ipv6: addrconf: Avoid calling netdevice notifiers with RCU read-side lock
[pandora-kernel.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
87
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90 #include <linux/export.h>
91
92 /* Set to 3 to get tracing... */
93 #define ACONF_DEBUG 2
94
95 #if ACONF_DEBUG >= 3
96 #define ADBG(x) printk x
97 #else
98 #define ADBG(x)
99 #endif
100
101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
102
103 static inline u32 cstamp_delta(unsigned long cstamp)
104 {
105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
106 }
107
108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
111
112 #ifdef CONFIG_SYSCTL
113 static void addrconf_sysctl_register(struct inet6_dev *idev);
114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 #else
116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
117 {
118 }
119
120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
121 {
122 }
123 #endif
124
125 #ifdef CONFIG_IPV6_PRIVACY
126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
128 static void ipv6_regen_rndid(unsigned long data);
129 #endif
130
131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
132 static int ipv6_count_addresses(struct inet6_dev *idev);
133
134 /*
135  *      Configured unicast address hash table
136  */
137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
138 static DEFINE_SPINLOCK(addrconf_hash_lock);
139
140 static void addrconf_verify(unsigned long);
141
142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
143 static DEFINE_SPINLOCK(addrconf_verify_lock);
144
145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
147
148 static void addrconf_type_change(struct net_device *dev,
149                                  unsigned long event);
150 static int addrconf_ifdown(struct net_device *dev, int how);
151
152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
153 static void addrconf_dad_timer(unsigned long data);
154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
155 static void addrconf_dad_run(struct inet6_dev *idev);
156 static void addrconf_rs_timer(unsigned long data);
157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
159
160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
161                                 struct prefix_info *pinfo);
162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
163                                struct net_device *dev);
164
165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
166
167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168         .forwarding             = 0,
169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
170         .mtu6                   = IPV6_MIN_MTU,
171         .accept_ra              = 1,
172         .accept_redirects       = 1,
173         .autoconf               = 1,
174         .force_mld_version      = 0,
175         .dad_transmits          = 1,
176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
179 #ifdef CONFIG_IPV6_PRIVACY
180         .use_tempaddr           = 0,
181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
183         .regen_max_retry        = REGEN_MAX_RETRY,
184         .max_desync_factor      = MAX_DESYNC_FACTOR,
185 #endif
186         .max_addresses          = IPV6_MAX_ADDRESSES,
187         .accept_ra_defrtr       = 1,
188         .accept_ra_pinfo        = 1,
189 #ifdef CONFIG_IPV6_ROUTER_PREF
190         .accept_ra_rtr_pref     = 1,
191         .rtr_probe_interval     = 60 * HZ,
192 #ifdef CONFIG_IPV6_ROUTE_INFO
193         .accept_ra_rt_info_max_plen = 0,
194 #endif
195 #endif
196         .proxy_ndp              = 0,
197         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
198         .disable_ipv6           = 0,
199         .accept_dad             = 1,
200 };
201
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203         .forwarding             = 0,
204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
205         .mtu6                   = IPV6_MIN_MTU,
206         .accept_ra              = 1,
207         .accept_redirects       = 1,
208         .autoconf               = 1,
209         .dad_transmits          = 1,
210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
213 #ifdef CONFIG_IPV6_PRIVACY
214         .use_tempaddr           = 0,
215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
217         .regen_max_retry        = REGEN_MAX_RETRY,
218         .max_desync_factor      = MAX_DESYNC_FACTOR,
219 #endif
220         .max_addresses          = IPV6_MAX_ADDRESSES,
221         .accept_ra_defrtr       = 1,
222         .accept_ra_pinfo        = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224         .accept_ra_rtr_pref     = 1,
225         .rtr_probe_interval     = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227         .accept_ra_rt_info_max_plen = 0,
228 #endif
229 #endif
230         .proxy_ndp              = 0,
231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
232         .disable_ipv6           = 0,
233         .accept_dad             = 1,
234 };
235
236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
241
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
244 {
245         return !qdisc_tx_is_noop(dev);
246 }
247
248 /* Check if a route is valid prefix route */
249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
250 {
251         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
252 }
253
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
255 {
256         if (del_timer(&ifp->timer))
257                 __in6_ifa_put(ifp);
258 }
259
260 enum addrconf_timer_t {
261         AC_NONE,
262         AC_DAD,
263         AC_RS,
264 };
265
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267                                enum addrconf_timer_t what,
268                                unsigned long when)
269 {
270         if (!del_timer(&ifp->timer))
271                 in6_ifa_hold(ifp);
272
273         switch (what) {
274         case AC_DAD:
275                 ifp->timer.function = addrconf_dad_timer;
276                 break;
277         case AC_RS:
278                 ifp->timer.function = addrconf_rs_timer;
279                 break;
280         default:
281                 break;
282         }
283         ifp->timer.expires = jiffies + when;
284         add_timer(&ifp->timer);
285 }
286
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
288 {
289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290                           sizeof(struct ipstats_mib),
291                           __alignof__(struct ipstats_mib)) < 0)
292                 goto err_ip;
293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294                                         GFP_KERNEL);
295         if (!idev->stats.icmpv6dev)
296                 goto err_icmp;
297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298                                            GFP_KERNEL);
299         if (!idev->stats.icmpv6msgdev)
300                 goto err_icmpmsg;
301
302         return 0;
303
304 err_icmpmsg:
305         kfree(idev->stats.icmpv6dev);
306 err_icmp:
307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 err_ip:
309         return -ENOMEM;
310 }
311
312 static void snmp6_free_dev(struct inet6_dev *idev)
313 {
314         kfree(idev->stats.icmpv6msgdev);
315         kfree(idev->stats.icmpv6dev);
316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
317 }
318
319 /* Nobody refers to this device, we may destroy it. */
320
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
322 {
323         struct net_device *dev = idev->dev;
324
325         WARN_ON(!list_empty(&idev->addr_list));
326         WARN_ON(idev->mc_list != NULL);
327
328 #ifdef NET_REFCNT_DEBUG
329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
330 #endif
331         dev_put(dev);
332         if (!idev->dead) {
333                 pr_warning("Freeing alive inet6 device %p\n", idev);
334                 return;
335         }
336         snmp6_free_dev(idev);
337         kfree_rcu(idev, rcu);
338 }
339
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
341
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
343 {
344         struct inet6_dev *ndev;
345
346         ASSERT_RTNL();
347
348         if (dev->mtu < IPV6_MIN_MTU)
349                 return NULL;
350
351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
352
353         if (ndev == NULL)
354                 return NULL;
355
356         rwlock_init(&ndev->lock);
357         ndev->dev = dev;
358         INIT_LIST_HEAD(&ndev->addr_list);
359
360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361         ndev->cnf.mtu6 = dev->mtu;
362         ndev->cnf.sysctl = NULL;
363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364         if (ndev->nd_parms == NULL) {
365                 kfree(ndev);
366                 return NULL;
367         }
368         if (ndev->cnf.forwarding)
369                 dev_disable_lro(dev);
370         /* We refer to the device */
371         dev_hold(dev);
372
373         if (snmp6_alloc_dev(ndev) < 0) {
374                 ADBG((KERN_WARNING
375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
376                         __func__, dev->name));
377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378                 dev_put(dev);
379                 kfree(ndev);
380                 return NULL;
381         }
382
383         if (snmp6_register_dev(ndev) < 0) {
384                 ADBG((KERN_WARNING
385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
386                         __func__, dev->name));
387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
388                 ndev->dead = 1;
389                 in6_dev_finish_destroy(ndev);
390                 return NULL;
391         }
392
393         /* One reference from device.  We must do this before
394          * we invoke __ipv6_regen_rndid().
395          */
396         in6_dev_hold(ndev);
397
398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399                 ndev->cnf.accept_dad = -1;
400
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
403                 printk(KERN_INFO
404                        "%s: Disabled Multicast RS\n",
405                        dev->name);
406                 ndev->cnf.rtr_solicits = 0;
407         }
408 #endif
409
410 #ifdef CONFIG_IPV6_PRIVACY
411         INIT_LIST_HEAD(&ndev->tempaddr_list);
412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413         if ((dev->flags&IFF_LOOPBACK) ||
414             dev->type == ARPHRD_TUNNEL ||
415             dev->type == ARPHRD_TUNNEL6 ||
416             dev->type == ARPHRD_SIT ||
417             dev->type == ARPHRD_NONE) {
418                 ndev->cnf.use_tempaddr = -1;
419         } else {
420                 in6_dev_hold(ndev);
421                 ipv6_regen_rndid((unsigned long) ndev);
422         }
423 #endif
424
425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
426                 ndev->if_flags |= IF_READY;
427
428         ipv6_mc_init_dev(ndev);
429         ndev->tstamp = jiffies;
430         addrconf_sysctl_register(ndev);
431         /* protected by rtnl_lock */
432         rcu_assign_pointer(dev->ip6_ptr, ndev);
433
434         /* Join all-node multicast group */
435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
436
437         /* Join all-router multicast group if forwarding is set */
438         if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
440
441         return ndev;
442 }
443
444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
445 {
446         struct inet6_dev *idev;
447
448         ASSERT_RTNL();
449
450         idev = __in6_dev_get(dev);
451         if (!idev) {
452                 idev = ipv6_add_dev(dev);
453                 if (!idev)
454                         return NULL;
455         }
456
457         if (dev->flags&IFF_UP)
458                 ipv6_mc_up(idev);
459         return idev;
460 }
461
462 #ifdef CONFIG_SYSCTL
463 static void dev_forward_change(struct inet6_dev *idev)
464 {
465         struct net_device *dev;
466         struct inet6_ifaddr *ifa;
467
468         if (!idev)
469                 return;
470         dev = idev->dev;
471         if (idev->cnf.forwarding)
472                 dev_disable_lro(dev);
473         if (dev && (dev->flags & IFF_MULTICAST)) {
474                 if (idev->cnf.forwarding)
475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
476                 else
477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
478         }
479
480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
481                 if (ifa->flags&IFA_F_TENTATIVE)
482                         continue;
483                 if (idev->cnf.forwarding)
484                         addrconf_join_anycast(ifa);
485                 else
486                         addrconf_leave_anycast(ifa);
487         }
488 }
489
490
491 static void addrconf_forward_change(struct net *net, __s32 newf)
492 {
493         struct net_device *dev;
494         struct inet6_dev *idev;
495
496         for_each_netdev(net, dev) {
497                 idev = __in6_dev_get(dev);
498                 if (idev) {
499                         int changed = (!idev->cnf.forwarding) ^ (!newf);
500                         idev->cnf.forwarding = newf;
501                         if (changed)
502                                 dev_forward_change(idev);
503                 }
504         }
505 }
506
507 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
508 {
509         struct net *net;
510
511         net = (struct net *)table->extra2;
512         if (p == &net->ipv6.devconf_dflt->forwarding)
513                 return 0;
514
515         if (!rtnl_trylock()) {
516                 /* Restore the original values before restarting */
517                 *p = old;
518                 return restart_syscall();
519         }
520
521         if (p == &net->ipv6.devconf_all->forwarding) {
522                 __s32 newf = net->ipv6.devconf_all->forwarding;
523                 net->ipv6.devconf_dflt->forwarding = newf;
524                 addrconf_forward_change(net, newf);
525         } else if ((!*p) ^ (!old))
526                 dev_forward_change((struct inet6_dev *)table->extra1);
527         rtnl_unlock();
528
529         if (*p)
530                 rt6_purge_dflt_routers(net);
531         return 1;
532 }
533 #endif
534
535 /* Nobody refers to this ifaddr, destroy it */
536 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
537 {
538         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
539
540 #ifdef NET_REFCNT_DEBUG
541         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
542 #endif
543
544         in6_dev_put(ifp->idev);
545
546         if (del_timer(&ifp->timer))
547                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
548
549         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
550                 pr_warning("Freeing alive inet6 address %p\n", ifp);
551                 return;
552         }
553         dst_release(&ifp->rt->dst);
554
555         kfree_rcu(ifp, rcu);
556 }
557
558 static void
559 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
560 {
561         struct list_head *p;
562         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
563
564         /*
565          * Each device address list is sorted in order of scope -
566          * global before linklocal.
567          */
568         list_for_each(p, &idev->addr_list) {
569                 struct inet6_ifaddr *ifa
570                         = list_entry(p, struct inet6_ifaddr, if_list);
571                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
572                         break;
573         }
574
575         list_add_tail(&ifp->if_list, p);
576 }
577
578 static u32 ipv6_addr_hash(const struct in6_addr *addr)
579 {
580         /*
581          * We perform the hash function over the last 64 bits of the address
582          * This will include the IEEE address token on links that support it.
583          */
584         return jhash_2words((__force u32)addr->s6_addr32[2],
585                             (__force u32)addr->s6_addr32[3], 0)
586                 & (IN6_ADDR_HSIZE - 1);
587 }
588
589 /* On success it returns ifp with increased reference count */
590
591 static struct inet6_ifaddr *
592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
593               int scope, u32 flags)
594 {
595         struct inet6_ifaddr *ifa = NULL;
596         struct rt6_info *rt;
597         unsigned int hash;
598         int err = 0;
599         int addr_type = ipv6_addr_type(addr);
600
601         if (addr_type == IPV6_ADDR_ANY ||
602             addr_type & IPV6_ADDR_MULTICAST ||
603             (!(idev->dev->flags & IFF_LOOPBACK) &&
604              addr_type & IPV6_ADDR_LOOPBACK))
605                 return ERR_PTR(-EADDRNOTAVAIL);
606
607         rcu_read_lock_bh();
608         if (idev->dead) {
609                 err = -ENODEV;                  /*XXX*/
610                 goto out2;
611         }
612
613         if (idev->cnf.disable_ipv6) {
614                 err = -EACCES;
615                 goto out2;
616         }
617
618         spin_lock(&addrconf_hash_lock);
619
620         /* Ignore adding duplicate addresses on an interface */
621         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
622                 ADBG(("ipv6_add_addr: already assigned\n"));
623                 err = -EEXIST;
624                 goto out;
625         }
626
627         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
628
629         if (ifa == NULL) {
630                 ADBG(("ipv6_add_addr: malloc failed\n"));
631                 err = -ENOBUFS;
632                 goto out;
633         }
634
635         rt = addrconf_dst_alloc(idev, addr, 0);
636         if (IS_ERR(rt)) {
637                 err = PTR_ERR(rt);
638                 goto out;
639         }
640
641         ipv6_addr_copy(&ifa->addr, addr);
642
643         spin_lock_init(&ifa->lock);
644         spin_lock_init(&ifa->state_lock);
645         init_timer(&ifa->timer);
646         INIT_HLIST_NODE(&ifa->addr_lst);
647         ifa->timer.data = (unsigned long) ifa;
648         ifa->scope = scope;
649         ifa->prefix_len = pfxlen;
650         ifa->flags = flags | IFA_F_TENTATIVE;
651         ifa->cstamp = ifa->tstamp = jiffies;
652
653         ifa->rt = rt;
654
655         /*
656          * part one of RFC 4429, section 3.3
657          * We should not configure an address as
658          * optimistic if we do not yet know the link
659          * layer address of our nexhop router
660          */
661
662         if (dst_get_neighbour_raw(&rt->dst) == NULL)
663                 ifa->flags &= ~IFA_F_OPTIMISTIC;
664
665         ifa->idev = idev;
666         in6_dev_hold(idev);
667         /* For caller */
668         in6_ifa_hold(ifa);
669
670         /* Add to big hash table */
671         hash = ipv6_addr_hash(addr);
672
673         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
674         spin_unlock(&addrconf_hash_lock);
675
676         write_lock(&idev->lock);
677         /* Add to inet6_dev unicast addr list. */
678         ipv6_link_dev_addr(idev, ifa);
679
680 #ifdef CONFIG_IPV6_PRIVACY
681         if (ifa->flags&IFA_F_TEMPORARY) {
682                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
683                 in6_ifa_hold(ifa);
684         }
685 #endif
686
687         in6_ifa_hold(ifa);
688         write_unlock(&idev->lock);
689 out2:
690         rcu_read_unlock_bh();
691
692         if (likely(err == 0))
693                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
694         else {
695                 kfree(ifa);
696                 ifa = ERR_PTR(err);
697         }
698
699         return ifa;
700 out:
701         spin_unlock(&addrconf_hash_lock);
702         goto out2;
703 }
704
705 /* This function wants to get referenced ifp and releases it before return */
706
707 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
708 {
709         struct inet6_ifaddr *ifa, *ifn;
710         struct inet6_dev *idev = ifp->idev;
711         int state;
712         int deleted = 0, onlink = 0;
713         unsigned long expires = jiffies;
714
715         spin_lock_bh(&ifp->state_lock);
716         state = ifp->state;
717         ifp->state = INET6_IFADDR_STATE_DEAD;
718         spin_unlock_bh(&ifp->state_lock);
719
720         if (state == INET6_IFADDR_STATE_DEAD)
721                 goto out;
722
723         spin_lock_bh(&addrconf_hash_lock);
724         hlist_del_init_rcu(&ifp->addr_lst);
725         spin_unlock_bh(&addrconf_hash_lock);
726
727         write_lock_bh(&idev->lock);
728 #ifdef CONFIG_IPV6_PRIVACY
729         if (ifp->flags&IFA_F_TEMPORARY) {
730                 list_del(&ifp->tmp_list);
731                 if (ifp->ifpub) {
732                         in6_ifa_put(ifp->ifpub);
733                         ifp->ifpub = NULL;
734                 }
735                 __in6_ifa_put(ifp);
736         }
737 #endif
738
739         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
740                 if (ifa == ifp) {
741                         list_del_init(&ifp->if_list);
742                         __in6_ifa_put(ifp);
743
744                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
745                                 break;
746                         deleted = 1;
747                         continue;
748                 } else if (ifp->flags & IFA_F_PERMANENT) {
749                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
750                                               ifp->prefix_len)) {
751                                 if (ifa->flags & IFA_F_PERMANENT) {
752                                         onlink = 1;
753                                         if (deleted)
754                                                 break;
755                                 } else {
756                                         unsigned long lifetime;
757
758                                         if (!onlink)
759                                                 onlink = -1;
760
761                                         spin_lock(&ifa->lock);
762
763                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
764                                         /*
765                                          * Note: Because this address is
766                                          * not permanent, lifetime <
767                                          * LONG_MAX / HZ here.
768                                          */
769                                         if (time_before(expires,
770                                                         ifa->tstamp + lifetime * HZ))
771                                                 expires = ifa->tstamp + lifetime * HZ;
772                                         spin_unlock(&ifa->lock);
773                                 }
774                         }
775                 }
776         }
777         write_unlock_bh(&idev->lock);
778
779         addrconf_del_timer(ifp);
780
781         ipv6_ifa_notify(RTM_DELADDR, ifp);
782
783         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
784
785         /*
786          * Purge or update corresponding prefix
787          *
788          * 1) we don't purge prefix here if address was not permanent.
789          *    prefix is managed by its own lifetime.
790          * 2) if there're no addresses, delete prefix.
791          * 3) if there're still other permanent address(es),
792          *    corresponding prefix is still permanent.
793          * 4) otherwise, update prefix lifetime to the
794          *    longest valid lifetime among the corresponding
795          *    addresses on the device.
796          *    Note: subsequent RA will update lifetime.
797          *
798          * --yoshfuji
799          */
800         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
801                 struct in6_addr prefix;
802                 struct rt6_info *rt;
803                 struct net *net = dev_net(ifp->idev->dev);
804                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
805                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
806
807                 if (rt && addrconf_is_prefix_route(rt)) {
808                         if (onlink == 0) {
809                                 ip6_del_rt(rt);
810                                 rt = NULL;
811                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
812                                 rt->rt6i_expires = expires;
813                                 rt->rt6i_flags |= RTF_EXPIRES;
814                         }
815                 }
816                 dst_release(&rt->dst);
817         }
818
819         /* clean up prefsrc entries */
820         rt6_remove_prefsrc(ifp);
821 out:
822         in6_ifa_put(ifp);
823 }
824
825 #ifdef CONFIG_IPV6_PRIVACY
826 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
827 {
828         struct inet6_dev *idev = ifp->idev;
829         struct in6_addr addr, *tmpaddr;
830         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
831         unsigned long regen_advance;
832         int tmp_plen;
833         int ret = 0;
834         int max_addresses;
835         u32 addr_flags;
836         unsigned long now = jiffies;
837
838         write_lock(&idev->lock);
839         if (ift) {
840                 spin_lock_bh(&ift->lock);
841                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
842                 spin_unlock_bh(&ift->lock);
843                 tmpaddr = &addr;
844         } else {
845                 tmpaddr = NULL;
846         }
847 retry:
848         in6_dev_hold(idev);
849         if (idev->cnf.use_tempaddr <= 0) {
850                 write_unlock(&idev->lock);
851                 printk(KERN_INFO
852                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
853                 in6_dev_put(idev);
854                 ret = -1;
855                 goto out;
856         }
857         spin_lock_bh(&ifp->lock);
858         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
859                 idev->cnf.use_tempaddr = -1;    /*XXX*/
860                 spin_unlock_bh(&ifp->lock);
861                 write_unlock(&idev->lock);
862                 printk(KERN_WARNING
863                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
864                 in6_dev_put(idev);
865                 ret = -1;
866                 goto out;
867         }
868         in6_ifa_hold(ifp);
869         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
870         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
871                 spin_unlock_bh(&ifp->lock);
872                 write_unlock(&idev->lock);
873                 printk(KERN_WARNING
874                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
875                 in6_ifa_put(ifp);
876                 in6_dev_put(idev);
877                 ret = -1;
878                 goto out;
879         }
880         memcpy(&addr.s6_addr[8], idev->rndid, 8);
881         age = (now - ifp->tstamp) / HZ;
882         tmp_valid_lft = min_t(__u32,
883                               ifp->valid_lft,
884                               idev->cnf.temp_valid_lft + age);
885         tmp_prefered_lft = min_t(__u32,
886                                  ifp->prefered_lft,
887                                  idev->cnf.temp_prefered_lft + age -
888                                  idev->cnf.max_desync_factor);
889         tmp_plen = ifp->prefix_len;
890         max_addresses = idev->cnf.max_addresses;
891         tmp_tstamp = ifp->tstamp;
892         spin_unlock_bh(&ifp->lock);
893
894         regen_advance = idev->cnf.regen_max_retry *
895                         idev->cnf.dad_transmits *
896                         idev->nd_parms->retrans_time / HZ;
897         write_unlock(&idev->lock);
898
899         /* A temporary address is created only if this calculated Preferred
900          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
901          * an implementation must not create a temporary address with a zero
902          * Preferred Lifetime.
903          */
904         if (tmp_prefered_lft <= regen_advance) {
905                 in6_ifa_put(ifp);
906                 in6_dev_put(idev);
907                 ret = -1;
908                 goto out;
909         }
910
911         addr_flags = IFA_F_TEMPORARY;
912         /* set in addrconf_prefix_rcv() */
913         if (ifp->flags & IFA_F_OPTIMISTIC)
914                 addr_flags |= IFA_F_OPTIMISTIC;
915
916         ift = !max_addresses ||
917               ipv6_count_addresses(idev) < max_addresses ?
918                 ipv6_add_addr(idev, &addr, tmp_plen,
919                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
920                               addr_flags) : NULL;
921         if (!ift || IS_ERR(ift)) {
922                 in6_ifa_put(ifp);
923                 in6_dev_put(idev);
924                 printk(KERN_INFO
925                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
926                 tmpaddr = &addr;
927                 write_lock(&idev->lock);
928                 goto retry;
929         }
930
931         spin_lock_bh(&ift->lock);
932         ift->ifpub = ifp;
933         ift->valid_lft = tmp_valid_lft;
934         ift->prefered_lft = tmp_prefered_lft;
935         ift->cstamp = now;
936         ift->tstamp = tmp_tstamp;
937         spin_unlock_bh(&ift->lock);
938
939         addrconf_dad_start(ift, 0);
940         in6_ifa_put(ift);
941         in6_dev_put(idev);
942 out:
943         return ret;
944 }
945 #endif
946
947 /*
948  *      Choose an appropriate source address (RFC3484)
949  */
950 enum {
951         IPV6_SADDR_RULE_INIT = 0,
952         IPV6_SADDR_RULE_LOCAL,
953         IPV6_SADDR_RULE_SCOPE,
954         IPV6_SADDR_RULE_PREFERRED,
955 #ifdef CONFIG_IPV6_MIP6
956         IPV6_SADDR_RULE_HOA,
957 #endif
958         IPV6_SADDR_RULE_OIF,
959         IPV6_SADDR_RULE_LABEL,
960 #ifdef CONFIG_IPV6_PRIVACY
961         IPV6_SADDR_RULE_PRIVACY,
962 #endif
963         IPV6_SADDR_RULE_ORCHID,
964         IPV6_SADDR_RULE_PREFIX,
965         IPV6_SADDR_RULE_MAX
966 };
967
968 struct ipv6_saddr_score {
969         int                     rule;
970         int                     addr_type;
971         struct inet6_ifaddr     *ifa;
972         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
973         int                     scopedist;
974         int                     matchlen;
975 };
976
977 struct ipv6_saddr_dst {
978         const struct in6_addr *addr;
979         int ifindex;
980         int scope;
981         int label;
982         unsigned int prefs;
983 };
984
985 static inline int ipv6_saddr_preferred(int type)
986 {
987         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
988                 return 1;
989         return 0;
990 }
991
992 static int ipv6_get_saddr_eval(struct net *net,
993                                struct ipv6_saddr_score *score,
994                                struct ipv6_saddr_dst *dst,
995                                int i)
996 {
997         int ret;
998
999         if (i <= score->rule) {
1000                 switch (i) {
1001                 case IPV6_SADDR_RULE_SCOPE:
1002                         ret = score->scopedist;
1003                         break;
1004                 case IPV6_SADDR_RULE_PREFIX:
1005                         ret = score->matchlen;
1006                         break;
1007                 default:
1008                         ret = !!test_bit(i, score->scorebits);
1009                 }
1010                 goto out;
1011         }
1012
1013         switch (i) {
1014         case IPV6_SADDR_RULE_INIT:
1015                 /* Rule 0: remember if hiscore is not ready yet */
1016                 ret = !!score->ifa;
1017                 break;
1018         case IPV6_SADDR_RULE_LOCAL:
1019                 /* Rule 1: Prefer same address */
1020                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1021                 break;
1022         case IPV6_SADDR_RULE_SCOPE:
1023                 /* Rule 2: Prefer appropriate scope
1024                  *
1025                  *      ret
1026                  *       ^
1027                  *    -1 |  d 15
1028                  *    ---+--+-+---> scope
1029                  *       |
1030                  *       |             d is scope of the destination.
1031                  *  B-d  |  \
1032                  *       |   \      <- smaller scope is better if
1033                  *  B-15 |    \        if scope is enough for destinaion.
1034                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1035                  * d-C-1 | /
1036                  *       |/         <- greater is better
1037                  *   -C  /             if scope is not enough for destination.
1038                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1039                  *
1040                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1041                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1042                  * Assume B = 0 and we get C > 29.
1043                  */
1044                 ret = __ipv6_addr_src_scope(score->addr_type);
1045                 if (ret >= dst->scope)
1046                         ret = -ret;
1047                 else
1048                         ret -= 128;     /* 30 is enough */
1049                 score->scopedist = ret;
1050                 break;
1051         case IPV6_SADDR_RULE_PREFERRED:
1052                 /* Rule 3: Avoid deprecated and optimistic addresses */
1053                 ret = ipv6_saddr_preferred(score->addr_type) ||
1054                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1055                 break;
1056 #ifdef CONFIG_IPV6_MIP6
1057         case IPV6_SADDR_RULE_HOA:
1058             {
1059                 /* Rule 4: Prefer home address */
1060                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1061                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1062                 break;
1063             }
1064 #endif
1065         case IPV6_SADDR_RULE_OIF:
1066                 /* Rule 5: Prefer outgoing interface */
1067                 ret = (!dst->ifindex ||
1068                        dst->ifindex == score->ifa->idev->dev->ifindex);
1069                 break;
1070         case IPV6_SADDR_RULE_LABEL:
1071                 /* Rule 6: Prefer matching label */
1072                 ret = ipv6_addr_label(net,
1073                                       &score->ifa->addr, score->addr_type,
1074                                       score->ifa->idev->dev->ifindex) == dst->label;
1075                 break;
1076 #ifdef CONFIG_IPV6_PRIVACY
1077         case IPV6_SADDR_RULE_PRIVACY:
1078             {
1079                 /* Rule 7: Prefer public address
1080                  * Note: prefer temporary address if use_tempaddr >= 2
1081                  */
1082                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1083                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1084                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1085                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1086                 break;
1087             }
1088 #endif
1089         case IPV6_SADDR_RULE_ORCHID:
1090                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1091                  *          non-ORCHID vs non-ORCHID
1092                  */
1093                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1094                         ipv6_addr_orchid(dst->addr));
1095                 break;
1096         case IPV6_SADDR_RULE_PREFIX:
1097                 /* Rule 8: Use longest matching prefix */
1098                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1099                                                        dst->addr);
1100                 break;
1101         default:
1102                 ret = 0;
1103         }
1104
1105         if (ret)
1106                 __set_bit(i, score->scorebits);
1107         score->rule = i;
1108 out:
1109         return ret;
1110 }
1111
1112 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1113                        const struct in6_addr *daddr, unsigned int prefs,
1114                        struct in6_addr *saddr)
1115 {
1116         struct ipv6_saddr_score scores[2],
1117                                 *score = &scores[0], *hiscore = &scores[1];
1118         struct ipv6_saddr_dst dst;
1119         struct net_device *dev;
1120         int dst_type;
1121
1122         dst_type = __ipv6_addr_type(daddr);
1123         dst.addr = daddr;
1124         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1125         dst.scope = __ipv6_addr_src_scope(dst_type);
1126         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1127         dst.prefs = prefs;
1128
1129         hiscore->rule = -1;
1130         hiscore->ifa = NULL;
1131
1132         rcu_read_lock();
1133
1134         for_each_netdev_rcu(net, dev) {
1135                 struct inet6_dev *idev;
1136
1137                 /* Candidate Source Address (section 4)
1138                  *  - multicast and link-local destination address,
1139                  *    the set of candidate source address MUST only
1140                  *    include addresses assigned to interfaces
1141                  *    belonging to the same link as the outgoing
1142                  *    interface.
1143                  * (- For site-local destination addresses, the
1144                  *    set of candidate source addresses MUST only
1145                  *    include addresses assigned to interfaces
1146                  *    belonging to the same site as the outgoing
1147                  *    interface.)
1148                  */
1149                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1150                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1151                     dst.ifindex && dev->ifindex != dst.ifindex)
1152                         continue;
1153
1154                 idev = __in6_dev_get(dev);
1155                 if (!idev)
1156                         continue;
1157
1158                 read_lock_bh(&idev->lock);
1159                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1160                         int i;
1161
1162                         /*
1163                          * - Tentative Address (RFC2462 section 5.4)
1164                          *  - A tentative address is not considered
1165                          *    "assigned to an interface" in the traditional
1166                          *    sense, unless it is also flagged as optimistic.
1167                          * - Candidate Source Address (section 4)
1168                          *  - In any case, anycast addresses, multicast
1169                          *    addresses, and the unspecified address MUST
1170                          *    NOT be included in a candidate set.
1171                          */
1172                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1173                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1174                                 continue;
1175
1176                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1177
1178                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1179                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1180                                 LIMIT_NETDEBUG(KERN_DEBUG
1181                                                "ADDRCONF: unspecified / multicast address "
1182                                                "assigned as unicast address on %s",
1183                                                dev->name);
1184                                 continue;
1185                         }
1186
1187                         score->rule = -1;
1188                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1189
1190                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1191                                 int minihiscore, miniscore;
1192
1193                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1194                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1195
1196                                 if (minihiscore > miniscore) {
1197                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1198                                             score->scopedist > 0) {
1199                                                 /*
1200                                                  * special case:
1201                                                  * each remaining entry
1202                                                  * has too small (not enough)
1203                                                  * scope, because ifa entries
1204                                                  * are sorted by their scope
1205                                                  * values.
1206                                                  */
1207                                                 goto try_nextdev;
1208                                         }
1209                                         break;
1210                                 } else if (minihiscore < miniscore) {
1211                                         if (hiscore->ifa)
1212                                                 in6_ifa_put(hiscore->ifa);
1213
1214                                         in6_ifa_hold(score->ifa);
1215
1216                                         swap(hiscore, score);
1217
1218                                         /* restore our iterator */
1219                                         score->ifa = hiscore->ifa;
1220
1221                                         break;
1222                                 }
1223                         }
1224                 }
1225 try_nextdev:
1226                 read_unlock_bh(&idev->lock);
1227         }
1228         rcu_read_unlock();
1229
1230         if (!hiscore->ifa)
1231                 return -EADDRNOTAVAIL;
1232
1233         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1234         in6_ifa_put(hiscore->ifa);
1235         return 0;
1236 }
1237 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1238
1239 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1240                     unsigned char banned_flags)
1241 {
1242         struct inet6_dev *idev;
1243         int err = -EADDRNOTAVAIL;
1244
1245         rcu_read_lock();
1246         idev = __in6_dev_get(dev);
1247         if (idev) {
1248                 struct inet6_ifaddr *ifp;
1249
1250                 read_lock_bh(&idev->lock);
1251                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1252                         if (ifp->scope == IFA_LINK &&
1253                             !(ifp->flags & banned_flags)) {
1254                                 ipv6_addr_copy(addr, &ifp->addr);
1255                                 err = 0;
1256                                 break;
1257                         }
1258                 }
1259                 read_unlock_bh(&idev->lock);
1260         }
1261         rcu_read_unlock();
1262         return err;
1263 }
1264
1265 static int ipv6_count_addresses(struct inet6_dev *idev)
1266 {
1267         int cnt = 0;
1268         struct inet6_ifaddr *ifp;
1269
1270         read_lock_bh(&idev->lock);
1271         list_for_each_entry(ifp, &idev->addr_list, if_list)
1272                 cnt++;
1273         read_unlock_bh(&idev->lock);
1274         return cnt;
1275 }
1276
1277 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1278                   struct net_device *dev, int strict)
1279 {
1280         struct inet6_ifaddr *ifp;
1281         struct hlist_node *node;
1282         unsigned int hash = ipv6_addr_hash(addr);
1283
1284         rcu_read_lock_bh();
1285         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
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                     (dev == NULL || ifp->idev->dev == dev ||
1291                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1292                         rcu_read_unlock_bh();
1293                         return 1;
1294                 }
1295         }
1296
1297         rcu_read_unlock_bh();
1298         return 0;
1299 }
1300 EXPORT_SYMBOL(ipv6_chk_addr);
1301
1302 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1303                                struct net_device *dev)
1304 {
1305         unsigned int hash = ipv6_addr_hash(addr);
1306         struct inet6_ifaddr *ifp;
1307         struct hlist_node *node;
1308
1309         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1310                 if (!net_eq(dev_net(ifp->idev->dev), net))
1311                         continue;
1312                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1313                         if (dev == NULL || ifp->idev->dev == dev)
1314                                 return true;
1315                 }
1316         }
1317         return false;
1318 }
1319
1320 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1321 {
1322         struct inet6_dev *idev;
1323         struct inet6_ifaddr *ifa;
1324         int     onlink;
1325
1326         onlink = 0;
1327         rcu_read_lock();
1328         idev = __in6_dev_get(dev);
1329         if (idev) {
1330                 read_lock_bh(&idev->lock);
1331                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1332                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1333                                                    ifa->prefix_len);
1334                         if (onlink)
1335                                 break;
1336                 }
1337                 read_unlock_bh(&idev->lock);
1338         }
1339         rcu_read_unlock();
1340         return onlink;
1341 }
1342
1343 EXPORT_SYMBOL(ipv6_chk_prefix);
1344
1345 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1346                                      struct net_device *dev, int strict)
1347 {
1348         struct inet6_ifaddr *ifp, *result = NULL;
1349         unsigned int hash = ipv6_addr_hash(addr);
1350         struct hlist_node *node;
1351
1352         rcu_read_lock_bh();
1353         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1354                 if (!net_eq(dev_net(ifp->idev->dev), net))
1355                         continue;
1356                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1357                         if (dev == NULL || ifp->idev->dev == dev ||
1358                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1359                                 result = ifp;
1360                                 in6_ifa_hold(ifp);
1361                                 break;
1362                         }
1363                 }
1364         }
1365         rcu_read_unlock_bh();
1366
1367         return result;
1368 }
1369
1370 /* Gets referenced address, destroys ifaddr */
1371
1372 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1373 {
1374         if (ifp->flags&IFA_F_PERMANENT) {
1375                 spin_lock_bh(&ifp->lock);
1376                 addrconf_del_timer(ifp);
1377                 ifp->flags |= IFA_F_TENTATIVE;
1378                 if (dad_failed)
1379                         ifp->flags |= IFA_F_DADFAILED;
1380                 spin_unlock_bh(&ifp->lock);
1381                 if (dad_failed)
1382                         ipv6_ifa_notify(0, ifp);
1383                 in6_ifa_put(ifp);
1384 #ifdef CONFIG_IPV6_PRIVACY
1385         } else if (ifp->flags&IFA_F_TEMPORARY) {
1386                 struct inet6_ifaddr *ifpub;
1387                 spin_lock_bh(&ifp->lock);
1388                 ifpub = ifp->ifpub;
1389                 if (ifpub) {
1390                         in6_ifa_hold(ifpub);
1391                         spin_unlock_bh(&ifp->lock);
1392                         ipv6_create_tempaddr(ifpub, ifp);
1393                         in6_ifa_put(ifpub);
1394                 } else {
1395                         spin_unlock_bh(&ifp->lock);
1396                 }
1397                 ipv6_del_addr(ifp);
1398 #endif
1399         } else
1400                 ipv6_del_addr(ifp);
1401 }
1402
1403 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1404 {
1405         int err = -ENOENT;
1406
1407         spin_lock(&ifp->state_lock);
1408         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1409                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1410                 err = 0;
1411         }
1412         spin_unlock(&ifp->state_lock);
1413
1414         return err;
1415 }
1416
1417 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1418 {
1419         struct inet6_dev *idev = ifp->idev;
1420
1421         if (addrconf_dad_end(ifp)) {
1422                 in6_ifa_put(ifp);
1423                 return;
1424         }
1425
1426         if (net_ratelimit())
1427                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1428                         ifp->idev->dev->name, &ifp->addr);
1429
1430         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1431                 struct in6_addr addr;
1432
1433                 addr.s6_addr32[0] = htonl(0xfe800000);
1434                 addr.s6_addr32[1] = 0;
1435
1436                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1437                     ipv6_addr_equal(&ifp->addr, &addr)) {
1438                         /* DAD failed for link-local based on MAC address */
1439                         idev->cnf.disable_ipv6 = 1;
1440
1441                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1442                                 ifp->idev->dev->name);
1443                 }
1444         }
1445
1446         addrconf_dad_stop(ifp, 1);
1447 }
1448
1449 /* Join to solicited addr multicast group. */
1450
1451 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1452 {
1453         struct in6_addr maddr;
1454
1455         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1456                 return;
1457
1458         addrconf_addr_solict_mult(addr, &maddr);
1459         ipv6_dev_mc_inc(dev, &maddr);
1460 }
1461
1462 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1463 {
1464         struct in6_addr maddr;
1465
1466         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1467                 return;
1468
1469         addrconf_addr_solict_mult(addr, &maddr);
1470         __ipv6_dev_mc_dec(idev, &maddr);
1471 }
1472
1473 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1474 {
1475         struct in6_addr addr;
1476         if (ifp->prefix_len == 127) /* RFC 6164 */
1477                 return;
1478         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1479         if (ipv6_addr_any(&addr))
1480                 return;
1481         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1482 }
1483
1484 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1485 {
1486         struct in6_addr addr;
1487         if (ifp->prefix_len == 127) /* RFC 6164 */
1488                 return;
1489         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490         if (ipv6_addr_any(&addr))
1491                 return;
1492         __ipv6_dev_ac_dec(ifp->idev, &addr);
1493 }
1494
1495 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1496 {
1497         if (dev->addr_len != ETH_ALEN)
1498                 return -1;
1499         memcpy(eui, dev->dev_addr, 3);
1500         memcpy(eui + 5, dev->dev_addr + 3, 3);
1501
1502         /*
1503          * The zSeries OSA network cards can be shared among various
1504          * OS instances, but the OSA cards have only one MAC address.
1505          * This leads to duplicate address conflicts in conjunction
1506          * with IPv6 if more than one instance uses the same card.
1507          *
1508          * The driver for these cards can deliver a unique 16-bit
1509          * identifier for each instance sharing the same card.  It is
1510          * placed instead of 0xFFFE in the interface identifier.  The
1511          * "u" bit of the interface identifier is not inverted in this
1512          * case.  Hence the resulting interface identifier has local
1513          * scope according to RFC2373.
1514          */
1515         if (dev->dev_id) {
1516                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1517                 eui[4] = dev->dev_id & 0xFF;
1518         } else {
1519                 eui[3] = 0xFF;
1520                 eui[4] = 0xFE;
1521                 eui[0] ^= 2;
1522         }
1523         return 0;
1524 }
1525
1526 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1527 {
1528         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1529         if (dev->addr_len != ARCNET_ALEN)
1530                 return -1;
1531         memset(eui, 0, 7);
1532         eui[7] = *(u8*)dev->dev_addr;
1533         return 0;
1534 }
1535
1536 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1537 {
1538         if (dev->addr_len != INFINIBAND_ALEN)
1539                 return -1;
1540         memcpy(eui, dev->dev_addr + 12, 8);
1541         eui[0] |= 2;
1542         return 0;
1543 }
1544
1545 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1546 {
1547         if (addr == 0)
1548                 return -1;
1549         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1550                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1551                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1552                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1553                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1554                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1555         eui[1] = 0;
1556         eui[2] = 0x5E;
1557         eui[3] = 0xFE;
1558         memcpy(eui + 4, &addr, 4);
1559         return 0;
1560 }
1561
1562 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1563 {
1564         if (dev->priv_flags & IFF_ISATAP)
1565                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1566         return -1;
1567 }
1568
1569 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1570 {
1571         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1572 }
1573
1574 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1575 {
1576         switch (dev->type) {
1577         case ARPHRD_ETHER:
1578         case ARPHRD_FDDI:
1579         case ARPHRD_IEEE802_TR:
1580                 return addrconf_ifid_eui48(eui, dev);
1581         case ARPHRD_ARCNET:
1582                 return addrconf_ifid_arcnet(eui, dev);
1583         case ARPHRD_INFINIBAND:
1584                 return addrconf_ifid_infiniband(eui, dev);
1585         case ARPHRD_SIT:
1586                 return addrconf_ifid_sit(eui, dev);
1587         case ARPHRD_IPGRE:
1588                 return addrconf_ifid_gre(eui, dev);
1589         }
1590         return -1;
1591 }
1592
1593 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1594 {
1595         int err = -1;
1596         struct inet6_ifaddr *ifp;
1597
1598         read_lock_bh(&idev->lock);
1599         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1600                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1601                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1602                         err = 0;
1603                         break;
1604                 }
1605         }
1606         read_unlock_bh(&idev->lock);
1607         return err;
1608 }
1609
1610 #ifdef CONFIG_IPV6_PRIVACY
1611 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1612 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1613 {
1614 regen:
1615         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1616         idev->rndid[0] &= ~0x02;
1617
1618         /*
1619          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1620          * check if generated address is not inappropriate
1621          *
1622          *  - Reserved subnet anycast (RFC 2526)
1623          *      11111101 11....11 1xxxxxxx
1624          *  - ISATAP (RFC4214) 6.1
1625          *      00-00-5E-FE-xx-xx-xx-xx
1626          *  - value 0
1627          *  - XXX: already assigned to an address on the device
1628          */
1629         if (idev->rndid[0] == 0xfd &&
1630             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1631             (idev->rndid[7]&0x80))
1632                 goto regen;
1633         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1634                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1635                         goto regen;
1636                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1637                         goto regen;
1638         }
1639
1640         return 0;
1641 }
1642
1643 static void ipv6_regen_rndid(unsigned long data)
1644 {
1645         struct inet6_dev *idev = (struct inet6_dev *) data;
1646         unsigned long expires;
1647
1648         rcu_read_lock_bh();
1649         write_lock_bh(&idev->lock);
1650
1651         if (idev->dead)
1652                 goto out;
1653
1654         if (__ipv6_regen_rndid(idev) < 0)
1655                 goto out;
1656
1657         expires = jiffies +
1658                 idev->cnf.temp_prefered_lft * HZ -
1659                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1660                 idev->cnf.max_desync_factor * HZ;
1661         if (time_before(expires, jiffies)) {
1662                 printk(KERN_WARNING
1663                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1664                         idev->dev->name);
1665                 goto out;
1666         }
1667
1668         if (!mod_timer(&idev->regen_timer, expires))
1669                 in6_dev_hold(idev);
1670
1671 out:
1672         write_unlock_bh(&idev->lock);
1673         rcu_read_unlock_bh();
1674         in6_dev_put(idev);
1675 }
1676
1677 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1678         int ret = 0;
1679
1680         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1681                 ret = __ipv6_regen_rndid(idev);
1682         return ret;
1683 }
1684 #endif
1685
1686 /*
1687  *      Add prefix route.
1688  */
1689
1690 static void
1691 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1692                       unsigned long expires, u32 flags)
1693 {
1694         struct fib6_config cfg = {
1695                 .fc_table = RT6_TABLE_PREFIX,
1696                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1697                 .fc_ifindex = dev->ifindex,
1698                 .fc_expires = expires,
1699                 .fc_dst_len = plen,
1700                 .fc_flags = RTF_UP | flags,
1701                 .fc_nlinfo.nl_net = dev_net(dev),
1702                 .fc_protocol = RTPROT_KERNEL,
1703         };
1704
1705         ipv6_addr_copy(&cfg.fc_dst, pfx);
1706
1707         /* Prevent useless cloning on PtP SIT.
1708            This thing is done here expecting that the whole
1709            class of non-broadcast devices need not cloning.
1710          */
1711 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1712         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1713                 cfg.fc_flags |= RTF_NONEXTHOP;
1714 #endif
1715
1716         ip6_route_add(&cfg);
1717 }
1718
1719
1720 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1721                                                   int plen,
1722                                                   const struct net_device *dev,
1723                                                   u32 flags, u32 noflags)
1724 {
1725         struct fib6_node *fn;
1726         struct rt6_info *rt = NULL;
1727         struct fib6_table *table;
1728
1729         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1730         if (table == NULL)
1731                 return NULL;
1732
1733         write_lock_bh(&table->tb6_lock);
1734         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1735         if (!fn)
1736                 goto out;
1737         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1738                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1739                         continue;
1740                 if ((rt->rt6i_flags & flags) != flags)
1741                         continue;
1742                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1743                         continue;
1744                 dst_hold(&rt->dst);
1745                 break;
1746         }
1747 out:
1748         write_unlock_bh(&table->tb6_lock);
1749         return rt;
1750 }
1751
1752
1753 /* Create "default" multicast route to the interface */
1754
1755 static void addrconf_add_mroute(struct net_device *dev)
1756 {
1757         struct fib6_config cfg = {
1758                 .fc_table = RT6_TABLE_LOCAL,
1759                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1760                 .fc_ifindex = dev->ifindex,
1761                 .fc_dst_len = 8,
1762                 .fc_flags = RTF_UP,
1763                 .fc_nlinfo.nl_net = dev_net(dev),
1764         };
1765
1766         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1767
1768         ip6_route_add(&cfg);
1769 }
1770
1771 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1772 static void sit_route_add(struct net_device *dev)
1773 {
1774         struct fib6_config cfg = {
1775                 .fc_table = RT6_TABLE_MAIN,
1776                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1777                 .fc_ifindex = dev->ifindex,
1778                 .fc_dst_len = 96,
1779                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1780                 .fc_nlinfo.nl_net = dev_net(dev),
1781         };
1782
1783         /* prefix length - 96 bits "::d.d.d.d" */
1784         ip6_route_add(&cfg);
1785 }
1786 #endif
1787
1788 static void addrconf_add_lroute(struct net_device *dev)
1789 {
1790         struct in6_addr addr;
1791
1792         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1793         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1794 }
1795
1796 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1797 {
1798         struct inet6_dev *idev;
1799
1800         ASSERT_RTNL();
1801
1802         idev = ipv6_find_idev(dev);
1803         if (!idev)
1804                 return ERR_PTR(-ENOBUFS);
1805
1806         if (idev->cnf.disable_ipv6)
1807                 return ERR_PTR(-EACCES);
1808
1809         /* Add default multicast route */
1810         if (!(dev->flags & IFF_LOOPBACK))
1811                 addrconf_add_mroute(dev);
1812
1813         /* Add link local route */
1814         addrconf_add_lroute(dev);
1815         return idev;
1816 }
1817
1818 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1819 {
1820         struct prefix_info *pinfo;
1821         __u32 valid_lft;
1822         __u32 prefered_lft;
1823         int addr_type;
1824         struct inet6_dev *in6_dev;
1825         struct net *net = dev_net(dev);
1826
1827         pinfo = (struct prefix_info *) opt;
1828
1829         if (len < sizeof(struct prefix_info)) {
1830                 ADBG(("addrconf: prefix option too short\n"));
1831                 return;
1832         }
1833
1834         /*
1835          *      Validation checks ([ADDRCONF], page 19)
1836          */
1837
1838         addr_type = ipv6_addr_type(&pinfo->prefix);
1839
1840         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1841                 return;
1842
1843         valid_lft = ntohl(pinfo->valid);
1844         prefered_lft = ntohl(pinfo->prefered);
1845
1846         if (prefered_lft > valid_lft) {
1847                 if (net_ratelimit())
1848                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1849                 return;
1850         }
1851
1852         in6_dev = in6_dev_get(dev);
1853
1854         if (in6_dev == NULL) {
1855                 if (net_ratelimit())
1856                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1857                 return;
1858         }
1859
1860         /*
1861          *      Two things going on here:
1862          *      1) Add routes for on-link prefixes
1863          *      2) Configure prefixes with the auto flag set
1864          */
1865
1866         if (pinfo->onlink) {
1867                 struct rt6_info *rt;
1868                 unsigned long rt_expires;
1869
1870                 /* Avoid arithmetic overflow. Really, we could
1871                  * save rt_expires in seconds, likely valid_lft,
1872                  * but it would require division in fib gc, that it
1873                  * not good.
1874                  */
1875                 if (HZ > USER_HZ)
1876                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1877                 else
1878                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1879
1880                 if (addrconf_finite_timeout(rt_expires))
1881                         rt_expires *= HZ;
1882
1883                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1884                                                pinfo->prefix_len,
1885                                                dev,
1886                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1887                                                RTF_GATEWAY | RTF_DEFAULT);
1888
1889                 if (rt) {
1890                         /* Autoconf prefix route */
1891                         if (valid_lft == 0) {
1892                                 ip6_del_rt(rt);
1893                                 rt = NULL;
1894                         } else if (addrconf_finite_timeout(rt_expires)) {
1895                                 /* not infinity */
1896                                 rt->rt6i_expires = jiffies + rt_expires;
1897                                 rt->rt6i_flags |= RTF_EXPIRES;
1898                         } else {
1899                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1900                                 rt->rt6i_expires = 0;
1901                         }
1902                 } else if (valid_lft) {
1903                         clock_t expires = 0;
1904                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1905                         if (addrconf_finite_timeout(rt_expires)) {
1906                                 /* not infinity */
1907                                 flags |= RTF_EXPIRES;
1908                                 expires = jiffies_to_clock_t(rt_expires);
1909                         }
1910                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1911                                               dev, expires, flags);
1912                 }
1913                 if (rt)
1914                         dst_release(&rt->dst);
1915         }
1916
1917         /* Try to figure out our local address for this prefix */
1918
1919         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1920                 struct inet6_ifaddr * ifp;
1921                 struct in6_addr addr;
1922                 int create = 0, update_lft = 0;
1923
1924                 if (pinfo->prefix_len == 64) {
1925                         memcpy(&addr, &pinfo->prefix, 8);
1926                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1927                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1928                                 in6_dev_put(in6_dev);
1929                                 return;
1930                         }
1931                         goto ok;
1932                 }
1933                 if (net_ratelimit())
1934                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1935                                pinfo->prefix_len);
1936                 in6_dev_put(in6_dev);
1937                 return;
1938
1939 ok:
1940
1941                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1942
1943                 if (ifp == NULL && valid_lft) {
1944                         int max_addresses = in6_dev->cnf.max_addresses;
1945                         u32 addr_flags = 0;
1946
1947 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1948                         if (in6_dev->cnf.optimistic_dad &&
1949                             !net->ipv6.devconf_all->forwarding)
1950                                 addr_flags = IFA_F_OPTIMISTIC;
1951 #endif
1952
1953                         /* Do not allow to create too much of autoconfigured
1954                          * addresses; this would be too easy way to crash kernel.
1955                          */
1956                         if (!max_addresses ||
1957                             ipv6_count_addresses(in6_dev) < max_addresses)
1958                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1959                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1960                                                     addr_flags);
1961
1962                         if (!ifp || IS_ERR(ifp)) {
1963                                 in6_dev_put(in6_dev);
1964                                 return;
1965                         }
1966
1967                         update_lft = create = 1;
1968                         ifp->cstamp = jiffies;
1969                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1970                 }
1971
1972                 if (ifp) {
1973                         int flags;
1974                         unsigned long now;
1975 #ifdef CONFIG_IPV6_PRIVACY
1976                         struct inet6_ifaddr *ift;
1977 #endif
1978                         u32 stored_lft;
1979
1980                         /* update lifetime (RFC2462 5.5.3 e) */
1981                         spin_lock(&ifp->lock);
1982                         now = jiffies;
1983                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1984                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1985                         else
1986                                 stored_lft = 0;
1987                         if (!update_lft && stored_lft) {
1988                                 if (valid_lft > MIN_VALID_LIFETIME ||
1989                                     valid_lft > stored_lft)
1990                                         update_lft = 1;
1991                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1992                                         /* valid_lft <= stored_lft is always true */
1993                                         /*
1994                                          * RFC 4862 Section 5.5.3e:
1995                                          * "Note that the preferred lifetime of
1996                                          *  the corresponding address is always
1997                                          *  reset to the Preferred Lifetime in
1998                                          *  the received Prefix Information
1999                                          *  option, regardless of whether the
2000                                          *  valid lifetime is also reset or
2001                                          *  ignored."
2002                                          *
2003                                          *  So if the preferred lifetime in
2004                                          *  this advertisement is different
2005                                          *  than what we have stored, but the
2006                                          *  valid lifetime is invalid, just
2007                                          *  reset prefered_lft.
2008                                          *
2009                                          *  We must set the valid lifetime
2010                                          *  to the stored lifetime since we'll
2011                                          *  be updating the timestamp below,
2012                                          *  else we'll set it back to the
2013                                          *  minimum.
2014                                          */
2015                                         if (prefered_lft != ifp->prefered_lft) {
2016                                                 valid_lft = stored_lft;
2017                                                 update_lft = 1;
2018                                         }
2019                                 } else {
2020                                         valid_lft = MIN_VALID_LIFETIME;
2021                                         if (valid_lft < prefered_lft)
2022                                                 prefered_lft = valid_lft;
2023                                         update_lft = 1;
2024                                 }
2025                         }
2026
2027                         if (update_lft) {
2028                                 ifp->valid_lft = valid_lft;
2029                                 ifp->prefered_lft = prefered_lft;
2030                                 ifp->tstamp = now;
2031                                 flags = ifp->flags;
2032                                 ifp->flags &= ~IFA_F_DEPRECATED;
2033                                 spin_unlock(&ifp->lock);
2034
2035                                 if (!(flags&IFA_F_TENTATIVE))
2036                                         ipv6_ifa_notify(0, ifp);
2037                         } else
2038                                 spin_unlock(&ifp->lock);
2039
2040 #ifdef CONFIG_IPV6_PRIVACY
2041                         read_lock_bh(&in6_dev->lock);
2042                         /* update all temporary addresses in the list */
2043                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2044                                             tmp_list) {
2045                                 int age, max_valid, max_prefered;
2046
2047                                 if (ifp != ift->ifpub)
2048                                         continue;
2049
2050                                 /*
2051                                  * RFC 4941 section 3.3:
2052                                  * If a received option will extend the lifetime
2053                                  * of a public address, the lifetimes of
2054                                  * temporary addresses should be extended,
2055                                  * subject to the overall constraint that no
2056                                  * temporary addresses should ever remain
2057                                  * "valid" or "preferred" for a time longer than
2058                                  * (TEMP_VALID_LIFETIME) or
2059                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2060                                  * respectively.
2061                                  */
2062                                 age = (now - ift->cstamp) / HZ;
2063                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2064                                 if (max_valid < 0)
2065                                         max_valid = 0;
2066
2067                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2068                                                in6_dev->cnf.max_desync_factor -
2069                                                age;
2070                                 if (max_prefered < 0)
2071                                         max_prefered = 0;
2072
2073                                 if (valid_lft > max_valid)
2074                                         valid_lft = max_valid;
2075
2076                                 if (prefered_lft > max_prefered)
2077                                         prefered_lft = max_prefered;
2078
2079                                 spin_lock(&ift->lock);
2080                                 flags = ift->flags;
2081                                 ift->valid_lft = valid_lft;
2082                                 ift->prefered_lft = prefered_lft;
2083                                 ift->tstamp = now;
2084                                 if (prefered_lft > 0)
2085                                         ift->flags &= ~IFA_F_DEPRECATED;
2086
2087                                 spin_unlock(&ift->lock);
2088                                 if (!(flags&IFA_F_TENTATIVE))
2089                                         ipv6_ifa_notify(0, ift);
2090                         }
2091
2092                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2093                                 /*
2094                                  * When a new public address is created as
2095                                  * described in [ADDRCONF], also create a new
2096                                  * temporary address. Also create a temporary
2097                                  * address if it's enabled but no temporary
2098                                  * address currently exists.
2099                                  */
2100                                 read_unlock_bh(&in6_dev->lock);
2101                                 ipv6_create_tempaddr(ifp, NULL);
2102                         } else {
2103                                 read_unlock_bh(&in6_dev->lock);
2104                         }
2105 #endif
2106                         in6_ifa_put(ifp);
2107                         addrconf_verify(0);
2108                 }
2109         }
2110         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2111         in6_dev_put(in6_dev);
2112 }
2113
2114 /*
2115  *      Set destination address.
2116  *      Special case for SIT interfaces where we create a new "virtual"
2117  *      device.
2118  */
2119 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2120 {
2121         struct in6_ifreq ireq;
2122         struct net_device *dev;
2123         int err = -EINVAL;
2124
2125         rtnl_lock();
2126
2127         err = -EFAULT;
2128         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2129                 goto err_exit;
2130
2131         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2132
2133         err = -ENODEV;
2134         if (dev == NULL)
2135                 goto err_exit;
2136
2137 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2138         if (dev->type == ARPHRD_SIT) {
2139                 const struct net_device_ops *ops = dev->netdev_ops;
2140                 struct ifreq ifr;
2141                 struct ip_tunnel_parm p;
2142
2143                 err = -EADDRNOTAVAIL;
2144                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2145                         goto err_exit;
2146
2147                 memset(&p, 0, sizeof(p));
2148                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2149                 p.iph.saddr = 0;
2150                 p.iph.version = 4;
2151                 p.iph.ihl = 5;
2152                 p.iph.protocol = IPPROTO_IPV6;
2153                 p.iph.ttl = 64;
2154                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2155
2156                 if (ops->ndo_do_ioctl) {
2157                         mm_segment_t oldfs = get_fs();
2158
2159                         set_fs(KERNEL_DS);
2160                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2161                         set_fs(oldfs);
2162                 } else
2163                         err = -EOPNOTSUPP;
2164
2165                 if (err == 0) {
2166                         err = -ENOBUFS;
2167                         dev = __dev_get_by_name(net, p.name);
2168                         if (!dev)
2169                                 goto err_exit;
2170                         err = dev_open(dev);
2171                 }
2172         }
2173 #endif
2174
2175 err_exit:
2176         rtnl_unlock();
2177         return err;
2178 }
2179
2180 /*
2181  *      Manual configuration of address on an interface
2182  */
2183 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2184                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2185                           __u32 valid_lft)
2186 {
2187         struct inet6_ifaddr *ifp;
2188         struct inet6_dev *idev;
2189         struct net_device *dev;
2190         int scope;
2191         u32 flags;
2192         clock_t expires;
2193         unsigned long timeout;
2194
2195         ASSERT_RTNL();
2196
2197         if (plen > 128)
2198                 return -EINVAL;
2199
2200         /* check the lifetime */
2201         if (!valid_lft || prefered_lft > valid_lft)
2202                 return -EINVAL;
2203
2204         dev = __dev_get_by_index(net, ifindex);
2205         if (!dev)
2206                 return -ENODEV;
2207
2208         idev = addrconf_add_dev(dev);
2209         if (IS_ERR(idev))
2210                 return PTR_ERR(idev);
2211
2212         scope = ipv6_addr_scope(pfx);
2213
2214         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2215         if (addrconf_finite_timeout(timeout)) {
2216                 expires = jiffies_to_clock_t(timeout * HZ);
2217                 valid_lft = timeout;
2218                 flags = RTF_EXPIRES;
2219         } else {
2220                 expires = 0;
2221                 flags = 0;
2222                 ifa_flags |= IFA_F_PERMANENT;
2223         }
2224
2225         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2226         if (addrconf_finite_timeout(timeout)) {
2227                 if (timeout == 0)
2228                         ifa_flags |= IFA_F_DEPRECATED;
2229                 prefered_lft = timeout;
2230         }
2231
2232         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2233
2234         if (!IS_ERR(ifp)) {
2235                 spin_lock_bh(&ifp->lock);
2236                 ifp->valid_lft = valid_lft;
2237                 ifp->prefered_lft = prefered_lft;
2238                 ifp->tstamp = jiffies;
2239                 spin_unlock_bh(&ifp->lock);
2240
2241                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2242                                       expires, flags);
2243                 /*
2244                  * Note that section 3.1 of RFC 4429 indicates
2245                  * that the Optimistic flag should not be set for
2246                  * manually configured addresses
2247                  */
2248                 addrconf_dad_start(ifp, 0);
2249                 in6_ifa_put(ifp);
2250                 addrconf_verify(0);
2251                 return 0;
2252         }
2253
2254         return PTR_ERR(ifp);
2255 }
2256
2257 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2258                           unsigned int plen)
2259 {
2260         struct inet6_ifaddr *ifp;
2261         struct inet6_dev *idev;
2262         struct net_device *dev;
2263
2264         if (plen > 128)
2265                 return -EINVAL;
2266
2267         dev = __dev_get_by_index(net, ifindex);
2268         if (!dev)
2269                 return -ENODEV;
2270
2271         if ((idev = __in6_dev_get(dev)) == NULL)
2272                 return -ENXIO;
2273
2274         read_lock_bh(&idev->lock);
2275         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2276                 if (ifp->prefix_len == plen &&
2277                     ipv6_addr_equal(pfx, &ifp->addr)) {
2278                         in6_ifa_hold(ifp);
2279                         read_unlock_bh(&idev->lock);
2280
2281                         ipv6_del_addr(ifp);
2282
2283                         /* If the last address is deleted administratively,
2284                            disable IPv6 on this interface.
2285                          */
2286                         if (list_empty(&idev->addr_list))
2287                                 addrconf_ifdown(idev->dev, 1);
2288                         return 0;
2289                 }
2290         }
2291         read_unlock_bh(&idev->lock);
2292         return -EADDRNOTAVAIL;
2293 }
2294
2295
2296 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2297 {
2298         struct in6_ifreq ireq;
2299         int err;
2300
2301         if (!capable(CAP_NET_ADMIN))
2302                 return -EPERM;
2303
2304         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2305                 return -EFAULT;
2306
2307         rtnl_lock();
2308         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2309                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2310                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2311         rtnl_unlock();
2312         return err;
2313 }
2314
2315 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2316 {
2317         struct in6_ifreq ireq;
2318         int err;
2319
2320         if (!capable(CAP_NET_ADMIN))
2321                 return -EPERM;
2322
2323         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2324                 return -EFAULT;
2325
2326         rtnl_lock();
2327         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2328                              ireq.ifr6_prefixlen);
2329         rtnl_unlock();
2330         return err;
2331 }
2332
2333 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2334                      int plen, int scope)
2335 {
2336         struct inet6_ifaddr *ifp;
2337
2338         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2339         if (!IS_ERR(ifp)) {
2340                 spin_lock_bh(&ifp->lock);
2341                 ifp->flags &= ~IFA_F_TENTATIVE;
2342                 spin_unlock_bh(&ifp->lock);
2343                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2344                 in6_ifa_put(ifp);
2345         }
2346 }
2347
2348 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2349 static void sit_add_v4_addrs(struct inet6_dev *idev)
2350 {
2351         struct in6_addr addr;
2352         struct net_device *dev;
2353         struct net *net = dev_net(idev->dev);
2354         int scope;
2355
2356         ASSERT_RTNL();
2357
2358         memset(&addr, 0, sizeof(struct in6_addr));
2359         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2360
2361         if (idev->dev->flags&IFF_POINTOPOINT) {
2362                 addr.s6_addr32[0] = htonl(0xfe800000);
2363                 scope = IFA_LINK;
2364         } else {
2365                 scope = IPV6_ADDR_COMPATv4;
2366         }
2367
2368         if (addr.s6_addr32[3]) {
2369                 add_addr(idev, &addr, 128, scope);
2370                 return;
2371         }
2372
2373         for_each_netdev(net, dev) {
2374                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2375                 if (in_dev && (dev->flags & IFF_UP)) {
2376                         struct in_ifaddr * ifa;
2377
2378                         int flag = scope;
2379
2380                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2381                                 int plen;
2382
2383                                 addr.s6_addr32[3] = ifa->ifa_local;
2384
2385                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2386                                         continue;
2387                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2388                                         if (idev->dev->flags&IFF_POINTOPOINT)
2389                                                 continue;
2390                                         flag |= IFA_HOST;
2391                                 }
2392                                 if (idev->dev->flags&IFF_POINTOPOINT)
2393                                         plen = 64;
2394                                 else
2395                                         plen = 96;
2396
2397                                 add_addr(idev, &addr, plen, flag);
2398                         }
2399                 }
2400         }
2401 }
2402 #endif
2403
2404 static void init_loopback(struct net_device *dev)
2405 {
2406         struct inet6_dev  *idev;
2407
2408         /* ::1 */
2409
2410         ASSERT_RTNL();
2411
2412         if ((idev = ipv6_find_idev(dev)) == NULL) {
2413                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2414                 return;
2415         }
2416
2417         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2418 }
2419
2420 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2421 {
2422         struct inet6_ifaddr * ifp;
2423         u32 addr_flags = IFA_F_PERMANENT;
2424
2425 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2426         if (idev->cnf.optimistic_dad &&
2427             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2428                 addr_flags |= IFA_F_OPTIMISTIC;
2429 #endif
2430
2431
2432         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2433         if (!IS_ERR(ifp)) {
2434                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2435                 addrconf_dad_start(ifp, 0);
2436                 in6_ifa_put(ifp);
2437         }
2438 }
2439
2440 static void addrconf_dev_config(struct net_device *dev)
2441 {
2442         struct in6_addr addr;
2443         struct inet6_dev    * idev;
2444
2445         ASSERT_RTNL();
2446
2447         if ((dev->type != ARPHRD_ETHER) &&
2448             (dev->type != ARPHRD_FDDI) &&
2449             (dev->type != ARPHRD_IEEE802_TR) &&
2450             (dev->type != ARPHRD_ARCNET) &&
2451             (dev->type != ARPHRD_INFINIBAND)) {
2452                 /* Alas, we support only Ethernet autoconfiguration. */
2453                 return;
2454         }
2455
2456         idev = addrconf_add_dev(dev);
2457         if (IS_ERR(idev))
2458                 return;
2459
2460         memset(&addr, 0, sizeof(struct in6_addr));
2461         addr.s6_addr32[0] = htonl(0xFE800000);
2462
2463         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2464                 addrconf_add_linklocal(idev, &addr);
2465 }
2466
2467 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2468 static void addrconf_sit_config(struct net_device *dev)
2469 {
2470         struct inet6_dev *idev;
2471
2472         ASSERT_RTNL();
2473
2474         /*
2475          * Configure the tunnel with one of our IPv4
2476          * addresses... we should configure all of
2477          * our v4 addrs in the tunnel
2478          */
2479
2480         if ((idev = ipv6_find_idev(dev)) == NULL) {
2481                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2482                 return;
2483         }
2484
2485         if (dev->priv_flags & IFF_ISATAP) {
2486                 struct in6_addr addr;
2487
2488                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2489                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2490                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2491                         addrconf_add_linklocal(idev, &addr);
2492                 return;
2493         }
2494
2495         sit_add_v4_addrs(idev);
2496
2497         if (dev->flags&IFF_POINTOPOINT) {
2498                 addrconf_add_mroute(dev);
2499                 addrconf_add_lroute(dev);
2500         } else
2501                 sit_route_add(dev);
2502 }
2503 #endif
2504
2505 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2506 static void addrconf_gre_config(struct net_device *dev)
2507 {
2508         struct inet6_dev *idev;
2509         struct in6_addr addr;
2510
2511         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2512
2513         ASSERT_RTNL();
2514
2515         if ((idev = ipv6_find_idev(dev)) == NULL) {
2516                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2517                 return;
2518         }
2519
2520         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2521         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2522
2523         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2524                 addrconf_add_linklocal(idev, &addr);
2525 }
2526 #endif
2527
2528 static inline int
2529 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2530 {
2531         struct in6_addr lladdr;
2532
2533         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2534                 addrconf_add_linklocal(idev, &lladdr);
2535                 return 0;
2536         }
2537         return -1;
2538 }
2539
2540 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2541 {
2542         struct net_device *link_dev;
2543         struct net *net = dev_net(idev->dev);
2544
2545         /* first try to inherit the link-local address from the link device */
2546         if (idev->dev->iflink &&
2547             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2548                 if (!ipv6_inherit_linklocal(idev, link_dev))
2549                         return;
2550         }
2551         /* then try to inherit it from any device */
2552         for_each_netdev(net, link_dev) {
2553                 if (!ipv6_inherit_linklocal(idev, link_dev))
2554                         return;
2555         }
2556         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2557 }
2558
2559 /*
2560  * Autoconfigure tunnel with a link-local address so routing protocols,
2561  * DHCPv6, MLD etc. can be run over the virtual link
2562  */
2563
2564 static void addrconf_ip6_tnl_config(struct net_device *dev)
2565 {
2566         struct inet6_dev *idev;
2567
2568         ASSERT_RTNL();
2569
2570         idev = addrconf_add_dev(dev);
2571         if (IS_ERR(idev)) {
2572                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2573                 return;
2574         }
2575         ip6_tnl_add_linklocal(idev);
2576 }
2577
2578 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2579                            void * data)
2580 {
2581         struct net_device *dev = (struct net_device *) data;
2582         struct inet6_dev *idev = __in6_dev_get(dev);
2583         int run_pending = 0;
2584         int err;
2585
2586         switch (event) {
2587         case NETDEV_REGISTER:
2588                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2589                         idev = ipv6_add_dev(dev);
2590                         if (!idev)
2591                                 return notifier_from_errno(-ENOMEM);
2592                 }
2593                 break;
2594
2595         case NETDEV_UP:
2596         case NETDEV_CHANGE:
2597                 if (dev->flags & IFF_SLAVE)
2598                         break;
2599
2600                 if (event == NETDEV_UP) {
2601                         if (!addrconf_qdisc_ok(dev)) {
2602                                 /* device is not ready yet. */
2603                                 printk(KERN_INFO
2604                                         "ADDRCONF(NETDEV_UP): %s: "
2605                                         "link is not ready\n",
2606                                         dev->name);
2607                                 break;
2608                         }
2609
2610                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2611                                 idev = ipv6_add_dev(dev);
2612
2613                         if (idev) {
2614                                 idev->if_flags |= IF_READY;
2615                                 run_pending = 1;
2616                         }
2617                 } else {
2618                         if (!addrconf_qdisc_ok(dev)) {
2619                                 /* device is still not ready. */
2620                                 break;
2621                         }
2622
2623                         if (idev) {
2624                                 if (idev->if_flags & IF_READY)
2625                                         /* device is already configured. */
2626                                         break;
2627                                 idev->if_flags |= IF_READY;
2628                         }
2629
2630                         printk(KERN_INFO
2631                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2632                                         "link becomes ready\n",
2633                                         dev->name);
2634
2635                         run_pending = 1;
2636                 }
2637
2638                 switch (dev->type) {
2639 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2640                 case ARPHRD_SIT:
2641                         addrconf_sit_config(dev);
2642                         break;
2643 #endif
2644 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2645                 case ARPHRD_IPGRE:
2646                         addrconf_gre_config(dev);
2647                         break;
2648 #endif
2649                 case ARPHRD_TUNNEL6:
2650                         addrconf_ip6_tnl_config(dev);
2651                         break;
2652                 case ARPHRD_LOOPBACK:
2653                         init_loopback(dev);
2654                         break;
2655
2656                 default:
2657                         addrconf_dev_config(dev);
2658                         break;
2659                 }
2660
2661                 if (idev) {
2662                         if (run_pending)
2663                                 addrconf_dad_run(idev);
2664
2665                         /*
2666                          * If the MTU changed during the interface down,
2667                          * when the interface up, the changed MTU must be
2668                          * reflected in the idev as well as routers.
2669                          */
2670                         if (idev->cnf.mtu6 != dev->mtu &&
2671                             dev->mtu >= IPV6_MIN_MTU) {
2672                                 rt6_mtu_change(dev, dev->mtu);
2673                                 idev->cnf.mtu6 = dev->mtu;
2674                         }
2675                         idev->tstamp = jiffies;
2676                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2677
2678                         /*
2679                          * If the changed mtu during down is lower than
2680                          * IPV6_MIN_MTU stop IPv6 on this interface.
2681                          */
2682                         if (dev->mtu < IPV6_MIN_MTU)
2683                                 addrconf_ifdown(dev, 1);
2684                 }
2685                 break;
2686
2687         case NETDEV_CHANGEMTU:
2688                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2689                         rt6_mtu_change(dev, dev->mtu);
2690                         idev->cnf.mtu6 = dev->mtu;
2691                         break;
2692                 }
2693
2694                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2695                         idev = ipv6_add_dev(dev);
2696                         if (idev)
2697                                 break;
2698                 }
2699
2700                 /*
2701                  * MTU falled under IPV6_MIN_MTU.
2702                  * Stop IPv6 on this interface.
2703                  */
2704
2705         case NETDEV_DOWN:
2706         case NETDEV_UNREGISTER:
2707                 /*
2708                  *      Remove all addresses from this interface.
2709                  */
2710                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2711                 break;
2712
2713         case NETDEV_CHANGENAME:
2714                 if (idev) {
2715                         snmp6_unregister_dev(idev);
2716                         addrconf_sysctl_unregister(idev);
2717                         addrconf_sysctl_register(idev);
2718                         err = snmp6_register_dev(idev);
2719                         if (err)
2720                                 return notifier_from_errno(err);
2721                 }
2722                 break;
2723
2724         case NETDEV_PRE_TYPE_CHANGE:
2725         case NETDEV_POST_TYPE_CHANGE:
2726                 addrconf_type_change(dev, event);
2727                 break;
2728         }
2729
2730         return NOTIFY_OK;
2731 }
2732
2733 /*
2734  *      addrconf module should be notified of a device going up
2735  */
2736 static struct notifier_block ipv6_dev_notf = {
2737         .notifier_call = addrconf_notify,
2738 };
2739
2740 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2741 {
2742         struct inet6_dev *idev;
2743         ASSERT_RTNL();
2744
2745         idev = __in6_dev_get(dev);
2746
2747         if (event == NETDEV_POST_TYPE_CHANGE)
2748                 ipv6_mc_remap(idev);
2749         else if (event == NETDEV_PRE_TYPE_CHANGE)
2750                 ipv6_mc_unmap(idev);
2751 }
2752
2753 static int addrconf_ifdown(struct net_device *dev, int how)
2754 {
2755         struct net *net = dev_net(dev);
2756         struct inet6_dev *idev;
2757         struct inet6_ifaddr *ifa;
2758         int state, i;
2759
2760         ASSERT_RTNL();
2761
2762         rt6_ifdown(net, dev);
2763         neigh_ifdown(&nd_tbl, dev);
2764
2765         idev = __in6_dev_get(dev);
2766         if (idev == NULL)
2767                 return -ENODEV;
2768
2769         /*
2770          * Step 1: remove reference to ipv6 device from parent device.
2771          *         Do not dev_put!
2772          */
2773         if (how) {
2774                 idev->dead = 1;
2775
2776                 /* protected by rtnl_lock */
2777                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2778
2779                 /* Step 1.5: remove snmp6 entry */
2780                 snmp6_unregister_dev(idev);
2781
2782         }
2783
2784         /* Step 2: clear hash table */
2785         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2786                 struct hlist_head *h = &inet6_addr_lst[i];
2787                 struct hlist_node *n;
2788
2789                 spin_lock_bh(&addrconf_hash_lock);
2790         restart:
2791                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2792                         if (ifa->idev == idev) {
2793                                 hlist_del_init_rcu(&ifa->addr_lst);
2794                                 addrconf_del_timer(ifa);
2795                                 goto restart;
2796                         }
2797                 }
2798                 spin_unlock_bh(&addrconf_hash_lock);
2799         }
2800
2801         write_lock_bh(&idev->lock);
2802
2803         /* Step 2: clear flags for stateless addrconf */
2804         if (!how)
2805                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2806
2807 #ifdef CONFIG_IPV6_PRIVACY
2808         if (how && del_timer(&idev->regen_timer))
2809                 in6_dev_put(idev);
2810
2811         /* Step 3: clear tempaddr list */
2812         while (!list_empty(&idev->tempaddr_list)) {
2813                 ifa = list_first_entry(&idev->tempaddr_list,
2814                                        struct inet6_ifaddr, tmp_list);
2815                 list_del(&ifa->tmp_list);
2816                 write_unlock_bh(&idev->lock);
2817                 spin_lock_bh(&ifa->lock);
2818
2819                 if (ifa->ifpub) {
2820                         in6_ifa_put(ifa->ifpub);
2821                         ifa->ifpub = NULL;
2822                 }
2823                 spin_unlock_bh(&ifa->lock);
2824                 in6_ifa_put(ifa);
2825                 write_lock_bh(&idev->lock);
2826         }
2827 #endif
2828
2829         while (!list_empty(&idev->addr_list)) {
2830                 ifa = list_first_entry(&idev->addr_list,
2831                                        struct inet6_ifaddr, if_list);
2832                 addrconf_del_timer(ifa);
2833
2834                 list_del(&ifa->if_list);
2835
2836                 write_unlock_bh(&idev->lock);
2837
2838                 spin_lock_bh(&ifa->state_lock);
2839                 state = ifa->state;
2840                 ifa->state = INET6_IFADDR_STATE_DEAD;
2841                 spin_unlock_bh(&ifa->state_lock);
2842
2843                 if (state != INET6_IFADDR_STATE_DEAD) {
2844                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2845                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2846                 }
2847                 in6_ifa_put(ifa);
2848
2849                 write_lock_bh(&idev->lock);
2850         }
2851
2852         write_unlock_bh(&idev->lock);
2853
2854         /* Step 5: Discard multicast list */
2855         if (how)
2856                 ipv6_mc_destroy_dev(idev);
2857         else
2858                 ipv6_mc_down(idev);
2859
2860         idev->tstamp = jiffies;
2861
2862         /* Last: Shot the device (if unregistered) */
2863         if (how) {
2864                 addrconf_sysctl_unregister(idev);
2865                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2866                 neigh_ifdown(&nd_tbl, dev);
2867                 in6_dev_put(idev);
2868         }
2869         return 0;
2870 }
2871
2872 static void addrconf_rs_timer(unsigned long data)
2873 {
2874         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2875         struct inet6_dev *idev = ifp->idev;
2876
2877         read_lock(&idev->lock);
2878         if (idev->dead || !(idev->if_flags & IF_READY))
2879                 goto out;
2880
2881         if (idev->cnf.forwarding)
2882                 goto out;
2883
2884         /* Announcement received after solicitation was sent */
2885         if (idev->if_flags & IF_RA_RCVD)
2886                 goto out;
2887
2888         spin_lock(&ifp->lock);
2889         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2890                 /* The wait after the last probe can be shorter */
2891                 addrconf_mod_timer(ifp, AC_RS,
2892                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2893                                    idev->cnf.rtr_solicit_delay :
2894                                    idev->cnf.rtr_solicit_interval);
2895                 spin_unlock(&ifp->lock);
2896
2897                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2898         } else {
2899                 spin_unlock(&ifp->lock);
2900                 /*
2901                  * Note: we do not support deprecated "all on-link"
2902                  * assumption any longer.
2903                  */
2904                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2905                        idev->dev->name);
2906         }
2907
2908 out:
2909         read_unlock(&idev->lock);
2910         in6_ifa_put(ifp);
2911 }
2912
2913 /*
2914  *      Duplicate Address Detection
2915  */
2916 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2917 {
2918         unsigned long rand_num;
2919         struct inet6_dev *idev = ifp->idev;
2920
2921         if (ifp->flags & IFA_F_OPTIMISTIC)
2922                 rand_num = 0;
2923         else
2924                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2925
2926         ifp->probes = idev->cnf.dad_transmits;
2927         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2928 }
2929
2930 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2931 {
2932         struct inet6_dev *idev = ifp->idev;
2933         struct net_device *dev = idev->dev;
2934
2935         addrconf_join_solict(dev, &ifp->addr);
2936
2937         net_srandom(ifp->addr.s6_addr32[3]);
2938
2939         read_lock_bh(&idev->lock);
2940         spin_lock(&ifp->lock);
2941         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2942                 goto out;
2943
2944         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2945             idev->cnf.accept_dad < 1 ||
2946             !(ifp->flags&IFA_F_TENTATIVE) ||
2947             ifp->flags & IFA_F_NODAD) {
2948                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2949                 spin_unlock(&ifp->lock);
2950                 read_unlock_bh(&idev->lock);
2951
2952                 addrconf_dad_completed(ifp);
2953                 return;
2954         }
2955
2956         if (!(idev->if_flags & IF_READY)) {
2957                 spin_unlock(&ifp->lock);
2958                 read_unlock_bh(&idev->lock);
2959                 /*
2960                  * If the device is not ready:
2961                  * - keep it tentative if it is a permanent address.
2962                  * - otherwise, kill it.
2963                  */
2964                 in6_ifa_hold(ifp);
2965                 addrconf_dad_stop(ifp, 0);
2966                 return;
2967         }
2968
2969         /*
2970          * Optimistic nodes can start receiving
2971          * Frames right away
2972          */
2973         if (ifp->flags & IFA_F_OPTIMISTIC)
2974                 ip6_ins_rt(ifp->rt);
2975
2976         addrconf_dad_kick(ifp);
2977 out:
2978         spin_unlock(&ifp->lock);
2979         read_unlock_bh(&idev->lock);
2980 }
2981
2982 static void addrconf_dad_timer(unsigned long data)
2983 {
2984         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2985         struct inet6_dev *idev = ifp->idev;
2986         struct in6_addr mcaddr;
2987
2988         if (!ifp->probes && addrconf_dad_end(ifp))
2989                 goto out;
2990
2991         read_lock(&idev->lock);
2992         if (idev->dead || !(idev->if_flags & IF_READY)) {
2993                 read_unlock(&idev->lock);
2994                 goto out;
2995         }
2996
2997         spin_lock(&ifp->lock);
2998         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2999                 spin_unlock(&ifp->lock);
3000                 read_unlock(&idev->lock);
3001                 goto out;
3002         }
3003
3004         if (ifp->probes == 0) {
3005                 /*
3006                  * DAD was successful
3007                  */
3008
3009                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3010                 spin_unlock(&ifp->lock);
3011                 read_unlock(&idev->lock);
3012
3013                 addrconf_dad_completed(ifp);
3014
3015                 goto out;
3016         }
3017
3018         ifp->probes--;
3019         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3020         spin_unlock(&ifp->lock);
3021         read_unlock(&idev->lock);
3022
3023         /* send a neighbour solicitation for our addr */
3024         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3025         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3026 out:
3027         in6_ifa_put(ifp);
3028 }
3029
3030 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3031 {
3032         struct net_device *dev = ifp->idev->dev;
3033
3034         /*
3035          *      Configure the address for reception. Now it is valid.
3036          */
3037
3038         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3039
3040         /* If added prefix is link local and we are prepared to process
3041            router advertisements, start sending router solicitations.
3042          */
3043
3044         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3045              ifp->idev->cnf.accept_ra == 2) &&
3046             ifp->idev->cnf.rtr_solicits > 0 &&
3047             (dev->flags&IFF_LOOPBACK) == 0 &&
3048             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3049                 /*
3050                  *      If a host as already performed a random delay
3051                  *      [...] as part of DAD [...] there is no need
3052                  *      to delay again before sending the first RS
3053                  */
3054                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3055
3056                 spin_lock_bh(&ifp->lock);
3057                 ifp->probes = 1;
3058                 ifp->idev->if_flags |= IF_RS_SENT;
3059                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3060                 spin_unlock_bh(&ifp->lock);
3061         }
3062 }
3063
3064 static void addrconf_dad_run(struct inet6_dev *idev)
3065 {
3066         struct inet6_ifaddr *ifp;
3067
3068         read_lock_bh(&idev->lock);
3069         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3070                 spin_lock(&ifp->lock);
3071                 if (ifp->flags & IFA_F_TENTATIVE &&
3072                     ifp->state == INET6_IFADDR_STATE_DAD)
3073                         addrconf_dad_kick(ifp);
3074                 spin_unlock(&ifp->lock);
3075         }
3076         read_unlock_bh(&idev->lock);
3077 }
3078
3079 #ifdef CONFIG_PROC_FS
3080 struct if6_iter_state {
3081         struct seq_net_private p;
3082         int bucket;
3083 };
3084
3085 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3086 {
3087         struct inet6_ifaddr *ifa = NULL;
3088         struct if6_iter_state *state = seq->private;
3089         struct net *net = seq_file_net(seq);
3090
3091         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3092                 struct hlist_node *n;
3093                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3094                                          addr_lst)
3095                         if (net_eq(dev_net(ifa->idev->dev), net))
3096                                 return ifa;
3097         }
3098         return NULL;
3099 }
3100
3101 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3102                                          struct inet6_ifaddr *ifa)
3103 {
3104         struct if6_iter_state *state = seq->private;
3105         struct net *net = seq_file_net(seq);
3106         struct hlist_node *n = &ifa->addr_lst;
3107
3108         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3109                 if (net_eq(dev_net(ifa->idev->dev), net))
3110                         return ifa;
3111
3112         while (++state->bucket < IN6_ADDR_HSIZE) {
3113                 hlist_for_each_entry_rcu_bh(ifa, n,
3114                                      &inet6_addr_lst[state->bucket], addr_lst) {
3115                         if (net_eq(dev_net(ifa->idev->dev), net))
3116                                 return ifa;
3117                 }
3118         }
3119
3120         return NULL;
3121 }
3122
3123 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3124 {
3125         struct inet6_ifaddr *ifa = if6_get_first(seq);
3126
3127         if (ifa)
3128                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3129                         --pos;
3130         return pos ? NULL : ifa;
3131 }
3132
3133 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3134         __acquires(rcu_bh)
3135 {
3136         rcu_read_lock_bh();
3137         return if6_get_idx(seq, *pos);
3138