[IPV6] address: Add put_ifaddrmsg() and rt_scope()
[safe/jmp/linux-2.6] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>     
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16
17 /*
18  *      Changes:
19  *
20  *      Janos Farkas                    :       delete timer on ifdown
21  *      <chexum@bankinf.banki.hu>
22  *      Andi Kleen                      :       kill double kfree on module
23  *                                              unload.
24  *      Maciej W. Rozycki               :       FDDI support
25  *      sekiya@USAGI                    :       Don't send too many RS
26  *                                              packets.
27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
28  *                                              packets.
29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
30  *                                              address validation timer.
31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
32  *                                              support.
33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
34  *                                              address on a same interface.
35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
37  *                                              seq_file.
38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
39  *                                              selection; consider scope,
40  *                                              status etc.
41  */
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/sched.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #ifdef CONFIG_SYSCTL
59 #include <linux/sysctl.h>
60 #endif
61 #include <linux/capability.h>
62 #include <linux/delay.h>
63 #include <linux/notifier.h>
64 #include <linux/string.h>
65
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 <linux/if_tunnel.h>
78 #include <linux/rtnetlink.h>
79
80 #ifdef CONFIG_IPV6_PRIVACY
81 #include <linux/random.h>
82 #endif
83
84 #include <asm/uaccess.h>
85
86 #include <linux/proc_fs.h>
87 #include <linux/seq_file.h>
88
89 /* Set to 3 to get tracing... */
90 #define ACONF_DEBUG 2
91
92 #if ACONF_DEBUG >= 3
93 #define ADBG(x) printk x
94 #else
95 #define ADBG(x)
96 #endif
97
98 #define INFINITY_LIFE_TIME      0xFFFFFFFF
99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
100
101 #ifdef CONFIG_SYSCTL
102 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p);
103 static void addrconf_sysctl_unregister(struct ipv6_devconf *p);
104 #endif
105
106 #ifdef CONFIG_IPV6_PRIVACY
107 static int __ipv6_regen_rndid(struct inet6_dev *idev);
108 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr); 
109 static void ipv6_regen_rndid(unsigned long data);
110
111 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
112 #endif
113
114 static int ipv6_count_addresses(struct inet6_dev *idev);
115
116 /*
117  *      Configured unicast address hash table
118  */
119 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
120 static DEFINE_RWLOCK(addrconf_hash_lock);
121
122 /* Protects inet6 devices */
123 DEFINE_RWLOCK(addrconf_lock);
124
125 static void addrconf_verify(unsigned long);
126
127 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
128 static DEFINE_SPINLOCK(addrconf_verify_lock);
129
130 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
131 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
132
133 static int addrconf_ifdown(struct net_device *dev, int how);
134
135 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
136 static void addrconf_dad_timer(unsigned long data);
137 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
138 static void addrconf_dad_run(struct inet6_dev *idev);
139 static void addrconf_rs_timer(unsigned long data);
140 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
141 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
142
143 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 
144                                 struct prefix_info *pinfo);
145 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
146
147 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
148
149 struct ipv6_devconf ipv6_devconf __read_mostly = {
150         .forwarding             = 0,
151         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
152         .mtu6                   = IPV6_MIN_MTU,
153         .accept_ra              = 1,
154         .accept_redirects       = 1,
155         .autoconf               = 1,
156         .force_mld_version      = 0,
157         .dad_transmits          = 1,
158         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
159         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
160         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
161 #ifdef CONFIG_IPV6_PRIVACY
162         .use_tempaddr           = 0,
163         .temp_valid_lft         = TEMP_VALID_LIFETIME,
164         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
165         .regen_max_retry        = REGEN_MAX_RETRY,
166         .max_desync_factor      = MAX_DESYNC_FACTOR,
167 #endif
168         .max_addresses          = IPV6_MAX_ADDRESSES,
169         .accept_ra_defrtr       = 1,
170         .accept_ra_pinfo        = 1,
171 #ifdef CONFIG_IPV6_ROUTER_PREF
172         .accept_ra_rtr_pref     = 1,
173         .rtr_probe_interval     = 60 * HZ,
174 #ifdef CONFIG_IPV6_ROUTE_INFO
175         .accept_ra_rt_info_max_plen = 0,
176 #endif
177 #endif
178 };
179
180 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
181         .forwarding             = 0,
182         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
183         .mtu6                   = IPV6_MIN_MTU,
184         .accept_ra              = 1,
185         .accept_redirects       = 1,
186         .autoconf               = 1,
187         .dad_transmits          = 1,
188         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
189         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
190         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
191 #ifdef CONFIG_IPV6_PRIVACY
192         .use_tempaddr           = 0,
193         .temp_valid_lft         = TEMP_VALID_LIFETIME,
194         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
195         .regen_max_retry        = REGEN_MAX_RETRY,
196         .max_desync_factor      = MAX_DESYNC_FACTOR,
197 #endif
198         .max_addresses          = IPV6_MAX_ADDRESSES,
199         .accept_ra_defrtr       = 1,
200         .accept_ra_pinfo        = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202         .accept_ra_rtr_pref     = 1,
203         .rtr_probe_interval     = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205         .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208 };
209
210 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
211 #if 0
212 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
213 #endif
214 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
215
216 #define IPV6_ADDR_SCOPE_TYPE(scope)     ((scope) << 16)
217
218 static inline unsigned ipv6_addr_scope2type(unsigned scope)
219 {
220         switch(scope) {
221         case IPV6_ADDR_SCOPE_NODELOCAL:
222                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_NODELOCAL) |
223                         IPV6_ADDR_LOOPBACK);
224         case IPV6_ADDR_SCOPE_LINKLOCAL:
225                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL) |
226                         IPV6_ADDR_LINKLOCAL);
227         case IPV6_ADDR_SCOPE_SITELOCAL:
228                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL) |
229                         IPV6_ADDR_SITELOCAL);
230         }
231         return IPV6_ADDR_SCOPE_TYPE(scope);
232 }
233
234 int __ipv6_addr_type(const struct in6_addr *addr)
235 {
236         u32 st;
237
238         st = addr->s6_addr32[0];
239
240         /* Consider all addresses with the first three bits different of
241            000 and 111 as unicasts.
242          */
243         if ((st & htonl(0xE0000000)) != htonl(0x00000000) &&
244             (st & htonl(0xE0000000)) != htonl(0xE0000000))
245                 return (IPV6_ADDR_UNICAST | 
246                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));
247
248         if ((st & htonl(0xFF000000)) == htonl(0xFF000000)) {
249                 /* multicast */
250                 /* addr-select 3.1 */
251                 return (IPV6_ADDR_MULTICAST |
252                         ipv6_addr_scope2type(IPV6_ADDR_MC_SCOPE(addr)));
253         }
254
255         if ((st & htonl(0xFFC00000)) == htonl(0xFE800000))
256                 return (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST | 
257                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));               /* addr-select 3.1 */
258         if ((st & htonl(0xFFC00000)) == htonl(0xFEC00000))
259                 return (IPV6_ADDR_SITELOCAL | IPV6_ADDR_UNICAST |
260                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL));               /* addr-select 3.1 */
261
262         if ((addr->s6_addr32[0] | addr->s6_addr32[1]) == 0) {
263                 if (addr->s6_addr32[2] == 0) {
264                         if (addr->s6_addr32[3] == 0)
265                                 return IPV6_ADDR_ANY;
266
267                         if (addr->s6_addr32[3] == htonl(0x00000001))
268                                 return (IPV6_ADDR_LOOPBACK | IPV6_ADDR_UNICAST |
269                                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));       /* addr-select 3.4 */
270
271                         return (IPV6_ADDR_COMPATv4 | IPV6_ADDR_UNICAST |
272                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
273                 }
274
275                 if (addr->s6_addr32[2] == htonl(0x0000ffff))
276                         return (IPV6_ADDR_MAPPED | 
277                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
278         }
279
280         return (IPV6_ADDR_RESERVED | 
281                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.4 */
282 }
283
284 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
285 {
286         if (del_timer(&ifp->timer))
287                 __in6_ifa_put(ifp);
288 }
289
290 enum addrconf_timer_t
291 {
292         AC_NONE,
293         AC_DAD,
294         AC_RS,
295 };
296
297 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
298                                enum addrconf_timer_t what,
299                                unsigned long when)
300 {
301         if (!del_timer(&ifp->timer))
302                 in6_ifa_hold(ifp);
303
304         switch (what) {
305         case AC_DAD:
306                 ifp->timer.function = addrconf_dad_timer;
307                 break;
308         case AC_RS:
309                 ifp->timer.function = addrconf_rs_timer;
310                 break;
311         default:;
312         }
313         ifp->timer.expires = jiffies + when;
314         add_timer(&ifp->timer);
315 }
316
317 /* Nobody refers to this device, we may destroy it. */
318
319 void in6_dev_finish_destroy(struct inet6_dev *idev)
320 {
321         struct net_device *dev = idev->dev;
322         BUG_TRAP(idev->addr_list==NULL);
323         BUG_TRAP(idev->mc_list==NULL);
324 #ifdef NET_REFCNT_DEBUG
325         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
326 #endif
327         dev_put(dev);
328         if (!idev->dead) {
329                 printk("Freeing alive inet6 device %p\n", idev);
330                 return;
331         }
332         snmp6_free_dev(idev);
333         kfree(idev);
334 }
335
336 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
337 {
338         struct inet6_dev *ndev;
339
340         ASSERT_RTNL();
341
342         if (dev->mtu < IPV6_MIN_MTU)
343                 return NULL;
344
345         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
346
347         if (ndev == NULL)
348                 return NULL;
349
350         rwlock_init(&ndev->lock);
351         ndev->dev = dev;
352         memcpy(&ndev->cnf, &ipv6_devconf_dflt, sizeof(ndev->cnf));
353         ndev->cnf.mtu6 = dev->mtu;
354         ndev->cnf.sysctl = NULL;
355         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
356         if (ndev->nd_parms == NULL) {
357                 kfree(ndev);
358                 return NULL;
359         }
360         /* We refer to the device */
361         dev_hold(dev);
362
363         if (snmp6_alloc_dev(ndev) < 0) {
364                 ADBG((KERN_WARNING
365                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
366                         __FUNCTION__, dev->name));
367                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
368                 ndev->dead = 1;
369                 in6_dev_finish_destroy(ndev);
370                 return NULL;
371         }
372
373         if (snmp6_register_dev(ndev) < 0) {
374                 ADBG((KERN_WARNING
375                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
376                         __FUNCTION__, dev->name));
377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378                 ndev->dead = 1;
379                 in6_dev_finish_destroy(ndev);
380                 return NULL;
381         }
382
383         /* One reference from device.  We must do this before
384          * we invoke __ipv6_regen_rndid().
385          */
386         in6_dev_hold(ndev);
387
388 #ifdef CONFIG_IPV6_PRIVACY
389         init_timer(&ndev->regen_timer);
390         ndev->regen_timer.function = ipv6_regen_rndid;
391         ndev->regen_timer.data = (unsigned long) ndev;
392         if ((dev->flags&IFF_LOOPBACK) ||
393             dev->type == ARPHRD_TUNNEL ||
394             dev->type == ARPHRD_NONE ||
395             dev->type == ARPHRD_SIT) {
396                 printk(KERN_INFO
397                        "%s: Disabled Privacy Extensions\n",
398                        dev->name);
399                 ndev->cnf.use_tempaddr = -1;
400         } else {
401                 in6_dev_hold(ndev);
402                 ipv6_regen_rndid((unsigned long) ndev);
403         }
404 #endif
405
406         if (netif_carrier_ok(dev))
407                 ndev->if_flags |= IF_READY;
408
409         write_lock_bh(&addrconf_lock);
410         dev->ip6_ptr = ndev;
411         write_unlock_bh(&addrconf_lock);
412
413         ipv6_mc_init_dev(ndev);
414         ndev->tstamp = jiffies;
415 #ifdef CONFIG_SYSCTL
416         neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
417                               NET_IPV6_NEIGH, "ipv6",
418                               &ndisc_ifinfo_sysctl_change,
419                               NULL);
420         addrconf_sysctl_register(ndev, &ndev->cnf);
421 #endif
422         return ndev;
423 }
424
425 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
426 {
427         struct inet6_dev *idev;
428
429         ASSERT_RTNL();
430
431         if ((idev = __in6_dev_get(dev)) == NULL) {
432                 if ((idev = ipv6_add_dev(dev)) == NULL)
433                         return NULL;
434         }
435
436         if (dev->flags&IFF_UP)
437                 ipv6_mc_up(idev);
438         return idev;
439 }
440
441 #ifdef CONFIG_SYSCTL
442 static void dev_forward_change(struct inet6_dev *idev)
443 {
444         struct net_device *dev;
445         struct inet6_ifaddr *ifa;
446         struct in6_addr addr;
447
448         if (!idev)
449                 return;
450         dev = idev->dev;
451         if (dev && (dev->flags & IFF_MULTICAST)) {
452                 ipv6_addr_all_routers(&addr);
453         
454                 if (idev->cnf.forwarding)
455                         ipv6_dev_mc_inc(dev, &addr);
456                 else
457                         ipv6_dev_mc_dec(dev, &addr);
458         }
459         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
460                 if (idev->cnf.forwarding)
461                         addrconf_join_anycast(ifa);
462                 else
463                         addrconf_leave_anycast(ifa);
464         }
465 }
466
467
468 static void addrconf_forward_change(void)
469 {
470         struct net_device *dev;
471         struct inet6_dev *idev;
472
473         read_lock(&dev_base_lock);
474         for (dev=dev_base; dev; dev=dev->next) {
475                 read_lock(&addrconf_lock);
476                 idev = __in6_dev_get(dev);
477                 if (idev) {
478                         int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
479                         idev->cnf.forwarding = ipv6_devconf.forwarding;
480                         if (changed)
481                                 dev_forward_change(idev);
482                 }
483                 read_unlock(&addrconf_lock);
484         }
485         read_unlock(&dev_base_lock);
486 }
487 #endif
488
489 /* Nobody refers to this ifaddr, destroy it */
490
491 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
492 {
493         BUG_TRAP(ifp->if_next==NULL);
494         BUG_TRAP(ifp->lst_next==NULL);
495 #ifdef NET_REFCNT_DEBUG
496         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
497 #endif
498
499         in6_dev_put(ifp->idev);
500
501         if (del_timer(&ifp->timer))
502                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
503
504         if (!ifp->dead) {
505                 printk("Freeing alive inet6 address %p\n", ifp);
506                 return;
507         }
508         dst_release(&ifp->rt->u.dst);
509
510         kfree(ifp);
511 }
512
513 static void
514 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
515 {
516         struct inet6_ifaddr *ifa, **ifap;
517         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
518
519         /*
520          * Each device address list is sorted in order of scope -
521          * global before linklocal.
522          */
523         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
524              ifap = &ifa->if_next) {
525                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
526                         break;
527         }
528
529         ifp->if_next = *ifap;
530         *ifap = ifp;
531 }
532
533 /* On success it returns ifp with increased reference count */
534
535 static struct inet6_ifaddr *
536 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
537               int scope, u32 flags)
538 {
539         struct inet6_ifaddr *ifa = NULL;
540         struct rt6_info *rt;
541         int hash;
542         int err = 0;
543
544         read_lock_bh(&addrconf_lock);
545         if (idev->dead) {
546                 err = -ENODEV;                  /*XXX*/
547                 goto out2;
548         }
549
550         write_lock(&addrconf_hash_lock);
551
552         /* Ignore adding duplicate addresses on an interface */
553         if (ipv6_chk_same_addr(addr, idev->dev)) {
554                 ADBG(("ipv6_add_addr: already assigned\n"));
555                 err = -EEXIST;
556                 goto out;
557         }
558
559         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
560
561         if (ifa == NULL) {
562                 ADBG(("ipv6_add_addr: malloc failed\n"));
563                 err = -ENOBUFS;
564                 goto out;
565         }
566
567         rt = addrconf_dst_alloc(idev, addr, 0);
568         if (IS_ERR(rt)) {
569                 err = PTR_ERR(rt);
570                 goto out;
571         }
572
573         ipv6_addr_copy(&ifa->addr, addr);
574
575         spin_lock_init(&ifa->lock);
576         init_timer(&ifa->timer);
577         ifa->timer.data = (unsigned long) ifa;
578         ifa->scope = scope;
579         ifa->prefix_len = pfxlen;
580         ifa->flags = flags | IFA_F_TENTATIVE;
581         ifa->cstamp = ifa->tstamp = jiffies;
582
583         ifa->rt = rt;
584
585         ifa->idev = idev;
586         in6_dev_hold(idev);
587         /* For caller */
588         in6_ifa_hold(ifa);
589
590         /* Add to big hash table */
591         hash = ipv6_addr_hash(addr);
592
593         ifa->lst_next = inet6_addr_lst[hash];
594         inet6_addr_lst[hash] = ifa;
595         in6_ifa_hold(ifa);
596         write_unlock(&addrconf_hash_lock);
597
598         write_lock(&idev->lock);
599         /* Add to inet6_dev unicast addr list. */
600         ipv6_link_dev_addr(idev, ifa);
601
602 #ifdef CONFIG_IPV6_PRIVACY
603         if (ifa->flags&IFA_F_TEMPORARY) {
604                 ifa->tmp_next = idev->tempaddr_list;
605                 idev->tempaddr_list = ifa;
606                 in6_ifa_hold(ifa);
607         }
608 #endif
609
610         in6_ifa_hold(ifa);
611         write_unlock(&idev->lock);
612 out2:
613         read_unlock_bh(&addrconf_lock);
614
615         if (likely(err == 0))
616                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
617         else {
618                 kfree(ifa);
619                 ifa = ERR_PTR(err);
620         }
621
622         return ifa;
623 out:
624         write_unlock(&addrconf_hash_lock);
625         goto out2;
626 }
627
628 /* This function wants to get referenced ifp and releases it before return */
629
630 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
631 {
632         struct inet6_ifaddr *ifa, **ifap;
633         struct inet6_dev *idev = ifp->idev;
634         int hash;
635         int deleted = 0, onlink = 0;
636         unsigned long expires = jiffies;
637
638         hash = ipv6_addr_hash(&ifp->addr);
639
640         ifp->dead = 1;
641
642         write_lock_bh(&addrconf_hash_lock);
643         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
644              ifap = &ifa->lst_next) {
645                 if (ifa == ifp) {
646                         *ifap = ifa->lst_next;
647                         __in6_ifa_put(ifp);
648                         ifa->lst_next = NULL;
649                         break;
650                 }
651         }
652         write_unlock_bh(&addrconf_hash_lock);
653
654         write_lock_bh(&idev->lock);
655 #ifdef CONFIG_IPV6_PRIVACY
656         if (ifp->flags&IFA_F_TEMPORARY) {
657                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
658                      ifap = &ifa->tmp_next) {
659                         if (ifa == ifp) {
660                                 *ifap = ifa->tmp_next;
661                                 if (ifp->ifpub) {
662                                         in6_ifa_put(ifp->ifpub);
663                                         ifp->ifpub = NULL;
664                                 }
665                                 __in6_ifa_put(ifp);
666                                 ifa->tmp_next = NULL;
667                                 break;
668                         }
669                 }
670         }
671 #endif
672
673         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
674                 if (ifa == ifp) {
675                         *ifap = ifa->if_next;
676                         __in6_ifa_put(ifp);
677                         ifa->if_next = NULL;
678                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
679                                 break;
680                         deleted = 1;
681                         continue;
682                 } else if (ifp->flags & IFA_F_PERMANENT) {
683                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
684                                               ifp->prefix_len)) {
685                                 if (ifa->flags & IFA_F_PERMANENT) {
686                                         onlink = 1;
687                                         if (deleted)
688                                                 break;
689                                 } else {
690                                         unsigned long lifetime;
691
692                                         if (!onlink)
693                                                 onlink = -1;
694
695                                         spin_lock(&ifa->lock);
696                                         lifetime = min_t(unsigned long,
697                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
698                                         if (time_before(expires,
699                                                         ifa->tstamp + lifetime * HZ))
700                                                 expires = ifa->tstamp + lifetime * HZ;
701                                         spin_unlock(&ifa->lock);
702                                 }
703                         }
704                 }
705                 ifap = &ifa->if_next;
706         }
707         write_unlock_bh(&idev->lock);
708
709         ipv6_ifa_notify(RTM_DELADDR, ifp);
710
711         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
712
713         addrconf_del_timer(ifp);
714
715         /*
716          * Purge or update corresponding prefix
717          *
718          * 1) we don't purge prefix here if address was not permanent.
719          *    prefix is managed by its own lifetime.
720          * 2) if there're no addresses, delete prefix.
721          * 3) if there're still other permanent address(es),
722          *    corresponding prefix is still permanent.
723          * 4) otherwise, update prefix lifetime to the
724          *    longest valid lifetime among the corresponding
725          *    addresses on the device.
726          *    Note: subsequent RA will update lifetime.
727          *
728          * --yoshfuji
729          */
730         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
731                 struct in6_addr prefix;
732                 struct rt6_info *rt;
733
734                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
735                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
736
737                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
738                         if (onlink == 0) {
739                                 ip6_del_rt(rt);
740                                 rt = NULL;
741                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
742                                 rt->rt6i_expires = expires;
743                                 rt->rt6i_flags |= RTF_EXPIRES;
744                         }
745                 }
746                 dst_release(&rt->u.dst);
747         }
748
749         in6_ifa_put(ifp);
750 }
751
752 #ifdef CONFIG_IPV6_PRIVACY
753 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
754 {
755         struct inet6_dev *idev = ifp->idev;
756         struct in6_addr addr, *tmpaddr;
757         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
758         int tmp_plen;
759         int ret = 0;
760         int max_addresses;
761
762         write_lock(&idev->lock);
763         if (ift) {
764                 spin_lock_bh(&ift->lock);
765                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
766                 spin_unlock_bh(&ift->lock);
767                 tmpaddr = &addr;
768         } else {
769                 tmpaddr = NULL;
770         }
771 retry:
772         in6_dev_hold(idev);
773         if (idev->cnf.use_tempaddr <= 0) {
774                 write_unlock(&idev->lock);
775                 printk(KERN_INFO
776                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
777                 in6_dev_put(idev);
778                 ret = -1;
779                 goto out;
780         }
781         spin_lock_bh(&ifp->lock);
782         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
783                 idev->cnf.use_tempaddr = -1;    /*XXX*/
784                 spin_unlock_bh(&ifp->lock);
785                 write_unlock(&idev->lock);
786                 printk(KERN_WARNING
787                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
788                 in6_dev_put(idev);
789                 ret = -1;
790                 goto out;
791         }
792         in6_ifa_hold(ifp);
793         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
794         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
795                 spin_unlock_bh(&ifp->lock);
796                 write_unlock(&idev->lock);
797                 printk(KERN_WARNING
798                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
799                 in6_ifa_put(ifp);
800                 in6_dev_put(idev);
801                 ret = -1;
802                 goto out;
803         }
804         memcpy(&addr.s6_addr[8], idev->rndid, 8);
805         tmp_valid_lft = min_t(__u32,
806                               ifp->valid_lft,
807                               idev->cnf.temp_valid_lft);
808         tmp_prefered_lft = min_t(__u32, 
809                                  ifp->prefered_lft, 
810                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
811         tmp_plen = ifp->prefix_len;
812         max_addresses = idev->cnf.max_addresses;
813         tmp_cstamp = ifp->cstamp;
814         tmp_tstamp = ifp->tstamp;
815         spin_unlock_bh(&ifp->lock);
816
817         write_unlock(&idev->lock);
818         ift = !max_addresses ||
819               ipv6_count_addresses(idev) < max_addresses ? 
820                 ipv6_add_addr(idev, &addr, tmp_plen,
821                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK, IFA_F_TEMPORARY) : NULL;
822         if (!ift || IS_ERR(ift)) {
823                 in6_ifa_put(ifp);
824                 in6_dev_put(idev);
825                 printk(KERN_INFO
826                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
827                 tmpaddr = &addr;
828                 write_lock(&idev->lock);
829                 goto retry;
830         }
831
832         spin_lock_bh(&ift->lock);
833         ift->ifpub = ifp;
834         ift->valid_lft = tmp_valid_lft;
835         ift->prefered_lft = tmp_prefered_lft;
836         ift->cstamp = tmp_cstamp;
837         ift->tstamp = tmp_tstamp;
838         spin_unlock_bh(&ift->lock);
839
840         addrconf_dad_start(ift, 0);
841         in6_ifa_put(ift);
842         in6_dev_put(idev);
843 out:
844         return ret;
845 }
846 #endif
847
848 /*
849  *      Choose an appropriate source address (RFC3484)
850  */
851 struct ipv6_saddr_score {
852         int             addr_type;
853         unsigned int    attrs;
854         int             matchlen;
855         int             scope;
856         unsigned int    rule;
857 };
858
859 #define IPV6_SADDR_SCORE_LOCAL          0x0001
860 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
861 #define IPV6_SADDR_SCORE_HOA            0x0008
862 #define IPV6_SADDR_SCORE_OIF            0x0010
863 #define IPV6_SADDR_SCORE_LABEL          0x0020
864 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
865
866 static int inline ipv6_saddr_preferred(int type)
867 {
868         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
869                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
870                 return 1;
871         return 0;
872 }
873
874 /* static matching label */
875 static int inline ipv6_saddr_label(const struct in6_addr *addr, int type)
876 {
877  /*
878   *     prefix (longest match)  label
879   *     -----------------------------
880   *     ::1/128                 0
881   *     ::/0                    1
882   *     2002::/16               2
883   *     ::/96                   3
884   *     ::ffff:0:0/96           4
885   *     fc00::/7                5
886   *     2001::/32               6
887   */
888         if (type & IPV6_ADDR_LOOPBACK)
889                 return 0;
890         else if (type & IPV6_ADDR_COMPATv4)
891                 return 3;
892         else if (type & IPV6_ADDR_MAPPED)
893                 return 4;
894         else if (addr->s6_addr32[0] == htonl(0x20010000))
895                 return 6;
896         else if (addr->s6_addr16[0] == htons(0x2002))
897                 return 2;
898         else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
899                 return 5;
900         return 1;
901 }
902
903 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
904                        struct in6_addr *daddr, struct in6_addr *saddr)
905 {
906         struct ipv6_saddr_score hiscore;
907         struct inet6_ifaddr *ifa_result = NULL;
908         int daddr_type = __ipv6_addr_type(daddr);
909         int daddr_scope = __ipv6_addr_src_scope(daddr_type);
910         u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
911         struct net_device *dev;
912
913         memset(&hiscore, 0, sizeof(hiscore));
914
915         read_lock(&dev_base_lock);
916         read_lock(&addrconf_lock);
917
918         for (dev = dev_base; dev; dev=dev->next) {
919                 struct inet6_dev *idev;
920                 struct inet6_ifaddr *ifa;
921
922                 /* Rule 0: Candidate Source Address (section 4)
923                  *  - multicast and link-local destination address,
924                  *    the set of candidate source address MUST only
925                  *    include addresses assigned to interfaces
926                  *    belonging to the same link as the outgoing
927                  *    interface.
928                  * (- For site-local destination addresses, the
929                  *    set of candidate source addresses MUST only
930                  *    include addresses assigned to interfaces
931                  *    belonging to the same site as the outgoing
932                  *    interface.)
933                  */
934                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
935                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
936                     daddr_dev && dev != daddr_dev)
937                         continue;
938
939                 idev = __in6_dev_get(dev);
940                 if (!idev)
941                         continue;
942
943                 read_lock_bh(&idev->lock);
944                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
945                         struct ipv6_saddr_score score;
946
947                         score.addr_type = __ipv6_addr_type(&ifa->addr);
948
949                         /* Rule 0:
950                          * - Tentative Address (RFC2462 section 5.4)
951                          *  - A tentative address is not considered
952                          *    "assigned to an interface" in the traditional
953                          *    sense.
954                          * - Candidate Source Address (section 4)
955                          *  - In any case, anycast addresses, multicast
956                          *    addresses, and the unspecified address MUST
957                          *    NOT be included in a candidate set.
958                          */
959                         if (ifa->flags & IFA_F_TENTATIVE)
960                                 continue;
961                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
962                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
963                                 LIMIT_NETDEBUG(KERN_DEBUG
964                                                "ADDRCONF: unspecified / multicast address"
965                                                "assigned as unicast address on %s",
966                                                dev->name);
967                                 continue;
968                         }
969
970                         score.attrs = 0;
971                         score.matchlen = 0;
972                         score.scope = 0;
973                         score.rule = 0;
974
975                         if (ifa_result == NULL) {
976                                 /* record it if the first available entry */
977                                 goto record_it;
978                         }
979
980                         /* Rule 1: Prefer same address */
981                         if (hiscore.rule < 1) {
982                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
983                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
984                                 hiscore.rule++;
985                         }
986                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
987                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
988                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
989                                         score.rule = 1;
990                                         goto record_it;
991                                 }
992                         } else {
993                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
994                                         continue;
995                         }
996
997                         /* Rule 2: Prefer appropriate scope */
998                         if (hiscore.rule < 2) {
999                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1000                                 hiscore.rule++;
1001                         }
1002                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1003                         if (hiscore.scope < score.scope) {
1004                                 if (hiscore.scope < daddr_scope) {
1005                                         score.rule = 2;
1006                                         goto record_it;
1007                                 } else
1008                                         continue;
1009                         } else if (score.scope < hiscore.scope) {
1010                                 if (score.scope < daddr_scope)
1011                                         break; /* addresses sorted by scope */
1012                                 else {
1013                                         score.rule = 2;
1014                                         goto record_it;
1015                                 }
1016                         }
1017
1018                         /* Rule 3: Avoid deprecated address */
1019                         if (hiscore.rule < 3) {
1020                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1021                                     !(ifa_result->flags & IFA_F_DEPRECATED))
1022                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1023                                 hiscore.rule++;
1024                         }
1025                         if (ipv6_saddr_preferred(score.addr_type) ||
1026                             !(ifa->flags & IFA_F_DEPRECATED)) {
1027                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1028                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1029                                         score.rule = 3;
1030                                         goto record_it;
1031                                 }
1032                         } else {
1033                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1034                                         continue;
1035                         }
1036
1037                         /* Rule 4: Prefer home address -- not implemented yet */
1038                         if (hiscore.rule < 4)
1039                                 hiscore.rule++;
1040
1041                         /* Rule 5: Prefer outgoing interface */
1042                         if (hiscore.rule < 5) {
1043                                 if (daddr_dev == NULL ||
1044                                     daddr_dev == ifa_result->idev->dev)
1045                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1046                                 hiscore.rule++;
1047                         }
1048                         if (daddr_dev == NULL ||
1049                             daddr_dev == ifa->idev->dev) {
1050                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1051                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1052                                         score.rule = 5;
1053                                         goto record_it;
1054                                 }
1055                         } else {
1056                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1057                                         continue;
1058                         }
1059
1060                         /* Rule 6: Prefer matching label */
1061                         if (hiscore.rule < 6) {
1062                                 if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1063                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1064                                 hiscore.rule++;
1065                         }
1066                         if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1067                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1068                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1069                                         score.rule = 6;
1070                                         goto record_it;
1071                                 }
1072                         } else {
1073                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1074                                         continue;
1075                         }
1076
1077 #ifdef CONFIG_IPV6_PRIVACY
1078                         /* Rule 7: Prefer public address
1079                          * Note: prefer temprary address if use_tempaddr >= 2
1080                          */
1081                         if (hiscore.rule < 7) {
1082                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1083                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1084                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1085                                 hiscore.rule++;
1086                         }
1087                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1088                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1089                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1090                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1091                                         score.rule = 7;
1092                                         goto record_it;
1093                                 }
1094                         } else {
1095                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1096                                         continue;
1097                         }
1098 #else
1099                         if (hiscore.rule < 7)
1100                                 hiscore.rule++;
1101 #endif
1102                         /* Rule 8: Use longest matching prefix */
1103                         if (hiscore.rule < 8) {
1104                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1105                                 hiscore.rule++;
1106                         }
1107                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1108                         if (score.matchlen > hiscore.matchlen) {
1109                                 score.rule = 8;
1110                                 goto record_it;
1111                         }
1112 #if 0
1113                         else if (score.matchlen < hiscore.matchlen)
1114                                 continue;
1115 #endif
1116
1117                         /* Final Rule: choose first available one */
1118                         continue;
1119 record_it:
1120                         if (ifa_result)
1121                                 in6_ifa_put(ifa_result);
1122                         in6_ifa_hold(ifa);
1123                         ifa_result = ifa;
1124                         hiscore = score;
1125                 }
1126                 read_unlock_bh(&idev->lock);
1127         }
1128         read_unlock(&addrconf_lock);
1129         read_unlock(&dev_base_lock);
1130
1131         if (!ifa_result)
1132                 return -EADDRNOTAVAIL;
1133         
1134         ipv6_addr_copy(saddr, &ifa_result->addr);
1135         in6_ifa_put(ifa_result);
1136         return 0;
1137 }
1138
1139
1140 int ipv6_get_saddr(struct dst_entry *dst,
1141                    struct in6_addr *daddr, struct in6_addr *saddr)
1142 {
1143         return ipv6_dev_get_saddr(dst ? ((struct rt6_info *)dst)->rt6i_idev->dev : NULL, daddr, saddr);
1144 }
1145
1146
1147 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
1148 {
1149         struct inet6_dev *idev;
1150         int err = -EADDRNOTAVAIL;
1151
1152         read_lock(&addrconf_lock);
1153         if ((idev = __in6_dev_get(dev)) != NULL) {
1154                 struct inet6_ifaddr *ifp;
1155
1156                 read_lock_bh(&idev->lock);
1157                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1158                         if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1159                                 ipv6_addr_copy(addr, &ifp->addr);
1160                                 err = 0;
1161                                 break;
1162                         }
1163                 }
1164                 read_unlock_bh(&idev->lock);
1165         }
1166         read_unlock(&addrconf_lock);
1167         return err;
1168 }
1169
1170 static int ipv6_count_addresses(struct inet6_dev *idev)
1171 {
1172         int cnt = 0;
1173         struct inet6_ifaddr *ifp;
1174
1175         read_lock_bh(&idev->lock);
1176         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1177                 cnt++;
1178         read_unlock_bh(&idev->lock);
1179         return cnt;
1180 }
1181
1182 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1183 {
1184         struct inet6_ifaddr * ifp;
1185         u8 hash = ipv6_addr_hash(addr);
1186
1187         read_lock_bh(&addrconf_hash_lock);
1188         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1189                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1190                     !(ifp->flags&IFA_F_TENTATIVE)) {
1191                         if (dev == NULL || ifp->idev->dev == dev ||
1192                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1193                                 break;
1194                 }
1195         }
1196         read_unlock_bh(&addrconf_hash_lock);
1197         return ifp != NULL;
1198 }
1199
1200 static
1201 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1202 {
1203         struct inet6_ifaddr * ifp;
1204         u8 hash = ipv6_addr_hash(addr);
1205
1206         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1207                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1208                         if (dev == NULL || ifp->idev->dev == dev)
1209                                 break;
1210                 }
1211         }
1212         return ifp != NULL;
1213 }
1214
1215 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1216 {
1217         struct inet6_ifaddr * ifp;
1218         u8 hash = ipv6_addr_hash(addr);
1219
1220         read_lock_bh(&addrconf_hash_lock);
1221         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1222                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1223                         if (dev == NULL || ifp->idev->dev == dev ||
1224                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1225                                 in6_ifa_hold(ifp);
1226                                 break;
1227                         }
1228                 }
1229         }
1230         read_unlock_bh(&addrconf_hash_lock);
1231
1232         return ifp;
1233 }
1234
1235 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1236 {
1237         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1238         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1239         u32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1240         u32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1241         int sk_ipv6only = ipv6_only_sock(sk);
1242         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1243         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1244         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1245
1246         if (!sk2_rcv_saddr && !sk_ipv6only)
1247                 return 1;
1248
1249         if (addr_type2 == IPV6_ADDR_ANY &&
1250             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1251                 return 1;
1252
1253         if (addr_type == IPV6_ADDR_ANY &&
1254             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1255                 return 1;
1256
1257         if (sk2_rcv_saddr6 &&
1258             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1259                 return 1;
1260
1261         if (addr_type == IPV6_ADDR_MAPPED &&
1262             !sk2_ipv6only &&
1263             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1264                 return 1;
1265
1266         return 0;
1267 }
1268
1269 /* Gets referenced address, destroys ifaddr */
1270
1271 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1272 {
1273         if (ifp->flags&IFA_F_PERMANENT) {
1274                 spin_lock_bh(&ifp->lock);
1275                 addrconf_del_timer(ifp);
1276                 ifp->flags |= IFA_F_TENTATIVE;
1277                 spin_unlock_bh(&ifp->lock);
1278                 in6_ifa_put(ifp);
1279 #ifdef CONFIG_IPV6_PRIVACY
1280         } else if (ifp->flags&IFA_F_TEMPORARY) {
1281                 struct inet6_ifaddr *ifpub;
1282                 spin_lock_bh(&ifp->lock);
1283                 ifpub = ifp->ifpub;
1284                 if (ifpub) {
1285                         in6_ifa_hold(ifpub);
1286                         spin_unlock_bh(&ifp->lock);
1287                         ipv6_create_tempaddr(ifpub, ifp);
1288                         in6_ifa_put(ifpub);
1289                 } else {
1290                         spin_unlock_bh(&ifp->lock);
1291                 }
1292                 ipv6_del_addr(ifp);
1293 #endif
1294         } else
1295                 ipv6_del_addr(ifp);
1296 }
1297
1298 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1299 {
1300         if (net_ratelimit())
1301                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1302         addrconf_dad_stop(ifp);
1303 }
1304
1305 /* Join to solicited addr multicast group. */
1306
1307 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1308 {
1309         struct in6_addr maddr;
1310
1311         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1312                 return;
1313
1314         addrconf_addr_solict_mult(addr, &maddr);
1315         ipv6_dev_mc_inc(dev, &maddr);
1316 }
1317
1318 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1319 {
1320         struct in6_addr maddr;
1321
1322         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1323                 return;
1324
1325         addrconf_addr_solict_mult(addr, &maddr);
1326         __ipv6_dev_mc_dec(idev, &maddr);
1327 }
1328
1329 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1330 {
1331         struct in6_addr addr;
1332         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1333         if (ipv6_addr_any(&addr))
1334                 return;
1335         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1336 }
1337
1338 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1339 {
1340         struct in6_addr addr;
1341         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1342         if (ipv6_addr_any(&addr))
1343                 return;
1344         __ipv6_dev_ac_dec(ifp->idev, &addr);
1345 }
1346
1347 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1348 {
1349         if (dev->addr_len != ETH_ALEN)
1350                 return -1;
1351         memcpy(eui, dev->dev_addr, 3);
1352         memcpy(eui + 5, dev->dev_addr + 3, 3);
1353
1354         /*
1355          * The zSeries OSA network cards can be shared among various
1356          * OS instances, but the OSA cards have only one MAC address.
1357          * This leads to duplicate address conflicts in conjunction
1358          * with IPv6 if more than one instance uses the same card.
1359          *
1360          * The driver for these cards can deliver a unique 16-bit
1361          * identifier for each instance sharing the same card.  It is
1362          * placed instead of 0xFFFE in the interface identifier.  The
1363          * "u" bit of the interface identifier is not inverted in this
1364          * case.  Hence the resulting interface identifier has local
1365          * scope according to RFC2373.
1366          */
1367         if (dev->dev_id) {
1368                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1369                 eui[4] = dev->dev_id & 0xFF;
1370         } else {
1371                 eui[3] = 0xFF;
1372                 eui[4] = 0xFE;
1373                 eui[0] ^= 2;
1374         }
1375         return 0;
1376 }
1377
1378 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1379 {
1380         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1381         if (dev->addr_len != ARCNET_ALEN)
1382                 return -1;
1383         memset(eui, 0, 7);
1384         eui[7] = *(u8*)dev->dev_addr;
1385         return 0;
1386 }
1387
1388 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1389 {
1390         if (dev->addr_len != INFINIBAND_ALEN)
1391                 return -1;
1392         memcpy(eui, dev->dev_addr + 12, 8);
1393         eui[0] |= 2;
1394         return 0;
1395 }
1396
1397 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1398 {
1399         switch (dev->type) {
1400         case ARPHRD_ETHER:
1401         case ARPHRD_FDDI:
1402         case ARPHRD_IEEE802_TR:
1403                 return addrconf_ifid_eui48(eui, dev);
1404         case ARPHRD_ARCNET:
1405                 return addrconf_ifid_arcnet(eui, dev);
1406         case ARPHRD_INFINIBAND:
1407                 return addrconf_ifid_infiniband(eui, dev);
1408         }
1409         return -1;
1410 }
1411
1412 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1413 {
1414         int err = -1;
1415         struct inet6_ifaddr *ifp;
1416
1417         read_lock_bh(&idev->lock);
1418         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1419                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1420                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1421                         err = 0;
1422                         break;
1423                 }
1424         }
1425         read_unlock_bh(&idev->lock);
1426         return err;
1427 }
1428
1429 #ifdef CONFIG_IPV6_PRIVACY
1430 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1431 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1432 {
1433 regen:
1434         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1435         idev->rndid[0] &= ~0x02;
1436
1437         /*
1438          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1439          * check if generated address is not inappropriate
1440          *
1441          *  - Reserved subnet anycast (RFC 2526)
1442          *      11111101 11....11 1xxxxxxx
1443          *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1444          *      00-00-5E-FE-xx-xx-xx-xx
1445          *  - value 0
1446          *  - XXX: already assigned to an address on the device
1447          */
1448         if (idev->rndid[0] == 0xfd && 
1449             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1450             (idev->rndid[7]&0x80))
1451                 goto regen;
1452         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1453                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1454                         goto regen;
1455                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1456                         goto regen;
1457         }
1458
1459         return 0;
1460 }
1461
1462 static void ipv6_regen_rndid(unsigned long data)
1463 {
1464         struct inet6_dev *idev = (struct inet6_dev *) data;
1465         unsigned long expires;
1466
1467         read_lock_bh(&addrconf_lock);
1468         write_lock_bh(&idev->lock);
1469
1470         if (idev->dead)
1471                 goto out;
1472
1473         if (__ipv6_regen_rndid(idev) < 0)
1474                 goto out;
1475         
1476         expires = jiffies +
1477                 idev->cnf.temp_prefered_lft * HZ - 
1478                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1479         if (time_before(expires, jiffies)) {
1480                 printk(KERN_WARNING
1481                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1482                         idev->dev->name);
1483                 goto out;
1484         }
1485
1486         if (!mod_timer(&idev->regen_timer, expires))
1487                 in6_dev_hold(idev);
1488
1489 out:
1490         write_unlock_bh(&idev->lock);
1491         read_unlock_bh(&addrconf_lock);
1492         in6_dev_put(idev);
1493 }
1494
1495 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1496         int ret = 0;
1497
1498         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1499                 ret = __ipv6_regen_rndid(idev);
1500         return ret;
1501 }
1502 #endif
1503
1504 /*
1505  *      Add prefix route.
1506  */
1507
1508 static void
1509 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1510                       unsigned long expires, u32 flags)
1511 {
1512         struct fib6_config cfg = {
1513                 .fc_table = RT6_TABLE_PREFIX,
1514                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1515                 .fc_ifindex = dev->ifindex,
1516                 .fc_expires = expires,
1517                 .fc_dst_len = plen,
1518                 .fc_flags = RTF_UP | flags,
1519         };
1520
1521         ipv6_addr_copy(&cfg.fc_dst, pfx);
1522
1523         /* Prevent useless cloning on PtP SIT.
1524            This thing is done here expecting that the whole
1525            class of non-broadcast devices need not cloning.
1526          */
1527         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1528                 cfg.fc_flags |= RTF_NONEXTHOP;
1529
1530         ip6_route_add(&cfg);
1531 }
1532
1533 /* Create "default" multicast route to the interface */
1534
1535 static void addrconf_add_mroute(struct net_device *dev)
1536 {
1537         struct fib6_config cfg = {
1538                 .fc_table = RT6_TABLE_LOCAL,
1539                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1540                 .fc_ifindex = dev->ifindex,
1541                 .fc_dst_len = 8,
1542                 .fc_flags = RTF_UP,
1543         };
1544
1545         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1546
1547         ip6_route_add(&cfg);
1548 }
1549
1550 static void sit_route_add(struct net_device *dev)
1551 {
1552         struct fib6_config cfg = {
1553                 .fc_table = RT6_TABLE_MAIN,
1554                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1555                 .fc_ifindex = dev->ifindex,
1556                 .fc_dst_len = 96,
1557                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1558         };
1559
1560         /* prefix length - 96 bits "::d.d.d.d" */
1561         ip6_route_add(&cfg);
1562 }
1563
1564 static void addrconf_add_lroute(struct net_device *dev)
1565 {
1566         struct in6_addr addr;
1567
1568         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1569         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1570 }
1571
1572 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1573 {
1574         struct inet6_dev *idev;
1575
1576         ASSERT_RTNL();
1577
1578         if ((idev = ipv6_find_idev(dev)) == NULL)
1579                 return NULL;
1580
1581         /* Add default multicast route */
1582         addrconf_add_mroute(dev);
1583
1584         /* Add link local route */
1585         addrconf_add_lroute(dev);
1586         return idev;
1587 }
1588
1589 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1590 {
1591         struct prefix_info *pinfo;
1592         __u32 valid_lft;
1593         __u32 prefered_lft;
1594         int addr_type;
1595         unsigned long rt_expires;
1596         struct inet6_dev *in6_dev;
1597
1598         pinfo = (struct prefix_info *) opt;
1599         
1600         if (len < sizeof(struct prefix_info)) {
1601                 ADBG(("addrconf: prefix option too short\n"));
1602                 return;
1603         }
1604         
1605         /*
1606          *      Validation checks ([ADDRCONF], page 19)
1607          */
1608
1609         addr_type = ipv6_addr_type(&pinfo->prefix);
1610
1611         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1612                 return;
1613
1614         valid_lft = ntohl(pinfo->valid);
1615         prefered_lft = ntohl(pinfo->prefered);
1616
1617         if (prefered_lft > valid_lft) {
1618                 if (net_ratelimit())
1619                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1620                 return;
1621         }
1622
1623         in6_dev = in6_dev_get(dev);
1624
1625         if (in6_dev == NULL) {
1626                 if (net_ratelimit())
1627                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1628                 return;
1629         }
1630
1631         /*
1632          *      Two things going on here:
1633          *      1) Add routes for on-link prefixes
1634          *      2) Configure prefixes with the auto flag set
1635          */
1636
1637         /* Avoid arithmetic overflow. Really, we could
1638            save rt_expires in seconds, likely valid_lft,
1639            but it would require division in fib gc, that it
1640            not good.
1641          */
1642         if (valid_lft >= 0x7FFFFFFF/HZ)
1643                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1644         else
1645                 rt_expires = valid_lft * HZ;
1646
1647         /*
1648          * We convert this (in jiffies) to clock_t later.
1649          * Avoid arithmetic overflow there as well.
1650          * Overflow can happen only if HZ < USER_HZ.
1651          */
1652         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1653                 rt_expires = 0x7FFFFFFF / USER_HZ;
1654
1655         if (pinfo->onlink) {
1656                 struct rt6_info *rt;
1657                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1658
1659                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1660                         if (rt->rt6i_flags&RTF_EXPIRES) {
1661                                 if (valid_lft == 0) {
1662                                         ip6_del_rt(rt);
1663                                         rt = NULL;
1664                                 } else {
1665                                         rt->rt6i_expires = jiffies + rt_expires;
1666                                 }
1667                         }
1668                 } else if (valid_lft) {
1669                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1670                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1671                 }
1672                 if (rt)
1673                         dst_release(&rt->u.dst);
1674         }
1675
1676         /* Try to figure out our local address for this prefix */
1677
1678         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1679                 struct inet6_ifaddr * ifp;
1680                 struct in6_addr addr;
1681                 int create = 0, update_lft = 0;
1682
1683                 if (pinfo->prefix_len == 64) {
1684                         memcpy(&addr, &pinfo->prefix, 8);
1685                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1686                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1687                                 in6_dev_put(in6_dev);
1688                                 return;
1689                         }
1690                         goto ok;
1691                 }
1692                 if (net_ratelimit())
1693                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1694                                pinfo->prefix_len);
1695                 in6_dev_put(in6_dev);
1696                 return;
1697
1698 ok:
1699
1700                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1701
1702                 if (ifp == NULL && valid_lft) {
1703                         int max_addresses = in6_dev->cnf.max_addresses;
1704
1705                         /* Do not allow to create too much of autoconfigured
1706                          * addresses; this would be too easy way to crash kernel.
1707                          */
1708                         if (!max_addresses ||
1709                             ipv6_count_addresses(in6_dev) < max_addresses)
1710                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1711                                                     addr_type&IPV6_ADDR_SCOPE_MASK, 0);
1712
1713                         if (!ifp || IS_ERR(ifp)) {
1714                                 in6_dev_put(in6_dev);
1715                                 return;
1716                         }
1717
1718                         update_lft = create = 1;
1719                         ifp->cstamp = jiffies;
1720                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1721                 }
1722
1723                 if (ifp) {
1724                         int flags;
1725                         unsigned long now;
1726 #ifdef CONFIG_IPV6_PRIVACY
1727                         struct inet6_ifaddr *ift;
1728 #endif
1729                         u32 stored_lft;
1730
1731                         /* update lifetime (RFC2462 5.5.3 e) */
1732                         spin_lock(&ifp->lock);
1733                         now = jiffies;
1734                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1735                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1736                         else
1737                                 stored_lft = 0;
1738                         if (!update_lft && stored_lft) {
1739                                 if (valid_lft > MIN_VALID_LIFETIME ||
1740                                     valid_lft > stored_lft)
1741                                         update_lft = 1;
1742                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1743                                         /* valid_lft <= stored_lft is always true */
1744                                         /* XXX: IPsec */
1745                                         update_lft = 0;
1746                                 } else {
1747                                         valid_lft = MIN_VALID_LIFETIME;
1748                                         if (valid_lft < prefered_lft)
1749                                                 prefered_lft = valid_lft;
1750                                         update_lft = 1;
1751                                 }
1752                         }
1753
1754                         if (update_lft) {
1755                                 ifp->valid_lft = valid_lft;
1756                                 ifp->prefered_lft = prefered_lft;
1757                                 ifp->tstamp = now;
1758                                 flags = ifp->flags;
1759                                 ifp->flags &= ~IFA_F_DEPRECATED;
1760                                 spin_unlock(&ifp->lock);
1761
1762                                 if (!(flags&IFA_F_TENTATIVE))
1763                                         ipv6_ifa_notify(0, ifp);
1764                         } else
1765                                 spin_unlock(&ifp->lock);
1766
1767 #ifdef CONFIG_IPV6_PRIVACY
1768                         read_lock_bh(&in6_dev->lock);
1769                         /* update all temporary addresses in the list */
1770                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1771                                 /*
1772                                  * When adjusting the lifetimes of an existing
1773                                  * temporary address, only lower the lifetimes.
1774                                  * Implementations must not increase the
1775                                  * lifetimes of an existing temporary address
1776                                  * when processing a Prefix Information Option.
1777                                  */
1778                                 spin_lock(&ift->lock);
1779                                 flags = ift->flags;
1780                                 if (ift->valid_lft > valid_lft &&
1781                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1782                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1783                                 if (ift->prefered_lft > prefered_lft &&
1784                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1785                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1786                                 spin_unlock(&ift->lock);
1787                                 if (!(flags&IFA_F_TENTATIVE))
1788                                         ipv6_ifa_notify(0, ift);
1789                         }
1790
1791                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1792                                 /*
1793                                  * When a new public address is created as described in [ADDRCONF],
1794                                  * also create a new temporary address.
1795                                  */
1796                                 read_unlock_bh(&in6_dev->lock); 
1797                                 ipv6_create_tempaddr(ifp, NULL);
1798                         } else {
1799                                 read_unlock_bh(&in6_dev->lock);
1800                         }
1801 #endif
1802                         in6_ifa_put(ifp);
1803                         addrconf_verify(0);
1804                 }
1805         }
1806         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1807         in6_dev_put(in6_dev);
1808 }
1809
1810 /*
1811  *      Set destination address.
1812  *      Special case for SIT interfaces where we create a new "virtual"
1813  *      device.
1814  */
1815 int addrconf_set_dstaddr(void __user *arg)
1816 {
1817         struct in6_ifreq ireq;
1818         struct net_device *dev;
1819         int err = -EINVAL;
1820
1821         rtnl_lock();
1822
1823         err = -EFAULT;
1824         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1825                 goto err_exit;
1826
1827         dev = __dev_get_by_index(ireq.ifr6_ifindex);
1828
1829         err = -ENODEV;
1830         if (dev == NULL)
1831                 goto err_exit;
1832
1833         if (dev->type == ARPHRD_SIT) {
1834                 struct ifreq ifr;
1835                 mm_segment_t    oldfs;
1836                 struct ip_tunnel_parm p;
1837
1838                 err = -EADDRNOTAVAIL;
1839                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1840                         goto err_exit;
1841
1842                 memset(&p, 0, sizeof(p));
1843                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1844                 p.iph.saddr = 0;
1845                 p.iph.version = 4;
1846                 p.iph.ihl = 5;
1847                 p.iph.protocol = IPPROTO_IPV6;
1848                 p.iph.ttl = 64;
1849                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1850
1851                 oldfs = get_fs(); set_fs(KERNEL_DS);
1852                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1853                 set_fs(oldfs);
1854
1855                 if (err == 0) {
1856                         err = -ENOBUFS;
1857                         if ((dev = __dev_get_by_name(p.name)) == NULL)
1858                                 goto err_exit;
1859                         err = dev_open(dev);
1860                 }
1861         }
1862
1863 err_exit:
1864         rtnl_unlock();
1865         return err;
1866 }
1867
1868 /*
1869  *      Manual configuration of address on an interface
1870  */
1871 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1872                           __u32 prefered_lft, __u32 valid_lft)
1873 {
1874         struct inet6_ifaddr *ifp;
1875         struct inet6_dev *idev;
1876         struct net_device *dev;
1877         __u8 ifa_flags = 0;
1878         int scope;
1879
1880         ASSERT_RTNL();
1881         
1882         /* check the lifetime */
1883         if (!valid_lft || prefered_lft > valid_lft)
1884                 return -EINVAL;
1885
1886         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1887                 return -ENODEV;
1888         
1889         if (!(dev->flags&IFF_UP))
1890                 return -ENETDOWN;
1891
1892         if ((idev = addrconf_add_dev(dev)) == NULL)
1893                 return -ENOBUFS;
1894
1895         scope = ipv6_addr_scope(pfx);
1896
1897         if (valid_lft == INFINITY_LIFE_TIME)
1898                 ifa_flags |= IFA_F_PERMANENT;
1899         else if (valid_lft >= 0x7FFFFFFF/HZ)
1900                 valid_lft = 0x7FFFFFFF/HZ;
1901
1902         if (prefered_lft == 0)
1903                 ifa_flags |= IFA_F_DEPRECATED;
1904         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1905                  (prefered_lft != INFINITY_LIFE_TIME))
1906                 prefered_lft = 0x7FFFFFFF/HZ;
1907
1908         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1909
1910         if (!IS_ERR(ifp)) {
1911                 spin_lock_bh(&ifp->lock);
1912                 ifp->valid_lft = valid_lft;
1913                 ifp->prefered_lft = prefered_lft;
1914                 ifp->tstamp = jiffies;
1915                 spin_unlock_bh(&ifp->lock);
1916
1917                 addrconf_dad_start(ifp, 0);
1918                 in6_ifa_put(ifp);
1919                 addrconf_verify(0);
1920                 return 0;
1921         }
1922
1923         return PTR_ERR(ifp);
1924 }
1925
1926 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1927 {
1928         struct inet6_ifaddr *ifp;
1929         struct inet6_dev *idev;
1930         struct net_device *dev;
1931         
1932         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1933                 return -ENODEV;
1934
1935         if ((idev = __in6_dev_get(dev)) == NULL)
1936                 return -ENXIO;
1937
1938         read_lock_bh(&idev->lock);
1939         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1940                 if (ifp->prefix_len == plen &&
1941                     ipv6_addr_equal(pfx, &ifp->addr)) {
1942                         in6_ifa_hold(ifp);
1943                         read_unlock_bh(&idev->lock);
1944                         
1945                         ipv6_del_addr(ifp);
1946
1947                         /* If the last address is deleted administratively,
1948                            disable IPv6 on this interface.
1949                          */
1950                         if (idev->addr_list == NULL)
1951                                 addrconf_ifdown(idev->dev, 1);
1952                         return 0;
1953                 }
1954         }
1955         read_unlock_bh(&idev->lock);
1956         return -EADDRNOTAVAIL;
1957 }
1958
1959
1960 int addrconf_add_ifaddr(void __user *arg)
1961 {
1962         struct in6_ifreq ireq;
1963         int err;
1964         
1965         if (!capable(CAP_NET_ADMIN))
1966                 return -EPERM;
1967         
1968         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1969                 return -EFAULT;
1970
1971         rtnl_lock();
1972         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
1973                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1974         rtnl_unlock();
1975         return err;
1976 }
1977
1978 int addrconf_del_ifaddr(void __user *arg)
1979 {
1980         struct in6_ifreq ireq;
1981         int err;
1982         
1983         if (!capable(CAP_NET_ADMIN))
1984                 return -EPERM;
1985
1986         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1987                 return -EFAULT;
1988
1989         rtnl_lock();
1990         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
1991         rtnl_unlock();
1992         return err;
1993 }
1994
1995 static void sit_add_v4_addrs(struct inet6_dev *idev)
1996 {
1997         struct inet6_ifaddr * ifp;
1998         struct in6_addr addr;
1999         struct net_device *dev;
2000         int scope;
2001
2002         ASSERT_RTNL();
2003
2004         memset(&addr, 0, sizeof(struct in6_addr));
2005         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2006
2007         if (idev->dev->flags&IFF_POINTOPOINT) {
2008                 addr.s6_addr32[0] = htonl(0xfe800000);
2009                 scope = IFA_LINK;
2010         } else {
2011                 scope = IPV6_ADDR_COMPATv4;
2012         }
2013
2014         if (addr.s6_addr32[3]) {
2015                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2016                 if (!IS_ERR(ifp)) {
2017                         spin_lock_bh(&ifp->lock);
2018                         ifp->flags &= ~IFA_F_TENTATIVE;
2019                         spin_unlock_bh(&ifp->lock);
2020                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2021                         in6_ifa_put(ifp);
2022                 }
2023                 return;
2024         }
2025
2026         for (dev = dev_base; dev != NULL; dev = dev->next) {
2027                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2028                 if (in_dev && (dev->flags & IFF_UP)) {
2029                         struct in_ifaddr * ifa;
2030
2031                         int flag = scope;
2032
2033                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2034                                 int plen;
2035
2036                                 addr.s6_addr32[3] = ifa->ifa_local;
2037
2038                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2039                                         continue;
2040                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2041                                         if (idev->dev->flags&IFF_POINTOPOINT)
2042                                                 continue;
2043                                         flag |= IFA_HOST;
2044                                 }
2045                                 if (idev->dev->flags&IFF_POINTOPOINT)
2046                                         plen = 64;
2047                                 else
2048                                         plen = 96;
2049
2050                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2051                                                     IFA_F_PERMANENT);
2052                                 if (!IS_ERR(ifp)) {
2053                                         spin_lock_bh(&ifp->lock);
2054                                         ifp->flags &= ~IFA_F_TENTATIVE;
2055                                         spin_unlock_bh(&ifp->lock);
2056                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2057                                         in6_ifa_put(ifp);
2058                                 }
2059                         }
2060                 }
2061         }
2062 }
2063
2064 static void init_loopback(struct net_device *dev)
2065 {
2066         struct inet6_dev  *idev;
2067         struct inet6_ifaddr * ifp;
2068
2069         /* ::1 */
2070
2071         ASSERT_RTNL();
2072
2073         if ((idev = ipv6_find_idev(dev)) == NULL) {
2074                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2075                 return;
2076         }
2077
2078         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2079         if (!IS_ERR(ifp)) {
2080                 spin_lock_bh(&ifp->lock);
2081                 ifp->flags &= ~IFA_F_TENTATIVE;
2082                 spin_unlock_bh(&ifp->lock);
2083                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2084                 in6_ifa_put(ifp);
2085         }
2086 }
2087
2088 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2089 {
2090         struct inet6_ifaddr * ifp;
2091
2092         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, IFA_F_PERMANENT);
2093         if (!IS_ERR(ifp)) {
2094                 addrconf_dad_start(ifp, 0);
2095                 in6_ifa_put(ifp);
2096         }
2097 }
2098
2099 static void addrconf_dev_config(struct net_device *dev)
2100 {
2101         struct in6_addr addr;
2102         struct inet6_dev    * idev;
2103
2104         ASSERT_RTNL();
2105
2106         if ((dev->type != ARPHRD_ETHER) && 
2107             (dev->type != ARPHRD_FDDI) &&
2108             (dev->type != ARPHRD_IEEE802_TR) &&
2109             (dev->type != ARPHRD_ARCNET) &&
2110             (dev->type != ARPHRD_INFINIBAND)) {
2111                 /* Alas, we support only Ethernet autoconfiguration. */
2112                 return;
2113         }
2114
2115         idev = addrconf_add_dev(dev);
2116         if (idev == NULL)
2117                 return;
2118
2119         memset(&addr, 0, sizeof(struct in6_addr));
2120         addr.s6_addr32[0] = htonl(0xFE800000);
2121
2122         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2123                 addrconf_add_linklocal(idev, &addr);
2124 }
2125
2126 static void addrconf_sit_config(struct net_device *dev)
2127 {
2128         struct inet6_dev *idev;
2129
2130         ASSERT_RTNL();
2131
2132         /* 
2133          * Configure the tunnel with one of our IPv4 
2134          * addresses... we should configure all of 
2135          * our v4 addrs in the tunnel
2136          */
2137
2138         if ((idev = ipv6_find_idev(dev)) == NULL) {
2139                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2140                 return;
2141         }
2142
2143         sit_add_v4_addrs(idev);
2144
2145         if (dev->flags&IFF_POINTOPOINT) {
2146                 addrconf_add_mroute(dev);
2147                 addrconf_add_lroute(dev);
2148         } else
2149                 sit_route_add(dev);
2150 }
2151
2152 static inline int
2153 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2154 {
2155         struct in6_addr lladdr;
2156
2157         if (!ipv6_get_lladdr(link_dev, &lladdr)) {
2158                 addrconf_add_linklocal(idev, &lladdr);
2159                 return 0;
2160         }
2161         return -1;
2162 }
2163
2164 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2165 {
2166         struct net_device *link_dev;
2167
2168         /* first try to inherit the link-local address from the link device */
2169         if (idev->dev->iflink &&
2170             (link_dev = __dev_get_by_index(idev->dev->iflink))) {
2171                 if (!ipv6_inherit_linklocal(idev, link_dev))
2172                         return;
2173         }
2174         /* then try to inherit it from any device */
2175         for (link_dev = dev_base; link_dev; link_dev = link_dev->next) {
2176                 if (!ipv6_inherit_linklocal(idev, link_dev))
2177                         return;
2178         }
2179         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2180 }
2181
2182 /*
2183  * Autoconfigure tunnel with a link-local address so routing protocols,
2184  * DHCPv6, MLD etc. can be run over the virtual link
2185  */
2186
2187 static void addrconf_ip6_tnl_config(struct net_device *dev)
2188 {
2189         struct inet6_dev *idev;
2190
2191         ASSERT_RTNL();
2192
2193         if ((idev = addrconf_add_dev(dev)) == NULL) {
2194                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2195                 return;
2196         }
2197         ip6_tnl_add_linklocal(idev);
2198 }
2199
2200 static int addrconf_notify(struct notifier_block *this, unsigned long event, 
2201                            void * data)
2202 {
2203         struct net_device *dev = (struct net_device *) data;
2204         struct inet6_dev *idev = __in6_dev_get(dev);
2205         int run_pending = 0;
2206
2207         switch(event) {
2208         case NETDEV_UP:
2209         case NETDEV_CHANGE:
2210                 if (event == NETDEV_UP) {
2211                         if (!netif_carrier_ok(dev)) {
2212                                 /* device is not ready yet. */
2213                                 printk(KERN_INFO
2214                                         "ADDRCONF(NETDEV_UP): %s: "
2215                                         "link is not ready\n",
2216                                         dev->name);
2217                                 break;
2218                         }
2219
2220                         if (idev)
2221                                 idev->if_flags |= IF_READY;
2222                 } else {
2223                         if (!netif_carrier_ok(dev)) {
2224                                 /* device is still not ready. */
2225                                 break;
2226                         }
2227
2228                         if (idev) {
2229                                 if (idev->if_flags & IF_READY) {
2230                                         /* device is already configured. */
2231                                         break;
2232                                 }
2233                                 idev->if_flags |= IF_READY;
2234                         }
2235
2236                         printk(KERN_INFO
2237                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2238                                         "link becomes ready\n",
2239                                         dev->name);
2240
2241                         run_pending = 1;
2242                 }
2243
2244                 switch(dev->type) {
2245                 case ARPHRD_SIT:
2246                         addrconf_sit_config(dev);
2247                         break;
2248                 case ARPHRD_TUNNEL6:
2249                         addrconf_ip6_tnl_config(dev);
2250                         break;
2251                 case ARPHRD_LOOPBACK:
2252                         init_loopback(dev);
2253                         break;
2254
2255                 default:
2256                         addrconf_dev_config(dev);
2257                         break;
2258                 };
2259                 if (idev) {
2260                         if (run_pending)
2261                                 addrconf_dad_run(idev);
2262
2263                         /* If the MTU changed during the interface down, when the
2264                            interface up, the changed MTU must be reflected in the
2265                            idev as well as routers.
2266                          */
2267                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2268                                 rt6_mtu_change(dev, dev->mtu);
2269                                 idev->cnf.mtu6 = dev->mtu;
2270                         }
2271                         idev->tstamp = jiffies;
2272                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2273                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2274                            stop IPv6 on this interface.
2275                          */
2276                         if (dev->mtu < IPV6_MIN_MTU)
2277                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2278                 }
2279                 break;
2280
2281         case NETDEV_CHANGEMTU:
2282                 if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2283                         rt6_mtu_change(dev, dev->mtu);
2284                         idev->cnf.mtu6 = dev->mtu;
2285                         break;
2286                 }
2287
2288                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2289
2290         case NETDEV_DOWN:
2291         case NETDEV_UNREGISTER:
2292                 /*
2293                  *      Remove all addresses from this interface.
2294                  */
2295                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2296                 break;
2297
2298         case NETDEV_CHANGENAME:
2299 #ifdef CONFIG_SYSCTL
2300                 if (idev) {
2301                         addrconf_sysctl_unregister(&idev->cnf);
2302                         neigh_sysctl_unregister(idev->nd_parms);
2303                         neigh_sysctl_register(dev, idev->nd_parms,
2304                                               NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2305                                               &ndisc_ifinfo_sysctl_change,
2306                                               NULL);
2307                         addrconf_sysctl_register(idev, &idev->cnf);
2308                 }
2309 #endif
2310                 break;
2311         };
2312
2313         return NOTIFY_OK;
2314 }
2315
2316 /*
2317  *      addrconf module should be notified of a device going up
2318  */
2319 static struct notifier_block ipv6_dev_notf = {
2320         .notifier_call = addrconf_notify,
2321         .priority = 0
2322 };
2323
2324 static int addrconf_ifdown(struct net_device *dev, int how)
2325 {
2326         struct inet6_dev *idev;
2327         struct inet6_ifaddr *ifa, **bifa;
2328         int i;
2329
2330         ASSERT_RTNL();
2331
2332         if (dev == &loopback_dev && how == 1)
2333                 how = 0;
2334
2335         rt6_ifdown(dev);
2336         neigh_ifdown(&nd_tbl, dev);
2337
2338         idev = __in6_dev_get(dev);
2339         if (idev == NULL)
2340                 return -ENODEV;
2341
2342         /* Step 1: remove reference to ipv6 device from parent device.
2343                    Do not dev_put!
2344          */
2345         if (how == 1) {
2346                 write_lock_bh(&addrconf_lock);
2347                 dev->ip6_ptr = NULL;
2348                 idev->dead = 1;
2349                 write_unlock_bh(&addrconf_lock);
2350
2351                 /* Step 1.5: remove snmp6 entry */
2352                 snmp6_unregister_dev(idev);
2353
2354         }
2355
2356         /* Step 2: clear hash table */
2357         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2358                 bifa = &inet6_addr_lst[i];
2359
2360                 write_lock_bh(&addrconf_hash_lock);
2361                 while ((ifa = *bifa) != NULL) {
2362                         if (ifa->idev == idev) {
2363                                 *bifa = ifa->lst_next;
2364                                 ifa->lst_next = NULL;
2365                                 addrconf_del_timer(ifa);
2366                                 in6_ifa_put(ifa);
2367                                 continue;
2368                         }
2369                         bifa = &ifa->lst_next;
2370                 }
2371                 write_unlock_bh(&addrconf_hash_lock);
2372         }
2373
2374         write_lock_bh(&idev->lock);
2375
2376         /* Step 3: clear flags for stateless addrconf */
2377         if (how != 1)
2378                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2379
2380         /* Step 4: clear address list */
2381 #ifdef CONFIG_IPV6_PRIVACY
2382         if (how == 1 && del_timer(&idev->regen_timer))
2383                 in6_dev_put(idev);
2384
2385         /* clear tempaddr list */
2386         while ((ifa = idev->tempaddr_list) != NULL) {
2387                 idev->tempaddr_list = ifa->tmp_next;
2388                 ifa->tmp_next = NULL;
2389                 ifa->dead = 1;
2390                 write_unlock_bh(&idev->lock);
2391                 spin_lock_bh(&ifa->lock);
2392
2393                 if (ifa->ifpub) {
2394                         in6_ifa_put(ifa->ifpub);
2395                         ifa->ifpub = NULL;
2396                 }
2397                 spin_unlock_bh(&ifa->lock);
2398                 in6_ifa_put(ifa);
2399                 write_lock_bh(&idev->lock);
2400         }
2401 #endif
2402         while ((ifa = idev->addr_list) != NULL) {
2403                 idev->addr_list = ifa->if_next;
2404                 ifa->if_next = NULL;
2405                 ifa->dead = 1;
2406                 addrconf_del_timer(ifa);
2407                 write_unlock_bh(&idev->lock);
2408
2409                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2410                 in6_ifa_put(ifa);
2411
2412                 write_lock_bh(&idev->lock);
2413         }
2414         write_unlock_bh(&idev->lock);
2415
2416         /* Step 5: Discard multicast list */
2417
2418         if (how == 1)
2419                 ipv6_mc_destroy_dev(idev);
2420         else
2421                 ipv6_mc_down(idev);
2422
2423         /* Step 5: netlink notification of this interface */
2424         idev->tstamp = jiffies;
2425         inet6_ifinfo_notify(RTM_DELLINK, idev);
2426         
2427         /* Shot the device (if unregistered) */
2428
2429         if (how == 1) {
2430 #ifdef CONFIG_SYSCTL
2431                 addrconf_sysctl_unregister(&idev->cnf);
2432                 neigh_sysctl_unregister(idev->nd_parms);
2433 #endif
2434                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2435                 neigh_ifdown(&nd_tbl, dev);
2436                 in6_dev_put(idev);
2437         }
2438         return 0;
2439 }
2440
2441 static void addrconf_rs_timer(unsigned long data)
2442 {
2443         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2444
2445         if (ifp->idev->cnf.forwarding)
2446                 goto out;
2447
2448         if (ifp->idev->if_flags & IF_RA_RCVD) {
2449                 /*
2450                  *      Announcement received after solicitation
2451                  *      was sent
2452                  */
2453                 goto out;
2454         }
2455
2456         spin_lock(&ifp->lock);
2457         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2458                 struct in6_addr all_routers;
2459
2460                 /* The wait after the last probe can be shorter */
2461                 addrconf_mod_timer(ifp, AC_RS,
2462                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2463                                    ifp->idev->cnf.rtr_solicit_delay :
2464                                    ifp->idev->cnf.rtr_solicit_interval);
2465                 spin_unlock(&ifp->lock);
2466
2467                 ipv6_addr_all_routers(&all_routers);
2468
2469                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2470         } else {
2471                 spin_unlock(&ifp->lock);
2472                 /*
2473                  * Note: we do not support deprecated "all on-link"
2474                  * assumption any longer.
2475                  */
2476                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2477                        ifp->idev->dev->name);
2478         }
2479
2480 out:
2481         in6_ifa_put(ifp);
2482 }
2483
2484 /*
2485  *      Duplicate Address Detection
2486  */
2487 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2488 {
2489         unsigned long rand_num;
2490         struct inet6_dev *idev = ifp->idev;
2491
2492         rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2493         ifp->probes = idev->cnf.dad_transmits;
2494         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2495 }
2496
2497 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2498 {
2499         struct inet6_dev *idev = ifp->idev;
2500         struct net_device *dev = idev->dev;
2501
2502         addrconf_join_solict(dev, &ifp->addr);
2503
2504         if (ifp->prefix_len != 128 && (ifp->flags&IFA_F_PERMANENT))
2505                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev, 0,
2506                                         flags);
2507
2508         net_srandom(ifp->addr.s6_addr32[3]);
2509
2510         read_lock_bh(&idev->lock);
2511         if (ifp->dead)
2512                 goto out;
2513         spin_lock_bh(&ifp->lock);
2514
2515         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2516             !(ifp->flags&IFA_F_TENTATIVE)) {
2517                 ifp->flags &= ~IFA_F_TENTATIVE;
2518                 spin_unlock_bh(&ifp->lock);
2519                 read_unlock_bh(&idev->lock);
2520
2521                 addrconf_dad_completed(ifp);
2522                 return;
2523         }
2524
2525         if (!(idev->if_flags & IF_READY)) {
2526                 spin_unlock_bh(&ifp->lock);
2527                 read_unlock_bh(&idev->lock);
2528                 /*
2529                  * If the defice is not ready:
2530                  * - keep it tentative if it is a permanent address.
2531                  * - otherwise, kill it.
2532                  */
2533                 in6_ifa_hold(ifp);
2534                 addrconf_dad_stop(ifp);
2535                 return;
2536         }
2537         addrconf_dad_kick(ifp);
2538         spin_unlock_bh(&ifp->lock);
2539 out:
2540         read_unlock_bh(&idev->lock);
2541 }
2542
2543 static void addrconf_dad_timer(unsigned long data)
2544 {
2545         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2546         struct inet6_dev *idev = ifp->idev;
2547         struct in6_addr unspec;
2548         struct in6_addr mcaddr;
2549
2550         read_lock_bh(&idev->lock);
2551         if (idev->dead) {
2552                 read_unlock_bh(&idev->lock);
2553                 goto out;
2554         }
2555         spin_lock_bh(&ifp->lock);
2556         if (ifp->probes == 0) {
2557                 /*
2558                  * DAD was successful
2559                  */
2560
2561                 ifp->flags &= ~IFA_F_TENTATIVE;
2562                 spin_unlock_bh(&ifp->lock);
2563                 read_unlock_bh(&idev->lock);
2564
2565                 addrconf_dad_completed(ifp);
2566
2567                 goto out;
2568         }
2569
2570         ifp->probes--;
2571         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2572         spin_unlock_bh(&ifp->lock);
2573         read_unlock_bh(&idev->lock);
2574
2575         /* send a neighbour solicitation for our addr */
2576         memset(&unspec, 0, sizeof(unspec));
2577         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2578         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2579 out:
2580         in6_ifa_put(ifp);
2581 }
2582
2583 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2584 {
2585         struct net_device *     dev = ifp->idev->dev;
2586
2587         /*
2588          *      Configure the address for reception. Now it is valid.
2589          */
2590
2591         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2592
2593         /* If added prefix is link local and forwarding is off,
2594            start sending router solicitations.
2595          */
2596
2597         if (ifp->idev->cnf.forwarding == 0 &&
2598             ifp->idev->cnf.rtr_solicits > 0 &&
2599             (dev->flags&IFF_LOOPBACK) == 0 &&
2600             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2601                 struct in6_addr all_routers;
2602
2603                 ipv6_addr_all_routers(&all_routers);
2604
2605                 /*
2606                  *      If a host as already performed a random delay
2607                  *      [...] as part of DAD [...] there is no need
2608                  *      to delay again before sending the first RS
2609                  */
2610                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2611
2612                 spin_lock_bh(&ifp->lock);
2613                 ifp->probes = 1;
2614                 ifp->idev->if_flags |= IF_RS_SENT;
2615                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2616                 spin_unlock_bh(&ifp->lock);
2617         }
2618 }
2619
2620 static void addrconf_dad_run(struct inet6_dev *idev) {
2621         struct inet6_ifaddr *ifp;
2622
2623         read_lock_bh(&idev->lock);
2624         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2625                 spin_lock_bh(&ifp->lock);
2626                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2627                         spin_unlock_bh(&ifp->lock);
2628                         continue;
2629                 }
2630                 spin_unlock_bh(&ifp->lock);
2631                 addrconf_dad_kick(ifp);
2632         }
2633         read_unlock_bh(&idev->lock);
2634 }
2635
2636 #ifdef CONFIG_PROC_FS
2637 struct if6_iter_state {
2638         int bucket;
2639 };
2640
2641 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2642 {
2643         struct inet6_ifaddr *ifa = NULL;
2644         struct if6_iter_state *state = seq->private;
2645
2646         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2647                 ifa = inet6_addr_lst[state->bucket];
2648                 if (ifa)
2649                         break;
2650         }
2651         return ifa;
2652 }
2653
2654 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2655 {
2656         struct if6_iter_state *state = seq->private;
2657
2658         ifa = ifa->lst_next;
2659 try_again:
2660         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2661                 ifa = inet6_addr_lst[state->bucket];
2662                 goto try_again;
2663         }
2664         return ifa;
2665 }
2666
2667 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2668 {
2669         struct inet6_ifaddr *ifa = if6_get_first(seq);
2670
2671         if (ifa)
2672                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2673                         --pos;
2674         return pos ? NULL : ifa;
2675 }
2676
2677 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2678 {
2679         read_lock_bh(&addrconf_hash_lock);
2680         return if6_get_idx(seq, *pos);
2681 }
2682
2683 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2684 {
2685         struct inet6_ifaddr *ifa;
2686
2687         ifa = if6_get_next(seq, v);
2688         ++*pos;
2689         return ifa;
2690 }
2691
2692 static void if6_seq_stop(struct seq_file *seq, void *v)
2693 {
2694         read_unlock_bh(&addrconf_hash_lock);
2695 }
2696
2697 static int if6_seq_show(struct seq_file *seq, void *v)
2698 {
2699         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2700         seq_printf(seq,
2701                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2702                    NIP6(ifp->addr),
2703                    ifp->idev->dev->ifindex,
2704                    ifp->prefix_len,
2705                    ifp->scope,
2706                    ifp->flags,
2707                    ifp->idev->dev->name);
2708         return 0;
2709 }
2710
2711 static struct seq_operations if6_seq_ops = {
2712         .start  = if6_seq_start,
2713         .next   = if6_seq_next,
2714         .show   = if6_seq_show,
2715         .stop   = if6_seq_stop,
2716 };
2717
2718 static int if6_seq_open(struct inode *inode, struct file *file)
2719 {
2720         struct seq_file *seq;
2721         int rc = -ENOMEM;
2722         struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2723
2724         if (!s)
2725                 goto out;
2726
2727         rc = seq_open(file, &if6_seq_ops);
2728         if (rc)
2729                 goto out_kfree;
2730
2731         seq = file->private_data;
2732         seq->private = s;
2733 out:
2734         return rc;
2735 out_kfree:
2736         kfree(s);
2737         goto out;
2738 }
2739
2740 static struct file_operations if6_fops = {
2741         .owner          = THIS_MODULE,
2742         .open           = if6_seq_open,
2743         .read           = seq_read,
2744         .llseek         = seq_lseek,
2745         .release        = seq_release_private,
2746 };
2747
2748 int __init if6_proc_init(void)
2749 {
2750         if (!proc_net_fops_create("if_inet6", S_IRUGO, &if6_fops))
2751                 return -ENOMEM;
2752         return 0;
2753 }
2754
2755 void if6_proc_exit(void)
2756 {
2757         proc_net_remove("if_inet6");
2758 }
2759 #endif  /* CONFIG_PROC_FS */
2760
2761 /*
2762  *      Periodic address status verification
2763  */
2764
2765 static void addrconf_verify(unsigned long foo)
2766 {
2767         struct inet6_ifaddr *ifp;
2768         unsigned long now, next;
2769         int i;
2770
2771         spin_lock_bh(&addrconf_verify_lock);
2772         now = jiffies;
2773         next = now + ADDR_CHECK_FREQUENCY;
2774
2775         del_timer(&addr_chk_timer);
2776
2777         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2778
2779 restart:
2780                 read_lock(&addrconf_hash_lock);
2781                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2782                         unsigned long age;
2783 #ifdef CONFIG_IPV6_PRIVACY
2784                         unsigned long regen_advance;
2785 #endif
2786
2787                         if (ifp->flags & IFA_F_PERMANENT)
2788                                 continue;
2789
2790                         spin_lock(&ifp->lock);
2791                         age = (now - ifp->tstamp) / HZ;
2792
2793 #ifdef CONFIG_IPV6_PRIVACY
2794                         regen_advance = ifp->idev->cnf.regen_max_retry * 
2795                                         ifp->idev->cnf.dad_transmits * 
2796                                         ifp->idev->nd_parms->retrans_time / HZ;
2797 #endif
2798
2799                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2800                             age >= ifp->valid_lft) {
2801                                 spin_unlock(&ifp->lock);
2802                                 in6_ifa_hold(ifp);
2803                                 read_unlock(&addrconf_hash_lock);
2804                                 ipv6_del_addr(ifp);
2805                                 goto restart;
2806                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2807                                 spin_unlock(&ifp->lock);
2808                                 continue;
2809                         } else if (age >= ifp->prefered_lft) {
2810                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2811                                 int deprecate = 0;
2812
2813                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2814                                         deprecate = 1;
2815                                         ifp->flags |= IFA_F_DEPRECATED;
2816                                 }
2817
2818                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2819                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2820
2821                                 spin_unlock(&ifp->lock);
2822
2823                                 if (deprecate) {
2824                                         in6_ifa_hold(ifp);
2825                                         read_unlock(&addrconf_hash_lock);
2826
2827                                         ipv6_ifa_notify(0, ifp);
2828                                         in6_ifa_put(ifp);
2829                                         goto restart;
2830                                 }
2831 #ifdef CONFIG_IPV6_PRIVACY
2832                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2833                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2834                                 if (age >= ifp->prefered_lft - regen_advance) {
2835                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2836                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2837                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2838                                         if (!ifp->regen_count && ifpub) {
2839                                                 ifp->regen_count++;
2840                                                 in6_ifa_hold(ifp);
2841                                                 in6_ifa_hold(ifpub);
2842                                                 spin_unlock(&ifp->lock);
2843                                                 read_unlock(&addrconf_hash_lock);
2844                                                 spin_lock(&ifpub->lock);
2845                                                 ifpub->regen_count = 0;
2846                                                 spin_unlock(&ifpub->lock);
2847                                                 ipv6_create_tempaddr(ifpub, ifp);
2848                                                 in6_ifa_put(ifpub);
2849                                                 in6_ifa_put(ifp);
2850                                                 goto restart;
2851                                         }
2852                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2853                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2854                                 spin_unlock(&ifp->lock);
2855 #endif
2856                         } else {
2857                                 /* ifp->prefered_lft <= ifp->valid_lft */
2858                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2859                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
2860                                 spin_unlock(&ifp->lock);
2861                         }
2862                 }
2863                 read_unlock(&addrconf_hash_lock);
2864         }
2865
2866         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2867         add_timer(&addr_chk_timer);
2868         spin_unlock_bh(&addrconf_verify_lock);
2869 }
2870
2871 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2872 {
2873         struct in6_addr *pfx = NULL;
2874
2875         if (addr)
2876                 pfx = nla_data(addr);
2877
2878         if (local) {
2879                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2880                         pfx = NULL;
2881                 else
2882                         pfx = nla_data(local);
2883         }
2884
2885         return pfx;
2886 }
2887
2888 static struct nla_policy ifa_ipv6_policy[IFA_MAX+1] __read_mostly = {
2889         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
2890         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
2891         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
2892 };
2893
2894 static int
2895 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2896 {
2897         struct ifaddrmsg *ifm;
2898         struct nlattr *tb[IFA_MAX+1];
2899         struct in6_addr *pfx;
2900         int err;
2901
2902         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2903         if (err < 0)
2904                 return err;
2905
2906         ifm = nlmsg_data(nlh);
2907         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2908         if (pfx == NULL)
2909                 return -EINVAL;
2910
2911         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
2912 }
2913
2914 static int
2915 inet6_addr_modify(int ifindex, struct in6_addr *pfx,
2916                   __u32 prefered_lft, __u32 valid_lft)
2917 {
2918         struct inet6_ifaddr *ifp = NULL;
2919         struct net_device *dev;
2920         int ifa_flags = 0;
2921
2922         if ((dev = __dev_get_by_index(ifindex)) == NULL)
2923                 return -ENODEV;
2924
2925         if (!(dev->flags&IFF_UP))
2926                 return -ENETDOWN;
2927
2928         if (!valid_lft || (prefered_lft > valid_lft))
2929                 return -EINVAL;
2930
2931         ifp = ipv6_get_ifaddr(pfx, dev, 1);
2932         if (ifp == NULL)
2933                 return -ENOENT;
2934
2935         if (valid_lft == INFINITY_LIFE_TIME)
2936                 ifa_flags = IFA_F_PERMANENT;
2937         else if (valid_lft >= 0x7FFFFFFF/HZ)
2938                 valid_lft = 0x7FFFFFFF/HZ;
2939
2940         if (prefered_lft == 0)
2941                 ifa_flags = IFA_F_DEPRECATED;
2942         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2943                  (prefered_lft != INFINITY_LIFE_TIME))
2944                 prefered_lft = 0x7FFFFFFF/HZ;
2945
2946         spin_lock_bh(&ifp->lock);
2947         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED|IFA_F_PERMANENT)) | ifa_flags;
2948
2949         ifp->tstamp = jiffies;
2950         ifp->valid_lft = valid_lft;
2951         ifp->prefered_lft = prefered_lft;
2952
2953         spin_unlock_bh(&ifp->lock);
2954         if (!(ifp->flags&IFA_F_TENTATIVE))
2955                 ipv6_ifa_notify(0, ifp);
2956         in6_ifa_put(ifp);
2957
2958         addrconf_verify(0);
2959
2960         return 0;
2961 }
2962
2963 static int
2964 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2965 {
2966         struct ifaddrmsg *ifm;
2967         struct nlattr *tb[IFA_MAX+1];
2968         struct in6_addr *pfx;
2969         u32 valid_lft, preferred_lft;
2970         int err;
2971
2972         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2973         if (err < 0)
2974                 return err;
2975
2976         ifm = nlmsg_data(nlh);
2977         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2978         if (pfx == NULL)
2979                 return -EINVAL;
2980
2981         if (tb[IFA_CACHEINFO]) {
2982                 struct ifa_cacheinfo *ci;
2983
2984                 ci = nla_data(tb[IFA_CACHEINFO]);
2985                 valid_lft = ci->ifa_valid;
2986                 preferred_lft = ci->ifa_prefered;
2987         } else {
2988                 preferred_lft = INFINITY_LIFE_TIME;
2989                 valid_lft = INFINITY_LIFE_TIME;
2990         }
2991
2992         if (nlh->nlmsg_flags & NLM_F_REPLACE) {
2993                 err = inet6_addr_modify(ifm->ifa_index, pfx,
2994                                         preferred_lft, valid_lft);
2995                 if (err == 0 || !(nlh->nlmsg_flags & NLM_F_CREATE))
2996                         return err;
2997         }
2998
2999         return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3000                               preferred_lft, valid_lft);
3001 }
3002
3003 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3004                           u8 scope, int ifindex)
3005 {
3006         struct ifaddrmsg *ifm;
3007
3008         ifm = nlmsg_data(nlh);
3009         ifm->ifa_family = AF_INET6;
3010         ifm->ifa_prefixlen = prefixlen;
3011         ifm->ifa_flags = flags;
3012         ifm->ifa_scope = scope;
3013         ifm->ifa_index = ifindex;
3014 }
3015
3016 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3017                          unsigned long tstamp, u32 preferred, u32 valid)
3018 {
3019         struct ifa_cacheinfo ci;
3020
3021         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3022                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3023         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3024                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3025         ci.ifa_prefered = preferred;
3026         ci.ifa_valid = valid;
3027
3028         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3029 }
3030
3031 static inline int rt_scope(int ifa_scope)
3032 {
3033         if (ifa_scope & IFA_HOST)
3034                 return RT_SCOPE_HOST;
3035         else if (ifa_scope & IFA_LINK)
3036                 return RT_SCOPE_LINK;
3037         else if (ifa_scope & IFA_SITE)
3038                 return RT_SCOPE_SITE;
3039         else
3040                 return RT_SCOPE_UNIVERSE;
3041 }
3042
3043 /* Maximum length of ifa_cacheinfo attributes */
3044 #define INET6_IFADDR_RTA_SPACE \
3045                 RTA_SPACE(16) /* IFA_ADDRESS */ + \
3046                 RTA_SPACE(sizeof(struct ifa_cacheinfo)) /* CACHEINFO */
3047
3048 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3049                              u32 pid, u32 seq, int event, unsigned int flags)
3050 {
3051         struct nlmsghdr  *nlh;
3052         unsigned char    *b = skb->tail;
3053         u32 preferred, valid;
3054
3055         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3056         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3057                       ifa->idev->dev->ifindex);
3058
3059         RTA_PUT(skb, IFA_ADDRESS, 16, &ifa->addr);
3060         if (!(ifa->flags&IFA_F_PERMANENT)) {
3061                 preferred = ifa->prefered_lft;
3062                 valid = ifa->valid_lft;
3063                 if (preferred != INFINITY_LIFE_TIME) {
3064                         long tval = (jiffies - ifa->tstamp)/HZ;
3065                         preferred -= tval;
3066                         if (valid != INFINITY_LIFE_TIME)
3067                                 valid -= tval;
3068                 }
3069         } else {
3070                 preferred = INFINITY_LIFE_TIME;
3071                 valid = INFINITY_LIFE_TIME;
3072         }
3073
3074         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3075                 goto rtattr_failure;
3076
3077         nlh->nlmsg_len = skb->tail - b;
3078         return skb->len;
3079
3080 nlmsg_failure:
3081 rtattr_failure:
3082         skb_trim(skb, b - skb->data);
3083         return -1;
3084 }
3085
3086 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3087                                 u32 pid, u32 seq, int event, u16 flags)
3088 {
3089         struct nlmsghdr  *nlh;
3090         unsigned char    *b = skb->tail;
3091         u8 scope = RT_SCOPE_UNIVERSE;
3092         int ifindex = ifmca->idev->dev->ifindex;
3093
3094         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3095                 scope = RT_SCOPE_SITE;
3096
3097         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3098         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3099         RTA_PUT(skb, IFA_MULTICAST, 16, &ifmca->mca_addr);
3100
3101         if (put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3102                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3103                 goto rtattr_failure;
3104
3105         nlh->nlmsg_len = skb->tail - b;
3106         return skb->len;
3107
3108 nlmsg_failure:
3109 rtattr_failure:
3110         skb_trim(skb, b - skb->data);
3111         return -1;
3112 }
3113
3114 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3115                                 u32 pid, u32 seq, int event, unsigned int flags)
3116 {
3117         struct nlmsghdr  *nlh;
3118         unsigned char    *b = skb->tail;
3119         u8 scope = RT_SCOPE_UNIVERSE;
3120         int ifindex = ifaca->aca_idev->dev->ifindex;
3121
3122         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3123                 scope = RT_SCOPE_SITE;
3124
3125         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3126         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3127         RTA_PUT(skb, IFA_ANYCAST, 16, &ifaca->aca_addr);
3128
3129         if (put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3130                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3131                 goto rtattr_failure;
3132
3133         nlh->nlmsg_len = skb->tail - b;
3134         return skb->len;
3135
3136 nlmsg_failure:
3137 rtattr_failure:
3138         skb_trim(skb, b - skb->data);
3139         return -1;
3140 }
3141
3142 enum addr_type_t
3143 {
3144         UNICAST_ADDR,
3145         MULTICAST_ADDR,
3146         ANYCAST_ADDR,
3147 };
3148
3149 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3150                            enum addr_type_t type)
3151 {
3152         int idx, ip_idx;
3153         int s_idx, s_ip_idx;
3154         int err = 1;
3155         struct net_device *dev;
3156         struct inet6_dev *idev = NULL;
3157         struct inet6_ifaddr *ifa;
3158         struct ifmcaddr6 *ifmca;
3159         struct ifacaddr6 *ifaca;
3160
3161         s_idx = cb->args[0];
3162         s_ip_idx = ip_idx = cb->args[1];
3163         read_lock(&dev_base_lock);
3164         
3165         for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
3166                 if (idx < s_idx)
3167                         continue;
3168                 if (idx > s_idx)
3169                         s_ip_idx = 0;
3170                 ip_idx = 0;
3171                 if ((idev = in6_dev_get(dev)) == NULL)
3172                         continue;
3173                 read_lock_bh(&idev->lock);
3174                 switch (type) {
3175                 case UNICAST_ADDR:
3176                         /* unicast address incl. temp addr */
3177                         for (ifa = idev->addr_list; ifa;
3178                              ifa = ifa->if_next, ip_idx++) {
3179                                 if (ip_idx < s_ip_idx)
3180                                         continue;
3181                                 if ((err = inet6_fill_ifaddr(skb, ifa, 
3182                                     NETLINK_CB(cb->skb).pid, 
3183                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3184                                     NLM_F_MULTI)) <= 0)
3185                                         goto done;
3186                         }
3187                         break;
3188                 case MULTICAST_ADDR:
3189                         /* multicast address */
3190                         for (ifmca = idev->mc_list; ifmca; 
3191                              ifmca = ifmca->next, ip_idx++) {
3192                                 if (ip_idx < s_ip_idx)
3193                                         continue;
3194                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca, 
3195                                     NETLINK_CB(cb->skb).pid, 
3196                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3197                                     NLM_F_MULTI)) <= 0)
3198                                         goto done;
3199                         }
3200                         break;
3201                 case ANYCAST_ADDR:
3202                         /* anycast address */
3203                         for (ifaca = idev->ac_list; ifaca;
3204                              ifaca = ifaca->aca_next, ip_idx++) {
3205                                 if (ip_idx < s_ip_idx)
3206                                         continue;
3207                                 if ((err = inet6_fill_ifacaddr(skb, ifaca, 
3208                                     NETLINK_CB(cb->skb).pid, 
3209                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3210                                     NLM_F_MULTI)) <= 0) 
3211                                         goto done;
3212                         }
3213                         break;
3214                 default:
3215                         break;
3216                 }
3217                 read_unlock_bh(&idev->lock);
3218                 in6_dev_put(idev);
3219         }
3220 done:
3221         if (err <= 0) {
3222                 read_unlock_bh(&idev->lock);
3223                 in6_dev_put(idev);
3224         }
3225         read_unlock(&dev_base_lock);
3226         cb->args[0] = idx;
3227         cb->args[1] = ip_idx;
3228         return skb->len;
3229 }
3230
3231 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3232 {
3233         enum addr_type_t type = UNICAST_ADDR;
3234         return inet6_dump_addr(skb, cb, type);
3235 }
3236
3237 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3238 {
3239         enum addr_type_t type = MULTICAST_ADDR;
3240         return inet6_dump_addr(skb, cb, type);
3241 }
3242
3243
3244 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3245 {
3246         enum addr_type_t type = ANYCAST_ADDR;
3247         return inet6_dump_addr(skb, cb, type);
3248 }
3249
3250 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3251                              void *arg)
3252 {
3253         struct ifaddrmsg *ifm;
3254         struct nlattr *tb[IFA_MAX+1];
3255         struct in6_addr *addr = NULL;
3256         struct net_device *dev = NULL;
3257         struct inet6_ifaddr *ifa;
3258         struct sk_buff *skb;
3259         int payload = sizeof(struct ifaddrmsg) + INET6_IFADDR_RTA_SPACE;
3260         int err;
3261
3262         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3263         if (err < 0)
3264                 goto errout;
3265
3266         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3267         if (addr == NULL) {
3268                 err = -EINVAL;
3269                 goto errout;
3270         }
3271
3272         ifm = nlmsg_data(nlh);
3273         if (ifm->ifa_index)
3274                 dev = __dev_get_by_index(ifm->ifa_index);
3275
3276         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3277                 err = -EADDRNOTAVAIL;
3278                 goto errout;
3279         }
3280
3281         if ((skb = nlmsg_new(nlmsg_total_size(payload), GFP_KERNEL)) == NULL) {
3282                 err = -ENOBUFS;
3283                 goto errout_ifa;
3284         }
3285
3286         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3287                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3288         if (err < 0) {
3289                 kfree_skb(skb);
3290                 goto errout_ifa;
3291         }
3292
3293         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3294 errout_ifa:
3295         in6_ifa_put(ifa);
3296 errout:
3297         return err;
3298 }
3299
3300 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3301 {
3302         struct sk_buff *skb;
3303         int payload = sizeof(struct ifaddrmsg) + INET6_IFADDR_RTA_SPACE;
3304         int err = -ENOBUFS;
3305
3306         skb = nlmsg_new(nlmsg_total_size(payload), GFP_ATOMIC);
3307         if (skb == NULL)
3308                 goto errout;
3309
3310         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3311         if (err < 0) {
3312                 kfree_skb(skb);
3313                 goto errout;
3314         }
3315
3316         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3317 errout:
3318         if (err < 0)
3319                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3320 }
3321
3322 static void inline ipv6_store_devconf(struct ipv6_devconf *cnf,
3323                                 __s32 *array, int bytes)
3324 {
3325         memset(array, 0, bytes);
3326         array[DEVCONF_FORWARDING] = cnf->forwarding;
3327         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3328         array[DEVCONF_MTU6] = cnf->mtu6;
3329         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3330         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3331         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3332         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3333         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3334         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3335         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3336         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3337 #ifdef CONFIG_IPV6_PRIVACY
3338         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3339         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3340         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3341         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3342         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3343 #endif
3344         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3345         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3346         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3347 #ifdef CONFIG_IPV6_ROUTER_PREF
3348         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3349         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3350 #ifdef CONFIV_IPV6_ROUTE_INFO
3351         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3352 #endif
3353 #endif
3354 }
3355
3356 /* Maximum length of ifinfomsg attributes */
3357 #define INET6_IFINFO_RTA_SPACE \
3358                 RTA_SPACE(IFNAMSIZ) /* IFNAME */ + \
3359                 RTA_SPACE(MAX_ADDR_LEN) /* ADDRESS */ + \
3360                 RTA_SPACE(sizeof(u32)) /* MTU */ + \
3361                 RTA_SPACE(sizeof(int)) /* LINK */ + \
3362                 RTA_SPACE(0) /* PROTINFO */ + \
3363                 RTA_SPACE(sizeof(u32)) /* FLAGS */ + \
3364                 RTA_SPACE(sizeof(struct ifla_cacheinfo)) /* CACHEINFO */ + \
3365                 RTA_SPACE(sizeof(__s32[DEVCONF_MAX])) /* CONF */
3366
3367 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, 
3368                              u32 pid, u32 seq, int event, unsigned int flags)
3369 {
3370         struct net_device       *dev = idev->dev;
3371         __s32                   *array = NULL;
3372         struct ifinfomsg        *r;
3373         struct nlmsghdr         *nlh;
3374         unsigned char           *b = skb->tail;
3375         struct rtattr           *subattr;
3376         __u32                   mtu = dev->mtu;
3377         struct ifla_cacheinfo   ci;
3378
3379         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*r), flags);
3380         r = NLMSG_DATA(nlh);
3381         r->ifi_family = AF_INET6;
3382         r->__ifi_pad = 0;
3383         r->ifi_type = dev->type;
3384         r->ifi_index = dev->ifindex;
3385         r->ifi_flags = dev_get_flags(dev);
3386         r->ifi_change = 0;
3387
3388         RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
3389
3390         if (dev->addr_len)
3391                 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3392
3393         RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
3394         if (dev->ifindex != dev->iflink)
3395                 RTA_PUT(skb, IFLA_LINK, sizeof(int), &dev->iflink);
3396                         
3397         subattr = (struct rtattr*)skb->tail;
3398
3399         RTA_PUT(skb, IFLA_PROTINFO, 0, NULL);
3400
3401         /* return the device flags */
3402         RTA_PUT(skb, IFLA_INET6_FLAGS, sizeof(__u32), &idev->if_flags);
3403
3404         /* return interface cacheinfo */
3405         ci.max_reasm_len = IPV6_MAXPLEN;
3406         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3407                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3408         ci.reachable_time = idev->nd_parms->reachable_time;
3409         ci.retrans_time = idev->nd_parms->retrans_time;
3410         RTA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3411         
3412         /* return the device sysctl params */
3413         if ((array = kmalloc(DEVCONF_MAX * sizeof(*array), GFP_ATOMIC)) == NULL)
3414                 goto rtattr_failure;
3415         ipv6_store_devconf(&idev->cnf, array, DEVCONF_MAX * sizeof(*array));
3416         RTA_PUT(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(*array), array);
3417
3418         /* XXX - Statistics/MC not implemented */
3419         subattr->rta_len = skb->tail - (u8*)subattr;
3420
3421         nlh->nlmsg_len = skb->tail - b;
3422         kfree(array);
3423         return skb->len;
3424
3425 nlmsg_failure:
3426 rtattr_failure:
3427         kfree(array);
3428         skb_trim(skb, b - skb->data);
3429         return -1;
3430 }
3431
3432 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3433 {
3434         int idx, err;
3435         int s_idx = cb->args[0];
3436         struct net_device *dev;
3437         struct inet6_dev *idev;
3438
3439         read_lock(&dev_base_lock);
3440         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
3441                 if (idx < s_idx)
3442                         continue;
3443                 if ((idev = in6_dev_get(dev)) == NULL)
3444                         continue;
3445                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid, 
3446                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3447                 in6_dev_put(idev);
3448                 if (err <= 0)
3449                         break;
3450         }
3451         read_unlock(&dev_base_lock);
3452         cb->args[0] = idx;
3453
3454         return skb->len;
3455 }
3456
3457 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3458 {
3459         struct sk_buff *skb;
3460         int payload = sizeof(struct ifinfomsg) + INET6_IFINFO_RTA_SPACE;
3461         int err = -ENOBUFS;
3462         
3463         skb = nlmsg_new(nlmsg_total_size(payload), GFP_ATOMIC);
3464         if (skb == NULL)
3465                 goto errout;
3466
3467         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3468         if (err < 0) {
3469                 kfree_skb(skb);
3470                 goto errout;
3471         }
3472
3473         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3474 errout:
3475         if (err < 0)
3476                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3477 }
3478
3479 /* Maximum length of prefix_cacheinfo attributes */
3480 #define INET6_PREFIX_RTA_SPACE \
3481                 RTA_SPACE(sizeof(((struct prefix_info *)NULL)->prefix)) /* ADDRESS */ + \
3482                 RTA_SPACE(sizeof(struct prefix_cacheinfo)) /* CACHEINFO */
3483
3484 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3485                         struct prefix_info *pinfo, u32 pid, u32 seq, 
3486                         int event, unsigned int flags)
3487 {
3488         struct prefixmsg        *pmsg;
3489         struct nlmsghdr         *nlh;
3490         unsigned char           *b = skb->tail;
3491         struct prefix_cacheinfo ci;
3492
3493         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*pmsg), flags);
3494         pmsg = NLMSG_DATA(nlh);
3495         pmsg->prefix_family = AF_INET6;
3496         pmsg->prefix_pad1 = 0;
3497         pmsg->prefix_pad2 = 0;
3498         pmsg->prefix_ifindex = idev->dev->ifindex;
3499         pmsg->prefix_len = pinfo->prefix_len;
3500         pmsg->prefix_type = pinfo->type;
3501         pmsg->prefix_pad3 = 0;
3502         
3503         pmsg->prefix_flags = 0;
3504         if (pinfo->onlink)
3505                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3506         if (pinfo->autoconf)
3507                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3508
3509         RTA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3510
3511         ci.preferred_time = ntohl(pinfo->prefered);
3512         ci.valid_time = ntohl(pinfo->valid);
3513         RTA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3514
3515         nlh->nlmsg_len = skb->tail - b;
3516         return skb->len;
3517
3518 nlmsg_failure:
3519 rtattr_failure:
3520         skb_trim(skb, b - skb->data);
3521         return -1;
3522 }
3523
3524 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 
3525                          struct prefix_info *pinfo)
3526 {
3527         struct sk_buff *skb;
3528         int payload = sizeof(struct prefixmsg) + INET6_PREFIX_RTA_SPACE;
3529         int err = -ENOBUFS;
3530
3531         skb = nlmsg_new(nlmsg_total_size(payload), GFP_ATOMIC);
3532         if (skb == NULL)
3533                 goto errout;
3534
3535         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3536         if (err < 0) {
3537                 kfree_skb(skb);
3538                 goto errout;
3539         }
3540
3541         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3542 errout:
3543         if (err < 0)
3544                 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3545 }
3546
3547 static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
3548         [RTM_GETLINK - RTM_BASE] = { .dumpit    = inet6_dump_ifinfo, },
3549         [RTM_NEWADDR - RTM_BASE] = { .doit      = inet6_rtm_newaddr, },
3550         [RTM_DELADDR - RTM_BASE] = { .doit      = inet6_rtm_deladdr, },
3551         [RTM_GETADDR - RTM_BASE] = { .doit      = inet6_rtm_getaddr,
3552                                      .dumpit    = inet6_dump_ifaddr, },
3553         [RTM_GETMULTICAST - RTM_BASE] = { .dumpit = inet6_dump_ifmcaddr, },
3554         [RTM_GETANYCAST - RTM_BASE] = { .dumpit = inet6_dump_ifacaddr, },
3555         [RTM_NEWROUTE - RTM_BASE] = { .doit     = inet6_rtm_newroute, },
3556         [RTM_DELROUTE - RTM_BASE] = { .doit     = inet6_rtm_delroute, },
3557         [RTM_GETROUTE - RTM_BASE] = { .doit     = inet6_rtm_getroute,
3558                                       .dumpit   = inet6_dump_fib, },
3559 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3560         [RTM_GETRULE  - RTM_BASE] = { .dumpit   = fib6_rules_dump,   },
3561 #endif
3562 };
3563
3564 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3565 {
3566         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3567
3568         switch (event) {
3569         case RTM_NEWADDR:
3570                 ip6_ins_rt(ifp->rt);
3571                 if (ifp->idev->cnf.forwarding)
3572                         addrconf_join_anycast(ifp);
3573                 break;
3574         case RTM_DELADDR:
3575                 if (ifp->idev->cnf.forwarding)
3576                         addrconf_leave_anycast(ifp);
3577                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3578                 dst_hold(&ifp->rt->u.dst);
3579                 if (ip6_del_rt(ifp->rt))
3580                         dst_free(&ifp->rt->u.dst);
3581                 break;
3582         }
3583 }
3584
3585 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3586 {
3587         read_lock_bh(&addrconf_lock);
3588         if (likely(ifp->idev->dead == 0))
3589                 __ipv6_ifa_notify(event, ifp);
3590         read_unlock_bh(&addrconf_lock);
3591 }
3592
3593 #ifdef CONFIG_SYSCTL
3594
3595 static
3596 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3597                            void __user *buffer, size_t *lenp, loff_t *ppos)
3598 {
3599         int *valp = ctl->data;
3600         int val = *valp;
3601         int ret;
3602
3603         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3604
3605         if (write && valp != &ipv6_devconf_dflt.forwarding) {
3606                 if (valp != &ipv6_devconf.forwarding) {
3607                         if ((!*valp) ^ (!val)) {
3608                                 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3609                                 if (idev == NULL)
3610                                         return ret;
3611                                 dev_forward_change(idev);
3612                         }
3613                 } else {
3614                         ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3615                         addrconf_forward_change();
3616                 }
3617                 if (*valp)
3618                         rt6_purge_dflt_routers();
3619         }
3620
3621         return ret;
3622 }
3623
3624 static int addrconf_sysctl_forward_strategy(ctl_table *table, 
3625                                             int __user *name, int nlen,
3626                                             void __user *oldval,
3627                                             size_t __user *oldlenp,
3628                                             void __user *newval, size_t newlen,
3629                                             void **context)
3630 {
3631         int *valp = table->data;
3632         int new;
3633
3634         if (!newval || !newlen)
3635                 return 0;
3636         if (newlen != sizeof(int))
3637                 return -EINVAL;
3638         if (get_user(new, (int __user *)newval))
3639                 return -EFAULT;
3640         if (new == *valp)
3641                 return 0;
3642         if (oldval && oldlenp) {
3643                 size_t len;
3644                 if (get_user(len, oldlenp))
3645                         return -EFAULT;
3646                 if (len) {
3647                         if (len > table->maxlen)
3648                                 len = table->maxlen;
3649                         if (copy_to_user(oldval, valp, len))
3650                                 return -EFAULT;
3651                         if (put_user(len, oldlenp))
3652                                 return -EFAULT;
3653                 }
3654         }
3655
3656         if (valp != &ipv6_devconf_dflt.forwarding) {
3657                 if (valp != &ipv6_devconf.forwarding) {
3658                         struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3659                         int changed;
3660                         if (unlikely(idev == NULL))
3661                                 return -ENODEV;
3662                         changed = (!*valp) ^ (!new);
3663                         *valp = new;
3664                         if (changed)
3665                                 dev_forward_change(idev);
3666                 } else {
3667                         *valp = new;
3668                         addrconf_forward_change();
3669                 }
3670
3671                 if (*valp)
3672                         rt6_purge_dflt_routers();
3673         } else
3674                 *valp = new;
3675
3676         return 1;
3677 }
3678
3679 static struct addrconf_sysctl_table
3680 {
3681         struct ctl_table_header *sysctl_header;
3682         ctl_table addrconf_vars[__NET_IPV6_MAX];
3683         ctl_table addrconf_dev[2];
3684         ctl_table addrconf_conf_dir[2];
3685         ctl_table addrconf_proto_dir[2];
3686         ctl_table addrconf_root_dir[2];
3687 } addrconf_sysctl __read_mostly = {
3688         .sysctl_header = NULL,
3689         .addrconf_vars = {
3690                 {
3691                         .ctl_name       =       NET_IPV6_FORWARDING,
3692                         .procname       =       "forwarding",
3693                         .data           =       &ipv6_devconf.forwarding,
3694                         .maxlen         =       sizeof(int),
3695                         .mode           =       0644,
3696                         .proc_handler   =       &addrconf_sysctl_forward,
3697                         .strategy       =       &addrconf_sysctl_forward_strategy,
3698                 },
3699                 {
3700                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3701                         .procname       =       "hop_limit",
3702                         .data           =       &ipv6_devconf.hop_limit,
3703                         .maxlen         =       sizeof(int),
3704                         .mode           =       0644,
3705                         .proc_handler   =       proc_dointvec,
3706                 },
3707                 {
3708                         .ctl_name       =       NET_IPV6_MTU,
3709                         .procname       =       "mtu",
3710                         .data           =       &ipv6_devconf.mtu6,
3711                         .maxlen         =       sizeof(int),
3712                         .mode           =       0644,
3713                         .proc_handler   =       &proc_dointvec,
3714                 },
3715                 {
3716                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3717                         .procname       =       "accept_ra",
3718                         .data           =       &ipv6_devconf.accept_ra,
3719                         .maxlen         =       sizeof(int),
3720                         .mode           =       0644,
3721                         .proc_handler   =       &proc_dointvec,
3722                 },
3723                 {
3724                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3725                         .procname       =       "accept_redirects",
3726                         .data           =       &ipv6_devconf.accept_redirects,
3727                         .maxlen         =       sizeof(int),
3728                         .mode           =       0644,
3729                         .proc_handler   =       &proc_dointvec,
3730                 },
3731                 {
3732                         .ctl_name       =       NET_IPV6_AUTOCONF,
3733                         .procname       =       "autoconf",
3734                         .data           =       &ipv6_devconf.autoconf,
3735                         .maxlen         =       sizeof(int),
3736                         .mode           =       0644,
3737                         .proc_handler   =       &proc_dointvec,
3738                 },
3739                 {
3740                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3741                         .procname       =       "dad_transmits",
3742                         .data           =       &ipv6_devconf.dad_transmits,
3743                         .maxlen         =       sizeof(int),
3744                         .mode           =       0644,
3745                         .proc_handler   =       &proc_dointvec,
3746                 },
3747                 {
3748                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3749                         .procname       =       "router_solicitations",
3750                         .data           =       &ipv6_devconf.rtr_solicits,
3751                         .maxlen         =       sizeof(int),
3752                         .mode           =       0644,
3753                         .proc_handler   =       &proc_dointvec,
3754                 },
3755                 {
3756                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3757                         .procname       =       "router_solicitation_interval",
3758                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3759                         .maxlen         =       sizeof(int),
3760                         .mode           =       0644,
3761                         .proc_handler   =       &proc_dointvec_jiffies,
3762                         .strategy       =       &sysctl_jiffies,
3763                 },
3764                 {
3765                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3766                         .procname       =       "router_solicitation_delay",
3767                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3768                         .maxlen         =       sizeof(int),
3769                         .mode           =       0644,
3770                         .proc_handler   =       &proc_dointvec_jiffies,
3771                         .strategy       =       &sysctl_jiffies,
3772                 },
3773                 {
3774                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3775                         .procname       =       "force_mld_version",
3776                         .data           =       &ipv6_devconf.force_mld_version,
3777                         .maxlen         =       sizeof(int),
3778                         .mode           =       0644,
3779                         .proc_handler   =       &proc_dointvec,
3780                 },
3781 #ifdef CONFIG_IPV6_PRIVACY
3782                 {
3783                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3784                         .procname       =       "use_tempaddr",
3785                         .data           =       &ipv6_devconf.use_tempaddr,
3786                         .maxlen         =       sizeof(int),
3787                         .mode           =       0644,
3788                         .proc_handler   =       &proc_dointvec,
3789                 },
3790                 {
3791                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3792                         .procname       =       "temp_valid_lft",
3793                         .data           =       &ipv6_devconf.temp_valid_lft,
3794                         .maxlen         =       sizeof(int),
3795                         .mode           =       0644,
3796                         .proc_handler   =       &proc_dointvec,
3797                 },
3798                 {
3799                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3800                         .procname       =       "temp_prefered_lft",
3801                         .data           =       &ipv6_devconf.temp_prefered_lft,
3802                         .maxlen         =       sizeof(int),
3803                         .mode           =       0644,
3804                         .proc_handler   =       &proc_dointvec,
3805                 },
3806                 {
3807                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3808                         .procname       =       "regen_max_retry",
3809                         .data           =       &ipv6_devconf.regen_max_retry,
3810                         .maxlen         =       sizeof(int),
3811                         .mode           =       0644,
3812                         .proc_handler   =       &proc_dointvec,
3813                 },
3814                 {
3815                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3816                         .procname       =       "max_desync_factor",
3817                         .data           =       &ipv6_devconf.max_desync_factor,
3818                         .maxlen         =       sizeof(int),
3819                         .mode           =       0644,
3820                         .proc_handler   =       &proc_dointvec,
3821                 },
3822 #endif
3823                 {
3824                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3825                         .procname       =       "max_addresses",
3826                         .data           =       &ipv6_devconf.max_addresses,
3827                         .maxlen         =       sizeof(int),
3828                         .mode           =       0644,
3829                         .proc_handler   =       &proc_dointvec,
3830                 },
3831                 {
3832                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3833                         .procname       =       "accept_ra_defrtr",
3834                         .data           =       &ipv6_devconf.accept_ra_defrtr,
3835                         .maxlen         =       sizeof(int),
3836                         .mode           =       0644,
3837                         .proc_handler   =       &proc_dointvec,
3838                 },
3839                 {
3840                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
3841                         .procname       =       "accept_ra_pinfo",
3842                         .data           =       &ipv6_devconf.accept_ra_pinfo,
3843                         .maxlen         =       sizeof(int),
3844                         .mode           =       0644,
3845                         .proc_handler   =       &proc_dointvec,
3846                 },
3847 #ifdef CONFIG_IPV6_ROUTER_PREF
3848                 {
3849                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
3850                         .procname       =       "accept_ra_rtr_pref",
3851                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
3852                         .maxlen         =       sizeof(int),
3853                         .mode           =       0644,
3854                         .proc_handler   =       &proc_dointvec,
3855                 },
3856                 {
3857                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
3858                         .procname       =       "router_probe_interval",
3859                         .data           =       &ipv6_devconf.rtr_probe_interval,
3860                         .maxlen         =       sizeof(int),
3861                         .mode           =       0644,
3862                         .proc_handler   =       &proc_dointvec_jiffies,
3863                         .strategy       =       &sysctl_jiffies,
3864                 },
3865 #ifdef CONFIV_IPV6_ROUTE_INFO
3866                 {
3867                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
3868                         .procname       =       "accept_ra_rt_info_max_plen",
3869                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
3870                         .maxlen         =       sizeof(int),
3871                         .mode           =       0644,
3872                         .proc_handler   =       &proc_dointvec,
3873                 },
3874 #endif
3875 #endif
3876                 {
3877                         .ctl_name       =       0,      /* sentinel */
3878                 }
3879         },
3880         .addrconf_dev = {
3881                 {
3882                         .ctl_name       =       NET_PROTO_CONF_ALL,
3883                         .procname       =       "all",
3884                         .mode           =       0555,
3885                         .child          =       addrconf_sysctl.addrconf_vars,
3886                 },
3887                 {
3888                         .ctl_name       =       0,      /* sentinel */
3889                 }
3890         },
3891         .addrconf_conf_dir = {
3892                 {
3893                         .ctl_name       =       NET_IPV6_CONF,
3894                         .procname       =       "conf",
3895                         .mode           =       0555,
3896                         .child          =       addrconf_sysctl.addrconf_dev,
3897                 },
3898                 {
3899                         .ctl_name       =       0,      /* sentinel */
3900                 }
3901         },
3902         .addrconf_proto_dir = {
3903                 {
3904                         .ctl_name       =       NET_IPV6,
3905                         .procname       =       "ipv6",
3906                         .mode           =       0555,
3907                         .child          =       addrconf_sysctl.addrconf_conf_dir,
3908                 },
3909                 {
3910                         .ctl_name       =       0,      /* sentinel */
3911                 }
3912         },
3913         .addrconf_root_dir = {
3914                 {
3915                         .ctl_name       =       CTL_NET,
3916                         .procname       =       "net",
3917                         .mode           =       0555,
3918                         .child          =       addrconf_sysctl.addrconf_proto_dir,
3919                 },
3920                 {
3921                         .ctl_name       =       0,      /* sentinel */
3922                 }
3923         },
3924 };
3925
3926 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
3927 {
3928         int i;
3929         struct net_device *dev = idev ? idev->dev : NULL;
3930         struct addrconf_sysctl_table *t;
3931         char *dev_name = NULL;
3932
3933         t = kmalloc(sizeof(*t), GFP_KERNEL);
3934         if (t == NULL)
3935                 return;
3936         memcpy(t, &addrconf_sysctl, sizeof(*t));
3937         for (i=0; t->addrconf_vars[i].data; i++) {
3938                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
3939                 t->addrconf_vars[i].de = NULL;
3940                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
3941         }
3942         if (dev) {
3943                 dev_name = dev->name; 
3944                 t->addrconf_dev[0].ctl_name = dev->ifindex;
3945         } else {
3946                 dev_name = "default";
3947                 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
3948         }
3949
3950         /* 
3951          * Make a copy of dev_name, because '.procname' is regarded as const 
3952          * by sysctl and we wouldn't want anyone to change it under our feet
3953          * (see SIOCSIFNAME).
3954          */     
3955         dev_name = kstrdup(dev_name, GFP_KERNEL);
3956         if (!dev_name)
3957             goto free;
3958
3959         t->addrconf_dev[0].procname = dev_name;
3960
3961         t->addrconf_dev[0].child = t->addrconf_vars;
3962         t->addrconf_dev[0].de = NULL;
3963         t->addrconf_conf_dir[0].child = t->addrconf_dev;
3964         t->addrconf_conf_dir[0].de = NULL;
3965         t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
3966         t->addrconf_proto_dir[0].de = NULL;
3967         t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
3968         t->addrconf_root_dir[0].de = NULL;
3969
3970         t->sysctl_header = register_sysctl_table(t->addrconf_root_dir, 0);
3971         if (t->sysctl_header == NULL)
3972                 goto free_procname;
3973         else
3974                 p->sysctl = t;
3975         return;
3976
3977         /* error path */
3978  free_procname:
3979         kfree(dev_name);
3980  free:
3981         kfree(t);
3982
3983         return;
3984 }
3985
3986 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
3987 {
3988         if (p->sysctl) {
3989                 struct addrconf_sysctl_table *t = p->sysctl;
3990                 p->sysctl = NULL;
3991                 unregister_sysctl_table(t->sysctl_header);
3992                 kfree(t->addrconf_dev[0].procname);
3993                 kfree(t);
3994         }
3995 }
3996
3997
3998 #endif
3999
4000 /*
4001  *      Device notifier
4002  */
4003
4004 int register_inet6addr_notifier(struct notifier_block *nb)
4005 {
4006         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4007 }
4008
4009 int unregister_inet6addr_notifier(struct notifier_block *nb)
4010 {
4011         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4012 }
4013
4014 /*
4015  *      Init / cleanup code
4016  */
4017
4018 int __init addrconf_init(void)
4019 {
4020         int err = 0;
4021
4022         /* The addrconf netdev notifier requires that loopback_dev
4023          * has it's ipv6 private information allocated and setup
4024          * before it can bring up and give link-local addresses
4025          * to other devices which are up.
4026          *
4027          * Unfortunately, loopback_dev is not necessarily the first
4028          * entry in the global dev_base list of net devices.  In fact,
4029          * it is likely to be the very last entry on that list.
4030          * So this causes the notifier registry below to try and
4031          * give link-local addresses to all devices besides loopback_dev
4032          * first, then loopback_dev, which cases all the non-loopback_dev
4033          * devices to fail to get a link-local address.
4034          *
4035          * So, as a temporary fix, allocate the ipv6 structure for
4036          * loopback_dev first by hand.
4037          * Longer term, all of the dependencies ipv6 has upon the loopback
4038          * device and it being up should be removed.
4039          */
4040         rtnl_lock();
4041         if (!ipv6_add_dev(&loopback_dev))
4042                 err = -ENOMEM;
4043         rtnl_unlock();
4044         if (err)
4045                 return err;
4046
4047         ip6_null_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4048
4049         register_netdevice_notifier(&ipv6_dev_notf);
4050
4051         addrconf_verify(0);
4052         rtnetlink_links[PF_INET6] = inet6_rtnetlink_table;
4053 #ifdef CONFIG_SYSCTL
4054         addrconf_sysctl.sysctl_header =
4055                 register_sysctl_table(addrconf_sysctl.addrconf_root_dir, 0);
4056         addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4057 #endif
4058
4059         return 0;
4060 }
4061
4062 void __exit addrconf_cleanup(void)
4063 {
4064         struct net_device *dev;
4065         struct inet6_dev *idev;
4066         struct inet6_ifaddr *ifa;
4067         int i;
4068
4069         unregister_netdevice_notifier(&ipv6_dev_notf);
4070
4071         rtnetlink_links[PF_INET6] = NULL;
4072 #ifdef CONFIG_SYSCTL
4073         addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4074         addrconf_sysctl_unregister(&ipv6_devconf);
4075 #endif
4076
4077         rtnl_lock();
4078
4079         /*
4080          *      clean dev list.
4081          */
4082
4083         for (dev=dev_base; dev; dev=dev->next) {
4084                 if ((idev = __in6_dev_get(dev)) == NULL)
4085                         continue;
4086                 addrconf_ifdown(dev, 1);
4087         }
4088         addrconf_ifdown(&loopback_dev, 2);
4089
4090         /*
4091          *      Check hash table.
4092          */
4093
4094         write_lock_bh(&addrconf_hash_lock);
4095         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4096                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4097                         struct inet6_ifaddr *bifa;
4098
4099                         bifa = ifa;
4100                         ifa = ifa->lst_next;
4101                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4102                         /* Do not free it; something is wrong.
4103                            Now we can investigate it with debugger.
4104                          */
4105                 }
4106         }
4107         write_unlock_bh(&addrconf_hash_lock);
4108
4109         del_timer(&addr_chk_timer);
4110
4111         rtnl_unlock();
4112
4113 #ifdef CONFIG_PROC_FS
4114         proc_net_remove("if_inet6");
4115 #endif
4116 }