prefered_lft)
1942                                                 prefered_lft = valid_lft;
1943                                         update_lft = 1;
1944                                 }
1945                         }
1946
1947                         if (update_lft) {
1948                                 ifp->valid_lft = valid_lft;
1949                                 ifp->prefered_lft = prefered_lft;
1950                                 ifp->tstamp = now;
1951                                 flags = ifp->flags;
1952                                 ifp->flags &= ~IFA_F_DEPRECATED;
1953                                 spin_unlock(&ifp->lock);
1954
1955                                 if (!(flags&IFA_F_TENTATIVE))
1956                                         ipv6_ifa_notify(0, ifp);
1957                         } else
1958                                 spin_unlock(&ifp->lock);
1959
1960 #ifdef CONFIG_IPV6_PRIVACY
1961                         read_lock_bh(&in6_dev->lock);
1962                         /* update all temporary addresses in the list */
1963                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1964                                 /*
1965                                  * When adjusting the lifetimes of an existing
1966                                  * temporary address, only lower the lifetimes.
1967                                  * Implementations must not increase the
1968                                  * lifetimes of an existing temporary address
1969                                  * when processing a Prefix Information Option.
1970                                  */
1971                                 if (ifp != ift->ifpub)
1972                                         continue;
1973
1974                                 spin_lock(&ift->lock);
1975                                 flags = ift->flags;
1976                                 if (ift->valid_lft > valid_lft &&
1977                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1978                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1979                                 if (ift->prefered_lft > prefered_lft &&
1980                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1981                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1982                                 spin_unlock(&ift->lock);
1983                                 if (!(flags&IFA_F_TENTATIVE))
1984                                         ipv6_ifa_notify(0, ift);
1985                         }
1986
1987                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1988                                 /*
1989                                  * When a new public address is created as described in [ADDRCONF],
1990                                  * also create a new temporary address.
1991                                  */
1992                                 read_unlock_bh(&in6_dev->lock);
1993                                 ipv6_create_tempaddr(ifp, NULL);
1994                         } else {
1995                                 read_unlock_bh(&in6_dev->lock);
1996                         }
1997 #endif
1998                         in6_ifa_put(ifp);
1999                         addrconf_verify(0);
2000                 }
2001         }
2002         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2003         in6_dev_put(in6_dev);
2004 }
2005
2006 /*
2007  *      Set destination address.
2008  *      Special case for SIT interfaces where we create a new "virtual"
2009  *      device.
2010  */
2011 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2012 {
2013         struct in6_ifreq ireq;
2014         struct net_device *dev;
2015         int err = -EINVAL;
2016
2017         rtnl_lock();
2018
2019         err = -EFAULT;
2020         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2021                 goto err_exit;
2022
2023         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2024
2025         err = -ENODEV;
2026         if (dev == NULL)
2027                 goto err_exit;
2028
2029 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2030         if (dev->type == ARPHRD_SIT) {
2031                 struct ifreq ifr;
2032                 mm_segment_t    oldfs;
2033                 struct ip_tunnel_parm p;
2034
2035                 err = -EADDRNOTAVAIL;
2036                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2037                         goto err_exit;
2038
2039                 memset(&p, 0, sizeof(p));
2040                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2041                 p.iph.saddr = 0;
2042                 p.iph.version = 4;
2043                 p.iph.ihl = 5;
2044                 p.iph.protocol = IPPROTO_IPV6;
2045                 p.iph.ttl = 64;
2046                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2047
2048                 oldfs = get_fs(); set_fs(KERNEL_DS);
2049                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2050                 set_fs(oldfs);
2051
2052                 if (err == 0) {
2053                         err = -ENOBUFS;
2054                         dev = __dev_get_by_name(net, p.name);
2055                         if (!dev)
2056                                 goto err_exit;
2057                         err = dev_open(dev);
2058                 }
2059         }
2060 #endif
2061
2062 err_exit:
2063         rtnl_unlock();
2064         return err;
2065 }
2066
2067 /*
2068  *      Manual configuration of address on an interface
2069  */
2070 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2071                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2072                           __u32 valid_lft)
2073 {
2074         struct inet6_ifaddr *ifp;
2075         struct inet6_dev *idev;
2076         struct net_device *dev;
2077         int scope;
2078         u32 flags;
2079         clock_t expires;
2080         unsigned long timeout;
2081
2082         ASSERT_RTNL();
2083
2084         if (plen > 128)
2085                 return -EINVAL;
2086
2087         /* check the lifetime */
2088         if (!valid_lft || prefered_lft > valid_lft)
2089                 return -EINVAL;
2090
2091         dev = __dev_get_by_index(net, ifindex);
2092         if (!dev)
2093                 return -ENODEV;
2094
2095         if ((idev = addrconf_add_dev(dev)) == NULL)
2096                 return -ENOBUFS;
2097
2098         scope = ipv6_addr_scope(pfx);
2099
2100         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2101         if (addrconf_finite_timeout(timeout)) {
2102                 expires = jiffies_to_clock_t(timeout * HZ);
2103                 valid_lft = timeout;
2104                 flags = RTF_EXPIRES;
2105         } else {
2106                 expires = 0;
2107                 flags = 0;
2108                 ifa_flags |= IFA_F_PERMANENT;
2109         }
2110
2111         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2112         if (addrconf_finite_timeout(timeout)) {
2113                 if (timeout == 0)
2114                         ifa_flags |= IFA_F_DEPRECATED;
2115                 prefered_lft = timeout;
2116         }
2117
2118         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2119
2120         if (!IS_ERR(ifp)) {
2121                 spin_lock_bh(&ifp->lock);
2122                 ifp->valid_lft = valid_lft;
2123                 ifp->prefered_lft = prefered_lft;
2124                 ifp->tstamp = jiffies;
2125                 spin_unlock_bh(&ifp->lock);
2126
2127                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2128                                       expires, flags);
2129                 /*
2130                  * Note that section 3.1 of RFC 4429 indicates
2131                  * that the Optimistic flag should not be set for
2132                  * manually configured addresses
2133                  */
2134                 addrconf_dad_start(ifp, 0);
2135                 in6_ifa_put(ifp);
2136                 addrconf_verify(0);
2137                 return 0;
2138         }
2139
2140         return PTR_ERR(ifp);
2141 }
2142
2143 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2144                           unsigned int plen)
2145 {
2146         struct inet6_ifaddr *ifp;
2147         struct inet6_dev *idev;
2148         struct net_device *dev;
2149
2150         if (plen > 128)
2151                 return -EINVAL;
2152
2153         dev = __dev_get_by_index(net, ifindex);
2154         if (!dev)
2155                 return -ENODEV;
2156
2157         if ((idev = __in6_dev_get(dev)) == NULL)
2158                 return -ENXIO;
2159
2160         read_lock_bh(&idev->lock);
2161         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2162                 if (ifp->prefix_len == plen &&
2163                     ipv6_addr_equal(pfx, &ifp->addr)) {
2164                         in6_ifa_hold(ifp);
2165                         read_unlock_bh(&idev->lock);
2166
2167                         ipv6_del_addr(ifp);
2168
2169                         /* If the last address is deleted administratively,
2170                            disable IPv6 on this interface.
2171                          */
2172                         if (idev->addr_list == NULL)
2173                                 addrconf_ifdown(idev->dev, 1);
2174                         return 0;
2175                 }
2176         }
2177         read_unlock_bh(&idev->lock);
2178         return -EADDRNOTAVAIL;
2179 }
2180
2181
2182 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2183 {
2184         struct in6_ifreq ireq;
2185         int err;
2186
2187         if (!capable(CAP_NET_ADMIN))
2188                 return -EPERM;
2189
2190         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2191                 return -EFAULT;
2192
2193         rtnl_lock();
2194         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2195                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2196                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2197         rtnl_unlock();
2198         return err;
2199 }
2200
2201 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2202 {
2203         struct in6_ifreq ireq;
2204         int err;
2205
2206         if (!capable(CAP_NET_ADMIN))
2207                 return -EPERM;
2208
2209         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2210                 return -EFAULT;
2211
2212         rtnl_lock();
2213         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2214                              ireq.ifr6_prefixlen);
2215         rtnl_unlock();
2216         return err;
2217 }
2218
2219 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2220 static void sit_add_v4_addrs(struct inet6_dev *idev)
2221 {
2222         struct inet6_ifaddr * ifp;
2223         struct in6_addr addr;
2224         struct net_device *dev;
2225         struct net *net = dev_net(idev->dev);
2226         int scope;
2227
2228         ASSERT_RTNL();
2229
2230         memset(&addr, 0, sizeof(struct in6_addr));
2231         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2232
2233         if (idev->dev->flags&IFF_POINTOPOINT) {
2234                 addr.s6_addr32[0] = htonl(0xfe800000);
2235                 scope = IFA_LINK;
2236         } else {
2237                 scope = IPV6_ADDR_COMPATv4;
2238         }
2239
2240         if (addr.s6_addr32[3]) {
2241                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2242                 if (!IS_ERR(ifp)) {
2243                         spin_lock_bh(&ifp->lock);
2244                         ifp->flags &= ~IFA_F_TENTATIVE;
2245                         spin_unlock_bh(&ifp->lock);
2246                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2247                         in6_ifa_put(ifp);
2248                 }
2249                 return;
2250         }
2251
2252         for_each_netdev(net, dev) {
2253                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2254                 if (in_dev && (dev->flags & IFF_UP)) {
2255                         struct in_ifaddr * ifa;
2256
2257                         int flag = scope;
2258
2259                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2260                                 int plen;
2261
2262                                 addr.s6_addr32[3] = ifa->ifa_local;
2263
2264                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2265                                         continue;
2266                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2267                                         if (idev->dev->flags&IFF_POINTOPOINT)
2268                                                 continue;
2269                                         flag |= IFA_HOST;
2270                                 }
2271                                 if (idev->dev->flags&IFF_POINTOPOINT)
2272                                         plen = 64;
2273                                 else
2274                                         plen = 96;
2275
2276                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2277                                                     IFA_F_PERMANENT);
2278                                 if (!IS_ERR(ifp)) {
2279                                         spin_lock_bh(&ifp->lock);
2280                                         ifp->flags &= ~IFA_F_TENTATIVE;
2281                                         spin_unlock_bh(&ifp->lock);
2282                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2283                                         in6_ifa_put(ifp);
2284                                 }
2285                         }
2286                 }
2287         }
2288 }
2289 #endif
2290
2291 static void init_loopback(struct net_device *dev)
2292 {
2293         struct inet6_dev  *idev;
2294         struct inet6_ifaddr * ifp;
2295
2296         /* ::1 */
2297
2298         ASSERT_RTNL();
2299
2300         if ((idev = ipv6_find_idev(dev)) == NULL) {
2301                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2302                 return;
2303         }
2304
2305         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2306         if (!IS_ERR(ifp)) {
2307                 spin_lock_bh(&ifp->lock);
2308                 ifp->flags &= ~IFA_F_TENTATIVE;
2309                 spin_unlock_bh(&ifp->lock);
2310                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2311                 in6_ifa_put(ifp);
2312         }
2313 }
2314
2315 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2316 {
2317         struct inet6_ifaddr * ifp;
2318         u32 addr_flags = IFA_F_PERMANENT;
2319
2320 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2321         if (idev->cnf.optimistic_dad &&
2322             !ipv6_devconf.forwarding)
2323                 addr_flags |= IFA_F_OPTIMISTIC;
2324 #endif
2325
2326
2327         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2328         if (!IS_ERR(ifp)) {
2329                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2330                 addrconf_dad_start(ifp, 0);
2331                 in6_ifa_put(ifp);
2332         }
2333 }
2334
2335 static void addrconf_dev_config(struct net_device *dev)
2336 {
2337         struct in6_addr addr;
2338         struct inet6_dev    * idev;
2339
2340         ASSERT_RTNL();
2341
2342         if ((dev->type != ARPHRD_ETHER) &&
2343             (dev->type != ARPHRD_FDDI) &&
2344             (dev->type != ARPHRD_IEEE802_TR) &&
2345             (dev->type != ARPHRD_ARCNET) &&
2346             (dev->type != ARPHRD_INFINIBAND)) {
2347                 /* Alas, we support only Ethernet autoconfiguration. */
2348                 return;
2349         }
2350
2351         idev = addrconf_add_dev(dev);
2352         if (idev == NULL)
2353                 return;
2354
2355         memset(&addr, 0, sizeof(struct in6_addr));
2356         addr.s6_addr32[0] = htonl(0xFE800000);
2357
2358         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2359                 addrconf_add_linklocal(idev, &addr);
2360 }
2361
2362 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2363 static void addrconf_sit_config(struct net_device *dev)
2364 {
2365         struct inet6_dev *idev;
2366
2367         ASSERT_RTNL();
2368
2369         /*
2370          * Configure the tunnel with one of our IPv4
2371          * addresses... we should configure all of
2372          * our v4 addrs in the tunnel
2373          */
2374
2375         if ((idev = ipv6_find_idev(dev)) == NULL) {
2376                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2377                 return;
2378         }
2379
2380         if (dev->priv_flags & IFF_ISATAP) {
2381                 struct in6_addr addr;
2382
2383                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2384                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2385                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2386                         addrconf_add_linklocal(idev, &addr);
2387                 return;
2388         }
2389
2390         sit_add_v4_addrs(idev);
2391
2392         if (dev->flags&IFF_POINTOPOINT) {
2393                 addrconf_add_mroute(dev);
2394                 addrconf_add_lroute(dev);
2395         } else
2396                 sit_route_add(dev);
2397 }
2398 #endif
2399
2400 static inline int
2401 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2402 {
2403         struct in6_addr lladdr;
2404
2405         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2406                 addrconf_add_linklocal(idev, &lladdr);
2407                 return 0;
2408         }
2409         return -1;
2410 }
2411
2412 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2413 {
2414         struct net_device *link_dev;
2415         struct net *net = dev_net(idev->dev);
2416
2417         /* first try to inherit the link-local address from the link device */
2418         if (idev->dev->iflink &&
2419             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2420                 if (!ipv6_inherit_linklocal(idev, link_dev))
2421                         return;
2422         }
2423         /* then try to inherit it from any device */
2424         for_each_netdev(net, link_dev) {
2425                 if (!ipv6_inherit_linklocal(idev, link_dev))
2426                         return;
2427         }
2428         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2429 }
2430
2431 /*
2432  * Autoconfigure tunnel with a link-local address so routing protocols,
2433  * DHCPv6, MLD etc. can be run over the virtual link
2434  */
2435
2436 static void addrconf_ip6_tnl_config(struct net_device *dev)
2437 {
2438         struct inet6_dev *idev;
2439
2440         ASSERT_RTNL();
2441
2442         if ((idev = addrconf_add_dev(dev)) == NULL) {
2443                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2444                 return;
2445         }
2446         ip6_tnl_add_linklocal(idev);
2447 }
2448
2449 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2450                            void * data)
2451 {
2452         struct net_device *dev = (struct net_device *) data;
2453         struct inet6_dev *idev = __in6_dev_get(dev);
2454         int run_pending = 0;
2455         int err;
2456
2457         switch(event) {
2458         case NETDEV_REGISTER:
2459                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2460                         idev = ipv6_add_dev(dev);
2461                         if (!idev)
2462                                 return notifier_from_errno(-ENOMEM);
2463                 }
2464                 break;
2465         case NETDEV_UP:
2466         case NETDEV_CHANGE:
2467                 if (dev->flags & IFF_SLAVE)
2468                         break;
2469
2470                 if (event == NETDEV_UP) {
2471                         if (!addrconf_qdisc_ok(dev)) {
2472                                 /* device is not ready yet. */
2473                                 printk(KERN_INFO
2474                                         "ADDRCONF(NETDEV_UP): %s: "
2475                                         "link is not ready\n",
2476                                         dev->name);
2477                                 break;
2478                         }
2479
2480                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2481                                 idev = ipv6_add_dev(dev);
2482
2483                         if (idev)
2484                                 idev->if_flags |= IF_READY;
2485                 } else {
2486                         if (!addrconf_qdisc_ok(dev)) {
2487                                 /* device is still not ready. */
2488                                 break;
2489                         }
2490
2491                         if (idev) {
2492                                 if (idev->if_flags & IF_READY) {
2493                                         /* device is already configured. */
2494                                         break;
2495                                 }
2496                                 idev->if_flags |= IF_READY;
2497                         }
2498
2499                         printk(KERN_INFO
2500                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2501                                         "link becomes ready\n",
2502                                         dev->name);
2503
2504                         run_pending = 1;
2505                 }
2506
2507                 switch(dev->type) {
2508 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2509                 case ARPHRD_SIT:
2510                         addrconf_sit_config(dev);
2511                         break;
2512 #endif
2513                 case ARPHRD_TUNNEL6:
2514                         addrconf_ip6_tnl_config(dev);
2515                         break;
2516                 case ARPHRD_LOOPBACK:
2517                         init_loopback(dev);
2518                         break;
2519
2520                 default:
2521                         addrconf_dev_config(dev);
2522                         break;
2523                 }
2524                 if (idev) {
2525                         if (run_pending)
2526                                 addrconf_dad_run(idev);
2527
2528                         /* If the MTU changed during the interface down, when the
2529                            interface up, the changed MTU must be reflected in the
2530                            idev as well as routers.
2531                          */
2532                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2533                                 rt6_mtu_change(dev, dev->mtu);
2534                                 idev->cnf.mtu6 = dev->mtu;
2535                         }
2536                         idev->tstamp = jiffies;
2537                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2538                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2539                            stop IPv6 on this interface.
2540                          */
2541                         if (dev->mtu < IPV6_MIN_MTU)
2542                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2543                 }
2544                 break;
2545
2546         case NETDEV_CHANGEMTU:
2547                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2548                         rt6_mtu_change(dev, dev->mtu);
2549                         idev->cnf.mtu6 = dev->mtu;
2550                         break;
2551                 }
2552
2553                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2554                         idev = ipv6_add_dev(dev);
2555                         if (idev)
2556                                 break;
2557                 }
2558
2559                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2560
2561         case NETDEV_DOWN:
2562         case NETDEV_UNREGISTER:
2563                 /*
2564                  *      Remove all addresses from this interface.
2565                  */
2566                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2567                 break;
2568
2569         case NETDEV_CHANGENAME:
2570                 if (idev) {
2571                         snmp6_unregister_dev(idev);
2572                         addrconf_sysctl_unregister(idev);
2573                         addrconf_sysctl_register(idev);
2574                         err = snmp6_register_dev(idev);
2575                         if (err)
2576                                 return notifier_from_errno(err);
2577                 }
2578                 break;
2579         }
2580
2581         return NOTIFY_OK;
2582 }
2583
2584 /*
2585  *      addrconf module should be notified of a device going up
2586  */
2587 static struct notifier_block ipv6_dev_notf = {
2588         .notifier_call = addrconf_notify,
2589         .priority = 0
2590 };
2591
2592 static int addrconf_ifdown(struct net_device *dev, int how)
2593 {
2594         struct inet6_dev *idev;
2595         struct inet6_ifaddr *ifa, **bifa;
2596         struct net *net = dev_net(dev);
2597         int i;
2598
2599         ASSERT_RTNL();
2600
2601         if ((dev->flags & IFF_LOOPBACK) && how == 1)
2602                 how = 0;
2603
2604         rt6_ifdown(net, dev);
2605         neigh_ifdown(&nd_tbl, dev);
2606
2607         idev = __in6_dev_get(dev);
2608         if (idev == NULL)
2609                 return -ENODEV;
2610
2611         /* Step 1: remove reference to ipv6 device from parent device.
2612                    Do not dev_put!
2613          */
2614         if (how) {
2615                 idev->dead = 1;
2616
2617                 /* protected by rtnl_lock */
2618                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2619
2620                 /* Step 1.5: remove snmp6 entry */
2621                 snmp6_unregister_dev(idev);
2622
2623         }
2624
2625         /* Step 2: clear hash table */
2626         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2627                 bifa = &inet6_addr_lst[i];
2628
2629                 write_lock_bh(&addrconf_hash_lock);
2630                 while ((ifa = *bifa) != NULL) {
2631                         if (ifa->idev == idev) {
2632                                 *bifa = ifa->lst_next;
2633                                 ifa->lst_next = NULL;
2634                                 addrconf_del_timer(ifa);
2635                                 in6_ifa_put(ifa);
2636                                 continue;
2637                         }
2638                         bifa = &ifa->lst_next;
2639                 }
2640                 write_unlock_bh(&addrconf_hash_lock);
2641         }
2642
2643         write_lock_bh(&idev->lock);
2644
2645         /* Step 3: clear flags for stateless addrconf */
2646         if (!how)
2647                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2648
2649         /* Step 4: clear address list */
2650 #ifdef CONFIG_IPV6_PRIVACY
2651         if (how && del_timer(&idev->regen_timer))
2652                 in6_dev_put(idev);
2653
2654         /* clear tempaddr list */
2655         while ((ifa = idev->tempaddr_list) != NULL) {
2656                 idev->tempaddr_list = ifa->tmp_next;
2657                 ifa->tmp_next = NULL;
2658                 ifa->dead = 1;
2659                 write_unlock_bh(&idev->lock);
2660                 spin_lock_bh(&ifa->lock);
2661
2662                 if (ifa->ifpub) {
2663                         in6_ifa_put(ifa->ifpub);
2664                         ifa->ifpub = NULL;
2665                 }
2666                 spin_unlock_bh(&ifa->lock);
2667                 in6_ifa_put(ifa);
2668                 write_lock_bh(&idev->lock);
2669         }
2670 #endif
2671         while ((ifa = idev->addr_list) != NULL) {
2672                 idev->addr_list = ifa->if_next;
2673                 ifa->if_next = NULL;
2674                 ifa->dead = 1;
2675                 addrconf_del_timer(ifa);
2676                 write_unlock_bh(&idev->lock);
2677
2678                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2679                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2680                 in6_ifa_put(ifa);
2681
2682                 write_lock_bh(&idev->lock);
2683         }
2684         write_unlock_bh(&idev->lock);
2685
2686         /* Step 5: Discard multicast list */
2687
2688         if (how)
2689                 ipv6_mc_destroy_dev(idev);
2690         else
2691                 ipv6_mc_down(idev);
2692
2693         idev->tstamp = jiffies;
2694
2695         /* Shot the device (if unregistered) */
2696
2697         if (how) {
2698                 addrconf_sysctl_unregister(idev);
2699                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2700                 neigh_ifdown(&nd_tbl, dev);
2701                 in6_dev_put(idev);
2702         }
2703         return 0;
2704 }
2705
2706 static void addrconf_rs_timer(unsigned long data)
2707 {
2708         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2709
2710         if (ifp->idev->cnf.forwarding)
2711                 goto out;
2712
2713         if (ifp->idev->if_flags & IF_RA_RCVD) {
2714                 /*
2715                  *      Announcement received after solicitation
2716                  *      was sent
2717                  */
2718                 goto out;
2719         }
2720
2721         spin_lock(&ifp->lock);
2722         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2723                 /* The wait after the last probe can be shorter */
2724                 addrconf_mod_timer(ifp, AC_RS,
2725                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2726                                    ifp->idev->cnf.rtr_solicit_delay :
2727                                    ifp->idev->cnf.rtr_solicit_interval);
2728                 spin_unlock(&ifp->lock);
2729
2730                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2731         } else {
2732                 spin_unlock(&ifp->lock);
2733                 /*
2734                  * Note: we do not support deprecated "all on-link"
2735                  * assumption any longer.
2736                  */
2737                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2738                        ifp->idev->dev->name);
2739         }
2740
2741 out:
2742         in6_ifa_put(ifp);
2743 }
2744
2745 /*
2746  *      Duplicate Address Detection
2747  */
2748 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2749 {
2750         unsigned long rand_num;
2751         struct inet6_dev *idev = ifp->idev;
2752
2753         if (ifp->flags & IFA_F_OPTIMISTIC)
2754                 rand_num = 0;
2755         else
2756                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2757
2758         ifp->probes = idev->cnf.dad_transmits;
2759         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2760 }
2761
2762 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2763 {
2764         struct inet6_dev *idev = ifp->idev;
2765         struct net_device *dev = idev->dev;
2766
2767         addrconf_join_solict(dev, &ifp->addr);
2768
2769         net_srandom(ifp->addr.s6_addr32[3]);
2770
2771         read_lock_bh(&idev->lock);
2772         if (ifp->dead)
2773                 goto out;
2774         spin_lock_bh(&ifp->lock);
2775
2776         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2777             idev->cnf.accept_dad < 1 ||
2778             !(ifp->flags&IFA_F_TENTATIVE) ||
2779             ifp->flags & IFA_F_NODAD) {
2780                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2781                 spin_unlock_bh(&ifp->lock);
2782                 read_unlock_bh(&idev->lock);
2783
2784                 addrconf_dad_completed(ifp);
2785                 return;
2786         }
2787
2788         if (!(idev->if_flags & IF_READY)) {
2789                 spin_unlock_bh(&ifp->lock);
2790                 read_unlock_bh(&idev->lock);
2791                 /*
2792                  * If the defice is not ready:
2793                  * - keep it tentative if it is a permanent address.
2794                  * - otherwise, kill it.
2795                  */
2796                 in6_ifa_hold(ifp);
2797                 addrconf_dad_stop(ifp);
2798                 return;
2799         }
2800
2801         /*
2802          * Optimistic nodes can start receiving
2803          * Frames right away
2804          */
2805         if(ifp->flags & IFA_F_OPTIMISTIC)
2806                 ip6_ins_rt(ifp->rt);
2807
2808         addrconf_dad_kick(ifp);
2809         spin_unlock_bh(&ifp->lock);
2810 out:
2811         read_unlock_bh(&idev->lock);
2812 }
2813
2814 static void addrconf_dad_timer(unsigned long data)
2815 {
2816         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2817         struct inet6_dev *idev = ifp->idev;
2818         struct in6_addr mcaddr;
2819
2820         read_lock_bh(&idev->lock);
2821         if (idev->dead) {
2822                 read_unlock_bh(&idev->lock);
2823                 goto out;
2824         }
2825         if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2826                 read_unlock_bh(&idev->lock);
2827                 addrconf_dad_failure(ifp);
2828                 return;
2829         }
2830         spin_lock_bh(&ifp->lock);
2831         if (ifp->probes == 0) {
2832                 /*
2833                  * DAD was successful
2834                  */
2835
2836                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2837                 spin_unlock_bh(&ifp->lock);
2838                 read_unlock_bh(&idev->lock);
2839
2840                 addrconf_dad_completed(ifp);
2841
2842                 goto out;
2843         }
2844
2845         ifp->probes--;
2846         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2847         spin_unlock_bh(&ifp->lock);
2848         read_unlock_bh(&idev->lock);
2849
2850         /* send a neighbour solicitation for our addr */
2851         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2852         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2853 out:
2854         in6_ifa_put(ifp);
2855 }
2856
2857 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2858 {
2859         struct net_device *     dev = ifp->idev->dev;
2860
2861         /*
2862          *      Configure the address for reception. Now it is valid.
2863          */
2864
2865         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2866
2867         /* If added prefix is link local and forwarding is off,
2868            start sending router solicitations.
2869          */
2870
2871         if (ifp->idev->cnf.forwarding == 0 &&
2872             ifp->idev->cnf.rtr_solicits > 0 &&
2873             (dev->flags&IFF_LOOPBACK) == 0 &&
2874             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2875                 /*
2876                  *      If a host as already performed a random delay
2877                  *      [...] as part of DAD [...] there is no need
2878                  *      to delay again before sending the first RS
2879                  */
2880                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2881
2882                 spin_lock_bh(&ifp->lock);
2883                 ifp->probes = 1;
2884                 ifp->idev->if_flags |= IF_RS_SENT;
2885                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2886                 spin_unlock_bh(&ifp->lock);
2887         }
2888 }
2889
2890 static void addrconf_dad_run(struct inet6_dev *idev) {
2891         struct inet6_ifaddr *ifp;
2892
2893         read_lock_bh(&idev->lock);
2894         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2895                 spin_lock_bh(&ifp->lock);
2896                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2897                         spin_unlock_bh(&ifp->lock);
2898                         continue;
2899                 }
2900                 spin_unlock_bh(&ifp->lock);
2901                 addrconf_dad_kick(ifp);
2902         }
2903         read_unlock_bh(&idev->lock);
2904 }
2905
2906 #ifdef CONFIG_PROC_FS
2907 struct if6_iter_state {
2908         struct seq_net_private p;
2909         int bucket;
2910 };
2911
2912 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2913 {
2914         struct inet6_ifaddr *ifa = NULL;
2915         struct if6_iter_state *state = seq->private;
2916         struct net *net = seq_file_net(seq);
2917
2918         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2919                 ifa = inet6_addr_lst[state->bucket];
2920
2921                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2922                         ifa = ifa->lst_next;
2923                 if (ifa)
2924                         break;
2925         }
2926         return ifa;
2927 }
2928
2929 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2930 {
2931         struct if6_iter_state *state = seq->private;
2932         struct net *net = seq_file_net(seq);
2933
2934         ifa = ifa->lst_next;
2935 try_again:
2936         if (ifa) {
2937                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2938                         ifa = ifa->lst_next;
2939                         goto try_again;
2940                 }
2941         }
2942
2943         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2944                 ifa = inet6_addr_lst[state->bucket];
2945                 goto try_again;
2946         }
2947
2948         return ifa;
2949 }
2950
2951 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2952 {
2953         struct inet6_ifaddr *ifa = if6_get_first(seq);
2954
2955         if (ifa)
2956                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2957                         --pos;
2958         return pos ? NULL : ifa;
2959 }
2960
2961 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2962         __acquires(addrconf_hash_lock)
2963 {
2964         read_lock_bh(&addrconf_hash_lock);
2965         return if6_get_idx(seq, *pos);
2966 }
2967
2968 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2969 {
2970         struct inet6_ifaddr *ifa;
2971
2972         ifa = if6_get_next(seq, v);
2973         ++*pos;
2974         return ifa;
2975 }
2976
2977 static void if6_seq_stop(struct seq_file *seq, void *v)
2978         __releases(addrconf_hash_lock)
2979 {
2980         read_unlock_bh(&addrconf_hash_lock);
2981 }
2982
2983 static int if6_seq_show(struct seq_file *seq, void *v)
2984 {
2985         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2986         seq_printf(seq,
2987                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2988                    NIP6(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         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3642 errout:
3643         if (err < 0)
3644                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3645 }
3646
3647 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3648                                 __s32 *array, int bytes)
3649 {
3650         BUG_ON(bytes < (DEVCONF_MAX * 4));
3651
3652         memset(array, 0, bytes);
3653         array[DEVCONF_FORWARDING] = cnf->forwarding;
3654         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3655         array[DEVCONF_MTU6] = cnf->mtu6;
3656         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3657         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3658         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3659         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3660         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3661         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3662         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3663         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3664 #ifdef CONFIG_IPV6_PRIVACY
3665         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3666         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3667         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3668         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3669         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3670 #endif
3671         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3672         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3673         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3674 #ifdef CONFIG_IPV6_ROUTER_PREF
3675         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3676         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3677 #ifdef CONFIG_IPV6_ROUTE_INFO
3678         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3679 #endif
3680 #endif
3681         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3682         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3683 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3684         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3685 #endif
3686 #ifdef CONFIG_IPV6_MROUTE
3687         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3688 #endif
3689         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3690         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3691 }
3692
3693 static inline size_t inet6_if_nlmsg_size(void)
3694 {
3695         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3696                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3697                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3698                + nla_total_size(4) /* IFLA_MTU */
3699                + nla_total_size(4) /* IFLA_LINK */
3700                + nla_total_size( /* IFLA_PROTINFO */
3701                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3702                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3703                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3704                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3705                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3706                  );
3707 }
3708
3709 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3710                                       int bytes)
3711 {
3712         int i;
3713         int pad = bytes - sizeof(u64) * items;
3714         BUG_ON(pad < 0);
3715
3716         /* Use put_unaligned() because stats may not be aligned for u64. */
3717         put_unaligned(items, &stats[0]);
3718         for (i = 1; i < items; i++)
3719                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3720
3721         memset(&stats[items], 0, pad);
3722 }
3723
3724 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3725                              int bytes)
3726 {
3727         switch(attrtype) {
3728         case IFLA_INET6_STATS:
3729                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3730                 break;
3731         case IFLA_INET6_ICMP6STATS:
3732                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3733                 break;
3734         }
3735 }
3736
3737 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3738                              u32 pid, u32 seq, int event, unsigned int flags)
3739 {
3740         struct net_device *dev = idev->dev;
3741         struct nlattr *nla;
3742         struct ifinfomsg *hdr;
3743         struct nlmsghdr *nlh;
3744         void *protoinfo;
3745         struct ifla_cacheinfo ci;
3746
3747         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3748         if (nlh == NULL)
3749                 return -EMSGSIZE;
3750
3751         hdr = nlmsg_data(nlh);
3752         hdr->ifi_family = AF_INET6;
3753         hdr->__ifi_pad = 0;
3754         hdr->ifi_type = dev->type;
3755         hdr->ifi_index = dev->ifindex;
3756         hdr->ifi_flags = dev_get_flags(dev);
3757         hdr->ifi_change = 0;
3758
3759         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3760
3761         if (dev->addr_len)
3762                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3763
3764         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3765         if (dev->ifindex != dev->iflink)
3766                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3767
3768         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3769         if (protoinfo == NULL)
3770                 goto nla_put_failure;
3771
3772         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3773
3774         ci.max_reasm_len = IPV6_MAXPLEN;
3775         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3776                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3777         ci.reachable_time = idev->nd_parms->reachable_time;
3778         ci.retrans_time = idev->nd_parms->retrans_time;
3779         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3780
3781         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3782         if (nla == NULL)
3783                 goto nla_put_failure;
3784         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3785
3786         /* XXX - MC not implemented */
3787
3788         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3789         if (nla == NULL)
3790                 goto nla_put_failure;
3791         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3792
3793         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3794         if (nla == NULL)
3795                 goto nla_put_failure;
3796         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3797
3798         nla_nest_end(skb, protoinfo);
3799         return nlmsg_end(skb, nlh);
3800
3801 nla_put_failure:
3802         nlmsg_cancel(skb, nlh);
3803         return -EMSGSIZE;
3804 }
3805
3806 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3807 {
3808         struct net *net = sock_net(skb->sk);
3809         int idx, err;
3810         int s_idx = cb->args[0];
3811         struct net_device *dev;
3812         struct inet6_dev *idev;
3813
3814         read_lock(&dev_base_lock);
3815         idx = 0;
3816         for_each_netdev(net, dev) {
3817                 if (idx < s_idx)
3818                         goto cont;
3819                 if ((idev = in6_dev_get(dev)) == NULL)
3820                         goto cont;
3821                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3822                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3823                 in6_dev_put(idev);
3824                 if (err <= 0)
3825                         break;
3826 cont:
3827                 idx++;
3828         }
3829         read_unlock(&dev_base_lock);
3830         cb->args[0] = idx;
3831
3832         return skb->len;
3833 }
3834
3835 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3836 {
3837         struct sk_buff *skb;
3838         struct net *net = dev_net(idev->dev);
3839         int err = -ENOBUFS;
3840
3841         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3842         if (skb == NULL)
3843                 goto errout;
3844
3845         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3846         if (err < 0) {
3847                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3848                 WARN_ON(err == -EMSGSIZE);
3849                 kfree_skb(skb);
3850                 goto errout;
3851         }
3852         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3853 errout:
3854         if (err < 0)
3855                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3856 }
3857
3858 static inline size_t inet6_prefix_nlmsg_size(void)
3859 {
3860         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3861                + nla_total_size(sizeof(struct in6_addr))
3862                + nla_total_size(sizeof(struct prefix_cacheinfo));
3863 }
3864
3865 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3866                              struct prefix_info *pinfo, u32 pid, u32 seq,
3867                              int event, unsigned int flags)
3868 {
3869         struct prefixmsg *pmsg;
3870         struct nlmsghdr *nlh;
3871         struct prefix_cacheinfo ci;
3872
3873         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3874         if (nlh == NULL)
3875                 return -EMSGSIZE;
3876
3877         pmsg = nlmsg_data(nlh);
3878         pmsg->prefix_family = AF_INET6;
3879         pmsg->prefix_pad1 = 0;
3880         pmsg->prefix_pad2 = 0;
3881         pmsg->prefix_ifindex = idev->dev->ifindex;
3882         pmsg->prefix_len = pinfo->prefix_len;
3883         pmsg->prefix_type = pinfo->type;
3884         pmsg->prefix_pad3 = 0;
3885         pmsg->prefix_flags = 0;
3886         if (pinfo->onlink)
3887                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3888         if (pinfo->autoconf)
3889                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3890
3891         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3892
3893         ci.preferred_time = ntohl(pinfo->prefered);
3894         ci.valid_time = ntohl(pinfo->valid);
3895         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3896
3897         return nlmsg_end(skb, nlh);
3898
3899 nla_put_failure:
3900         nlmsg_cancel(skb, nlh);
3901         return -EMSGSIZE;
3902 }
3903
3904 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3905                          struct prefix_info *pinfo)
3906 {
3907         struct sk_buff *skb;
3908         struct net *net = dev_net(idev->dev);
3909         int err = -ENOBUFS;
3910
3911         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3912         if (skb == NULL)
3913                 goto errout;
3914
3915         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3916         if (err < 0) {
3917                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3918                 WARN_ON(err == -EMSGSIZE);
3919                 kfree_skb(skb);
3920                 goto errout;
3921         }
3922         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3923 errout:
3924         if (err < 0)
3925                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3926 }
3927
3928 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3929 {
3930         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3931
3932         switch (event) {
3933         case RTM_NEWADDR:
3934                 /*
3935                  * If the address was optimistic
3936                  * we inserted the route at the start of
3937                  * our DAD process, so we don't need
3938                  * to do it again
3939                  */
3940                 if (!(ifp->rt->rt6i_node))
3941                         ip6_ins_rt(ifp->rt);
3942                 if (ifp->idev->cnf.forwarding)
3943                         addrconf_join_anycast(ifp);
3944                 break;
3945         case RTM_DELADDR:
3946                 if (ifp->idev->cnf.forwarding)
3947                         addrconf_leave_anycast(ifp);
3948                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3949                 dst_hold(&ifp->rt->u.dst);
3950                 if (ip6_del_rt(ifp->rt))
3951                         dst_free(&ifp->rt->u.dst);
3952                 break;
3953         }
3954 }
3955
3956 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3957 {
3958         rcu_read_lock_bh();
3959         if (likely(ifp->idev->dead == 0))
3960                 __ipv6_ifa_notify(event, ifp);
3961         rcu_read_unlock_bh();
3962 }
3963
3964 #ifdef CONFIG_SYSCTL
3965
3966 static
3967 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3968                            void __user *buffer, size_t *lenp, loff_t *ppos)
3969 {
3970         int *valp = ctl->data;
3971         int val = *valp;
3972         int ret;
3973
3974         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3975
3976         if (write)
3977                 addrconf_fixup_forwarding(ctl, valp, val);
3978         return ret;
3979 }
3980
3981 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3982                                             int __user *name, int nlen,
3983                                             void __user *oldval,
3984                                             size_t __user *oldlenp,
3985                                             void __user *newval, size_t newlen)
3986 {
3987         int *valp = table->data;
3988         int val = *valp;
3989         int new;
3990
3991         if (!newval || !newlen)
3992                 return 0;
3993         if (newlen != sizeof(int))
3994                 return -EINVAL;
3995         if (get_user(new, (int __user *)newval))
3996                 return -EFAULT;
3997         if (new == *valp)
3998                 return 0;
3999         if (oldval && oldlenp) {
4000                 size_t len;
4001                 if (get_user(len, oldlenp))
4002                         return -EFAULT;
4003                 if (len) {
4004                         if (len > table->maxlen)
4005                                 len = table->maxlen;
4006                         if (copy_to_user(oldval, valp, len))
4007                                 return -EFAULT;
4008                         if (put_user(len, oldlenp))
4009                                 return -EFAULT;
4010                 }
4011         }
4012
4013         *valp = new;
4014         addrconf_fixup_forwarding(table, valp, val);
4015         return 1;
4016 }
4017
4018 static struct addrconf_sysctl_table
4019 {
4020         struct ctl_table_header *sysctl_header;
4021         ctl_table addrconf_vars[DEVCONF_MAX+1];
4022         char *dev_name;
4023 } addrconf_sysctl __read_mostly = {
4024         .sysctl_header = NULL,
4025         .addrconf_vars = {
4026                 {
4027                         .ctl_name       =       NET_IPV6_FORWARDING,
4028                         .procname       =       "forwarding",
4029                         .data           =       &ipv6_devconf.forwarding,
4030                         .maxlen         =       sizeof(int),
4031                         .mode           =       0644,
4032                         .proc_handler   =       &addrconf_sysctl_forward,
4033                         .strategy       =       &addrconf_sysctl_forward_strategy,
4034                 },
4035                 {
4036                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4037                         .procname       =       "hop_limit",
4038                         .data           =       &ipv6_devconf.hop_limit,
4039                         .maxlen         =       sizeof(int),
4040                         .mode           =       0644,
4041                         .proc_handler   =       proc_dointvec,
4042                 },
4043                 {
4044                         .ctl_name       =       NET_IPV6_MTU,
4045                         .procname       =       "mtu",
4046                         .data           =       &ipv6_devconf.mtu6,
4047                         .maxlen         =       sizeof(int),
4048                         .mode           =       0644,
4049                         .proc_handler   =       &proc_dointvec,
4050                 },
4051                 {
4052                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4053                         .procname       =       "accept_ra",
4054                         .data           =       &ipv6_devconf.accept_ra,
4055                         .maxlen         =       sizeof(int),
4056                         .mode           =       0644,
4057                         .proc_handler   =       &proc_dointvec,
4058                 },
4059                 {
4060                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4061                         .procname       =       "accept_redirects",
4062                         .data           =       &ipv6_devconf.accept_redirects,
4063                         .maxlen         =       sizeof(int),
4064                         .mode           =       0644,
4065                         .proc_handler   =       &proc_dointvec,
4066                 },
4067                 {
4068                         .ctl_name       =       NET_IPV6_AUTOCONF,
4069                         .procname       =       "autoconf",
4070                         .data           =       &ipv6_devconf.autoconf,
4071                         .maxlen         =       sizeof(int),
4072                         .mode           =       0644,
4073                         .proc_handler   =       &proc_dointvec,
4074                 },
4075                 {
4076                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4077                         .procname       =       "dad_transmits",
4078                         .data           =       &ipv6_devconf.dad_transmits,
4079                         .maxlen         =       sizeof(int),
4080                         .mode           =       0644,
4081                         .proc_handler   =       &proc_dointvec,
4082                 },
4083                 {
4084                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4085                         .procname       =       "router_solicitations",
4086                         .data           =       &ipv6_devconf.rtr_solicits,
4087                         .maxlen         =       sizeof(int),
4088                         .mode           =       0644,
4089                         .proc_handler   =       &proc_dointvec,
4090                 },
4091                 {
4092                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4093                         .procname       =       "router_solicitation_interval",
4094                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4095                         .maxlen         =       sizeof(int),
4096                         .mode           =       0644,
4097                         .proc_handler   =       &proc_dointvec_jiffies,
4098                         .strategy       =       &sysctl_jiffies,
4099                 },
4100                 {
4101                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4102                         .procname       =       "router_solicitation_delay",
4103                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4104                         .maxlen         =       sizeof(int),
4105                         .mode           =       0644,
4106                         .proc_handler   =       &proc_dointvec_jiffies,
4107                         .strategy       =       &sysctl_jiffies,
4108                 },
4109                 {
4110                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4111                         .procname       =       "force_mld_version",
4112                         .data           =       &ipv6_devconf.force_mld_version,
4113                         .maxlen         =       sizeof(int),
4114                         .mode           =       0644,
4115                         .proc_handler   =       &proc_dointvec,
4116                 },
4117 #ifdef CONFIG_IPV6_PRIVACY
4118                 {
4119                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4120                         .procname       =       "use_tempaddr",
4121                         .data           =       &ipv6_devconf.use_tempaddr,
4122                         .maxlen         =       sizeof(int),
4123                         .mode           =       0644,
4124                         .proc_handler   =       &proc_dointvec,
4125                 },
4126                 {
4127                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4128                         .procname       =       "temp_valid_lft",
4129                         .data           =       &ipv6_devconf.temp_valid_lft,
4130                         .maxlen         =       sizeof(int),
4131                         .mode           =       0644,
4132                         .proc_handler   =       &proc_dointvec,
4133                 },
4134                 {
4135                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4136                         .procname       =       "temp_prefered_lft",
4137                         .data           =       &ipv6_devconf.temp_prefered_lft,
4138                         .maxlen         =       sizeof(int),
4139                         .mode           =       0644,
4140                         .proc_handler   =       &proc_dointvec,
4141                 },
4142                 {
4143                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4144                         .procname       =       "regen_max_retry",
4145                         .data           =       &ipv6_devconf.regen_max_retry,
4146                         .maxlen         =       sizeof(int),
4147                         .mode           =       0644,
4148                         .proc_handler   =       &proc_dointvec,
4149                 },
4150                 {
4151                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4152                         .procname       =       "max_desync_factor",
4153                         .data           =       &ipv6_devconf.max_desync_factor,
4154                         .maxlen         =       sizeof(int),
4155                         .mode           =       0644,
4156                         .proc_handler   =       &proc_dointvec,
4157                 },
4158 #endif
4159                 {
4160                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4161                         .procname       =       "max_addresses",
4162                         .data           =       &ipv6_devconf.max_addresses,
4163                         .maxlen         =       sizeof(int),
4164                         .mode           =       0644,
4165                         .proc_handler   =       &proc_dointvec,
4166                 },
4167                 {
4168                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4169                         .procname       =       "accept_ra_defrtr",
4170                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4171                         .maxlen         =       sizeof(int),
4172                         .mode           =       0644,
4173                         .proc_handler   =       &proc_dointvec,
4174                 },
4175                 {
4176                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4177                         .procname       =       "accept_ra_pinfo",
4178                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4179                         .maxlen         =       sizeof(int),
4180                         .mode           =       0644,
4181                         .proc_handler   =       &proc_dointvec,
4182                 },
4183 #ifdef CONFIG_IPV6_ROUTER_PREF
4184                 {
4185                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4186                         .procname       =       "accept_ra_rtr_pref",
4187                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4188                         .maxlen         =       sizeof(int),
4189                         .mode           =       0644,
4190                         .proc_handler   =       &proc_dointvec,
4191                 },
4192                 {
4193                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4194                         .procname       =       "router_probe_interval",
4195                         .data           =       &ipv6_devconf.rtr_probe_interval,
4196                         .maxlen         =       sizeof(int),
4197                         .mode           =       0644,
4198                         .proc_handler   =       &proc_dointvec_jiffies,
4199                         .strategy       =       &sysctl_jiffies,
4200                 },
4201 #ifdef CONFIG_IPV6_ROUTE_INFO
4202                 {
4203                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4204                         .procname       =       "accept_ra_rt_info_max_plen",
4205                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4206                         .maxlen         =       sizeof(int),
4207                         .mode           =       0644,
4208                         .proc_handler   =       &proc_dointvec,
4209                 },
4210 #endif
4211 #endif
4212                 {
4213                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4214                         .procname       =       "proxy_ndp",
4215                         .data           =       &ipv6_devconf.proxy_ndp,
4216                         .maxlen         =       sizeof(int),
4217                         .mode           =       0644,
4218                         .proc_handler   =       &proc_dointvec,
4219                 },
4220                 {
4221                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4222                         .procname       =       "accept_source_route",
4223                         .data           =       &ipv6_devconf.accept_source_route,
4224                         .maxlen         =       sizeof(int),
4225                         .mode           =       0644,
4226                         .proc_handler   =       &proc_dointvec,
4227                 },
4228 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4229                 {
4230                         .ctl_name       =       CTL_UNNUMBERED,
4231                         .procname       =       "optimistic_dad",
4232                         .data           =       &ipv6_devconf.optimistic_dad,
4233                         .maxlen         =       sizeof(int),
4234                         .mode           =       0644,
4235                         .proc_handler   =       &proc_dointvec,
4236
4237                 },
4238 #endif
4239 #ifdef CONFIG_IPV6_MROUTE
4240                 {
4241                         .ctl_name       =       CTL_UNNUMBERED,
4242                         .procname       =       "mc_forwarding",
4243                         .data           =       &ipv6_devconf.mc_forwarding,
4244                         .maxlen         =       sizeof(int),
4245                         .mode           =       0644,
4246                         .proc_handler   =       &proc_dointvec,
4247                 },
4248 #endif
4249                 {
4250                         .ctl_name       =       CTL_UNNUMBERED,
4251                         .procname       =       "disable_ipv6",
4252                         .data           =       &ipv6_devconf.disable_ipv6,
4253                         .maxlen         =       sizeof(int),
4254                         .mode           =       0644,
4255                         .proc_handler   =       &proc_dointvec,
4256                 },
4257                 {
4258                         .ctl_name       =       CTL_UNNUMBERED,
4259                         .procname       =       "accept_dad",
4260                         .data           =       &ipv6_devconf.accept_dad,
4261                         .maxlen         =       sizeof(int),
4262                         .mode           =       0644,
4263                         .proc_handler   =       &proc_dointvec,
4264                 },
4265                 {
4266                         .ctl_name       =       0,      /* sentinel */
4267                 }
4268         },
4269 };
4270
4271 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4272                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4273 {
4274         int i;
4275         struct addrconf_sysctl_table *t;
4276
4277 #define ADDRCONF_CTL_PATH_DEV   3
4278
4279         struct ctl_path addrconf_ctl_path[] = {
4280                 { .procname = "net", .ctl_name = CTL_NET, },
4281                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4282                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4283                 { /* to be set */ },
4284                 { },
4285         };
4286
4287
4288         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4289         if (t == NULL)
4290                 goto out;
4291
4292         for (i=0; t->addrconf_vars[i].data; i++) {
4293                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4294                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4295                 t->addrconf_vars[i].extra2 = net;
4296         }
4297
4298         /*
4299          * Make a copy of dev_name, because '.procname' is regarded as const
4300          * by sysctl and we wouldn't want anyone to change it under our feet
4301          * (see SIOCSIFNAME).
4302          */
4303         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4304         if (!t->dev_name)
4305                 goto free;
4306
4307         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4308         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4309
4310         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4311                         t->addrconf_vars);
4312         if (t->sysctl_header == NULL)
4313                 goto free_procname;
4314
4315         p->sysctl = t;
4316         return 0;
4317
4318 free_procname:
4319         kfree(t->dev_name);
4320 free:
4321         kfree(t);
4322 out:
4323         return -ENOBUFS;
4324 }
4325
4326 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4327 {
4328         struct addrconf_sysctl_table *t;
4329
4330         if (p->sysctl == NULL)
4331                 return;
4332
4333         t = p->sysctl;
4334         p->sysctl = NULL;
4335         unregister_sysctl_table(t->sysctl_header);
4336         kfree(t->dev_name);
4337         kfree(t);
4338 }
4339
4340 static void addrconf_sysctl_register(struct inet6_dev *idev)
4341 {
4342         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4343                               NET_IPV6_NEIGH, "ipv6",
4344                               &ndisc_ifinfo_sysctl_change,
4345                               ndisc_ifinfo_sysctl_strategy);
4346         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4347                         idev->dev->ifindex, idev, &idev->cnf);
4348 }
4349
4350 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4351 {
4352         __addrconf_sysctl_unregister(&idev->cnf);
4353         neigh_sysctl_unregister(idev->nd_parms);
4354 }
4355
4356
4357 #endif
4358
4359 static int addrconf_init_net(struct net *net)
4360 {
4361         int err;
4362         struct ipv6_devconf *all, *dflt;
4363
4364         err = -ENOMEM;
4365         all = &ipv6_devconf;
4366         dflt = &ipv6_devconf_dflt;
4367
4368         if (net != &init_net) {
4369                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4370                 if (all == NULL)
4371                         goto err_alloc_all;
4372
4373                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4374                 if (dflt == NULL)
4375                         goto err_alloc_dflt;
4376         }
4377
4378         net->ipv6.devconf_all = all;
4379         net->ipv6.devconf_dflt = dflt;
4380
4381 #ifdef CONFIG_SYSCTL
4382         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4383                         NULL, all);
4384         if (err < 0)
4385                 goto err_reg_all;
4386
4387         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4388                         NULL, dflt);
4389         if (err < 0)
4390                 goto err_reg_dflt;
4391 #endif
4392         return 0;
4393
4394 #ifdef CONFIG_SYSCTL
4395 err_reg_dflt:
4396         __addrconf_sysctl_unregister(all);
4397 err_reg_all:
4398         kfree(dflt);
4399 #endif
4400 err_alloc_dflt:
4401         kfree(all);
4402 err_alloc_all:
4403         return err;
4404 }
4405
4406 static void addrconf_exit_net(struct net *net)
4407 {
4408 #ifdef CONFIG_SYSCTL
4409         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4410         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4411 #endif
4412         if (net != &init_net) {
4413                 kfree(net->ipv6.devconf_dflt);
4414                 kfree(net->ipv6.devconf_all);
4415         }
4416 }
4417
4418 static struct pernet_operations addrconf_ops = {
4419         .init = addrconf_init_net,
4420         .exit = addrconf_exit_net,
4421 };
4422
4423 /*
4424  *      Device notifier
4425  */
4426
4427 int register_inet6addr_notifier(struct notifier_block *nb)
4428 {
4429         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4430 }
4431
4432 EXPORT_SYMBOL(register_inet6addr_notifier);
4433
4434 int unregister_inet6addr_notifier(struct notifier_block *nb)
4435 {
4436         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4437 }
4438
4439 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4440
4441 static void addrconf_net_exit(struct net *net)
4442 {
4443         struct net_device *dev;
4444
4445         rtnl_lock();
4446         /* clean dev list */
4447         for_each_netdev(net, dev) {
4448                 if (__in6_dev_get(dev) == NULL)
4449                         continue;
4450                 addrconf_ifdown(dev, 1);
4451         }
4452         addrconf_ifdown(net->loopback_dev, 2);
4453         rtnl_unlock();
4454 }
4455
4456 static struct pernet_operations addrconf_net_ops = {
4457         .exit = addrconf_net_exit,
4458 };
4459
4460 /*
4461  *      Init / cleanup code
4462  */
4463
4464 int __init addrconf_init(void)
4465 {
4466         int err;
4467
4468         if ((err = ipv6_addr_label_init()) < 0) {
4469                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4470                         err);
4471                 return err;
4472         }
4473
4474         register_pernet_subsys(&addrconf_ops);
4475
4476         /* The addrconf netdev notifier requires that loopback_dev
4477          * has it's ipv6 private information allocated and setup
4478          * before it can bring up and give link-local addresses
4479          * to other devices which are up.
4480          *
4481          * Unfortunately, loopback_dev is not necessarily the first
4482          * entry in the global dev_base list of net devices.  In fact,
4483          * it is likely to be the very last entry on that list.
4484          * So this causes the notifier registry below to try and
4485          * give link-local addresses to all devices besides loopback_dev
4486          * first, then loopback_dev, which cases all the non-loopback_dev
4487          * devices to fail to get a link-local address.
4488          *
4489          * So, as a temporary fix, allocate the ipv6 structure for
4490          * loopback_dev first by hand.
4491          * Longer term, all of the dependencies ipv6 has upon the loopback
4492          * device and it being up should be removed.
4493          */
4494         rtnl_lock();
4495         if (!ipv6_add_dev(init_net.loopback_dev))
4496                 err = -ENOMEM;
4497         rtnl_unlock();
4498         if (err)
4499                 goto errlo;
4500
4501         err = register_pernet_device(&addrconf_net_ops);
4502         if (err)
4503                 return err;
4504
4505         register_netdevice_notifier(&ipv6_dev_notf);
4506
4507         addrconf_verify(0);
4508
4509         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4510         if (err < 0)
4511                 goto errout;
4512
4513         /* Only the first call to __rtnl_register can fail */
4514         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4515         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4516         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4517         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4518         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4519
4520         ipv6_addr_label_rtnl_register();
4521
4522         return 0;
4523 errout:
4524         unregister_netdevice_notifier(&ipv6_dev_notf);
4525 errlo:
4526         unregister_pernet_subsys(&addrconf_ops);
4527
4528         return err;
4529 }
4530
4531 void addrconf_cleanup(void)
4532 {
4533         struct inet6_ifaddr *ifa;
4534         int i;
4535
4536         unregister_netdevice_notifier(&ipv6_dev_notf);
4537         unregister_pernet_device(&addrconf_net_ops);
4538
4539         unregister_pernet_subsys(&addrconf_ops);
4540
4541         rtnl_lock();
4542
4543         /*
4544          *      Check hash table.
4545          */
4546         write_lock_bh(&addrconf_hash_lock);
4547         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4548                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4549                         struct inet6_ifaddr *bifa;
4550
4551                         bifa = ifa;
4552                         ifa = ifa->lst_next;
4553                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4554                         /* Do not free it; something is wrong.
4555                            Now we can investigate it with debugger.
4556                          */
4557                 }
4558         }
4559         write_unlock_bh(&addrconf_hash_lock);
4560
4561         del_timer(&addr_chk_timer);
4562         rtnl_unlock();
4563
4564         unregister_pernet_subsys(&addrconf_net_ops);
4565 }