fedora core 6 1.2949 + vserver 2.2.0
[linux-2.6.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>     
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      $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 static void addrconf_verify(unsigned long);
123
124 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
125 static DEFINE_SPINLOCK(addrconf_verify_lock);
126
127 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
128 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
129
130 static int addrconf_ifdown(struct net_device *dev, int how);
131
132 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
133 static void addrconf_dad_timer(unsigned long data);
134 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
135 static void addrconf_dad_run(struct inet6_dev *idev);
136 static void addrconf_rs_timer(unsigned long data);
137 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
138 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
139
140 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 
141                                 struct prefix_info *pinfo);
142 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
143
144 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
145
146 struct ipv6_devconf ipv6_devconf __read_mostly = {
147         .forwarding             = 0,
148         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
149         .mtu6                   = IPV6_MIN_MTU,
150         .accept_ra              = 1,
151         .accept_redirects       = 1,
152         .autoconf               = 1,
153         .force_mld_version      = 0,
154         .dad_transmits          = 1,
155         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
156         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
157         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
158 #ifdef CONFIG_IPV6_PRIVACY
159         .use_tempaddr           = 0,
160         .temp_valid_lft         = TEMP_VALID_LIFETIME,
161         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
162         .regen_max_retry        = REGEN_MAX_RETRY,
163         .max_desync_factor      = MAX_DESYNC_FACTOR,
164 #endif
165         .max_addresses          = IPV6_MAX_ADDRESSES,
166         .accept_ra_defrtr       = 1,
167         .accept_ra_pinfo        = 1,
168 #ifdef CONFIG_IPV6_ROUTER_PREF
169         .accept_ra_rtr_pref     = 1,
170         .rtr_probe_interval     = 60 * HZ,
171 #ifdef CONFIG_IPV6_ROUTE_INFO
172         .accept_ra_rt_info_max_plen = 0,
173 #endif
174 #endif
175         .proxy_ndp              = 0,
176         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
177 };
178
179 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
180         .forwarding             = 0,
181         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
182         .mtu6                   = IPV6_MIN_MTU,
183         .accept_ra              = 1,
184         .accept_redirects       = 1,
185         .autoconf               = 1,
186         .dad_transmits          = 1,
187         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
188         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
189         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
190 #ifdef CONFIG_IPV6_PRIVACY
191         .use_tempaddr           = 0,
192         .temp_valid_lft         = TEMP_VALID_LIFETIME,
193         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
194         .regen_max_retry        = REGEN_MAX_RETRY,
195         .max_desync_factor      = MAX_DESYNC_FACTOR,
196 #endif
197         .max_addresses          = IPV6_MAX_ADDRESSES,
198         .accept_ra_defrtr       = 1,
199         .accept_ra_pinfo        = 1,
200 #ifdef CONFIG_IPV6_ROUTER_PREF
201         .accept_ra_rtr_pref     = 1,
202         .rtr_probe_interval     = 60 * HZ,
203 #ifdef CONFIG_IPV6_ROUTE_INFO
204         .accept_ra_rt_info_max_plen = 0,
205 #endif
206 #endif
207         .proxy_ndp              = 0,
208         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
209 };
210
211 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
212 #if 0
213 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
214 #endif
215 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
216
217 #define IPV6_ADDR_SCOPE_TYPE(scope)     ((scope) << 16)
218
219 static inline unsigned ipv6_addr_scope2type(unsigned scope)
220 {
221         switch(scope) {
222         case IPV6_ADDR_SCOPE_NODELOCAL:
223                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_NODELOCAL) |
224                         IPV6_ADDR_LOOPBACK);
225         case IPV6_ADDR_SCOPE_LINKLOCAL:
226                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL) |
227                         IPV6_ADDR_LINKLOCAL);
228         case IPV6_ADDR_SCOPE_SITELOCAL:
229                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL) |
230                         IPV6_ADDR_SITELOCAL);
231         }
232         return IPV6_ADDR_SCOPE_TYPE(scope);
233 }
234
235 int __ipv6_addr_type(const struct in6_addr *addr)
236 {
237         __be32 st;
238
239         st = addr->s6_addr32[0];
240
241         /* Consider all addresses with the first three bits different of
242            000 and 111 as unicasts.
243          */
244         if ((st & htonl(0xE0000000)) != htonl(0x00000000) &&
245             (st & htonl(0xE0000000)) != htonl(0xE0000000))
246                 return (IPV6_ADDR_UNICAST | 
247                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));
248
249         if ((st & htonl(0xFF000000)) == htonl(0xFF000000)) {
250                 /* multicast */
251                 /* addr-select 3.1 */
252                 return (IPV6_ADDR_MULTICAST |
253                         ipv6_addr_scope2type(IPV6_ADDR_MC_SCOPE(addr)));
254         }
255
256         if ((st & htonl(0xFFC00000)) == htonl(0xFE800000))
257                 return (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST | 
258                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));               /* addr-select 3.1 */
259         if ((st & htonl(0xFFC00000)) == htonl(0xFEC00000))
260                 return (IPV6_ADDR_SITELOCAL | IPV6_ADDR_UNICAST |
261                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL));               /* addr-select 3.1 */
262
263         if ((addr->s6_addr32[0] | addr->s6_addr32[1]) == 0) {
264                 if (addr->s6_addr32[2] == 0) {
265                         if (addr->s6_addr32[3] == 0)
266                                 return IPV6_ADDR_ANY;
267
268                         if (addr->s6_addr32[3] == htonl(0x00000001))
269                                 return (IPV6_ADDR_LOOPBACK | IPV6_ADDR_UNICAST |
270                                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));       /* addr-select 3.4 */
271
272                         return (IPV6_ADDR_COMPATv4 | IPV6_ADDR_UNICAST |
273                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
274                 }
275
276                 if (addr->s6_addr32[2] == htonl(0x0000ffff))
277                         return (IPV6_ADDR_MAPPED | 
278                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
279         }
280
281         return (IPV6_ADDR_RESERVED | 
282                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.4 */
283 }
284
285 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
286 {
287         if (del_timer(&ifp->timer))
288                 __in6_ifa_put(ifp);
289 }
290
291 enum addrconf_timer_t
292 {
293         AC_NONE,
294         AC_DAD,
295         AC_RS,
296 };
297
298 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
299                                enum addrconf_timer_t what,
300                                unsigned long when)
301 {
302         if (!del_timer(&ifp->timer))
303                 in6_ifa_hold(ifp);
304
305         switch (what) {
306         case AC_DAD:
307                 ifp->timer.function = addrconf_dad_timer;
308                 break;
309         case AC_RS:
310                 ifp->timer.function = addrconf_rs_timer;
311                 break;
312         default:;
313         }
314         ifp->timer.expires = jiffies + when;
315         add_timer(&ifp->timer);
316 }
317
318 /* Nobody refers to this device, we may destroy it. */
319
320 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
321 {
322         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
323         kfree(idev);
324 }
325
326 void in6_dev_finish_destroy(struct inet6_dev *idev)
327 {
328         struct net_device *dev = idev->dev;
329         BUG_TRAP(idev->addr_list==NULL);
330         BUG_TRAP(idev->mc_list==NULL);
331 #ifdef NET_REFCNT_DEBUG
332         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
333 #endif
334         dev_put(dev);
335         if (!idev->dead) {
336                 printk("Freeing alive inet6 device %p\n", idev);
337                 return;
338         }
339         snmp6_free_dev(idev);
340         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
341 }
342
343 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 {
345         struct inet6_dev *ndev;
346         struct in6_addr maddr;
347
348         ASSERT_RTNL();
349
350         if (dev->mtu < IPV6_MIN_MTU)
351                 return NULL;
352
353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
354
355         if (ndev == NULL)
356                 return NULL;
357
358         rwlock_init(&ndev->lock);
359         ndev->dev = dev;
360         memcpy(&ndev->cnf, &ipv6_devconf_dflt, sizeof(ndev->cnf));
361         ndev->cnf.mtu6 = dev->mtu;
362         ndev->cnf.sysctl = NULL;
363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364         if (ndev->nd_parms == NULL) {
365                 kfree(ndev);
366                 return NULL;
367         }
368         /* We refer to the device */
369         dev_hold(dev);
370
371         if (snmp6_alloc_dev(ndev) < 0) {
372                 ADBG((KERN_WARNING
373                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
374                         __FUNCTION__, dev->name));
375                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
376                 ndev->dead = 1;
377                 in6_dev_finish_destroy(ndev);
378                 return NULL;
379         }
380
381         if (snmp6_register_dev(ndev) < 0) {
382                 ADBG((KERN_WARNING
383                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
384                         __FUNCTION__, dev->name));
385                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
386                 ndev->dead = 1;
387                 in6_dev_finish_destroy(ndev);
388                 return NULL;
389         }
390
391         /* One reference from device.  We must do this before
392          * we invoke __ipv6_regen_rndid().
393          */
394         in6_dev_hold(ndev);
395
396 #ifdef CONFIG_IPV6_PRIVACY
397         init_timer(&ndev->regen_timer);
398         ndev->regen_timer.function = ipv6_regen_rndid;
399         ndev->regen_timer.data = (unsigned long) ndev;
400         if ((dev->flags&IFF_LOOPBACK) ||
401             dev->type == ARPHRD_TUNNEL ||
402 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
403             dev->type == ARPHRD_SIT ||
404 #endif
405             dev->type == ARPHRD_NONE) {
406                 printk(KERN_INFO
407                        "%s: Disabled Privacy Extensions\n",
408                        dev->name);
409                 ndev->cnf.use_tempaddr = -1;
410         } else {
411                 in6_dev_hold(ndev);
412                 ipv6_regen_rndid((unsigned long) ndev);
413         }
414 #endif
415
416         if (netif_carrier_ok(dev))
417                 ndev->if_flags |= IF_READY;
418
419
420         ipv6_mc_init_dev(ndev);
421         ndev->tstamp = jiffies;
422 #ifdef CONFIG_SYSCTL
423         neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
424                               NET_IPV6_NEIGH, "ipv6",
425                               &ndisc_ifinfo_sysctl_change,
426                               NULL);
427         addrconf_sysctl_register(ndev, &ndev->cnf);
428 #endif
429         /* protected by rtnl_lock */
430         rcu_assign_pointer(dev->ip6_ptr, ndev);
431
432         /* Join all-node multicast group */
433         ipv6_addr_all_nodes(&maddr);
434         ipv6_dev_mc_inc(dev, &maddr);
435
436         return ndev;
437 }
438
439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
440 {
441         struct inet6_dev *idev;
442
443         ASSERT_RTNL();
444
445         if ((idev = __in6_dev_get(dev)) == NULL) {
446                 if ((idev = ipv6_add_dev(dev)) == NULL)
447                         return NULL;
448         }
449
450         if (dev->flags&IFF_UP)
451                 ipv6_mc_up(idev);
452         return idev;
453 }
454
455 #ifdef CONFIG_SYSCTL
456 static void dev_forward_change(struct inet6_dev *idev)
457 {
458         struct net_device *dev;
459         struct inet6_ifaddr *ifa;
460         struct in6_addr addr;
461
462         if (!idev)
463                 return;
464         dev = idev->dev;
465         if (dev && (dev->flags & IFF_MULTICAST)) {
466                 ipv6_addr_all_routers(&addr);
467         
468                 if (idev->cnf.forwarding)
469                         ipv6_dev_mc_inc(dev, &addr);
470                 else
471                         ipv6_dev_mc_dec(dev, &addr);
472         }
473         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
474                 if (ifa->flags&IFA_F_TENTATIVE)
475                         continue;
476                 if (idev->cnf.forwarding)
477                         addrconf_join_anycast(ifa);
478                 else
479                         addrconf_leave_anycast(ifa);
480         }
481 }
482
483
484 static void addrconf_forward_change(void)
485 {
486         struct net_device *dev;
487         struct inet6_dev *idev;
488
489         read_lock(&dev_base_lock);
490         for (dev=dev_base; dev; dev=dev->next) {
491                 rcu_read_lock();
492                 idev = __in6_dev_get(dev);
493                 if (idev) {
494                         int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
495                         idev->cnf.forwarding = ipv6_devconf.forwarding;
496                         if (changed)
497                                 dev_forward_change(idev);
498                 }
499                 rcu_read_unlock();
500         }
501         read_unlock(&dev_base_lock);
502 }
503 #endif
504
505 /* Nobody refers to this ifaddr, destroy it */
506
507 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
508 {
509         BUG_TRAP(ifp->if_next==NULL);
510         BUG_TRAP(ifp->lst_next==NULL);
511 #ifdef NET_REFCNT_DEBUG
512         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
513 #endif
514
515         in6_dev_put(ifp->idev);
516
517         if (del_timer(&ifp->timer))
518                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
519
520         if (!ifp->dead) {
521                 printk("Freeing alive inet6 address %p\n", ifp);
522                 return;
523         }
524         dst_release(&ifp->rt->u.dst);
525
526         kfree(ifp);
527 }
528
529 static void
530 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
531 {
532         struct inet6_ifaddr *ifa, **ifap;
533         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
534
535         /*
536          * Each device address list is sorted in order of scope -
537          * global before linklocal.
538          */
539         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
540              ifap = &ifa->if_next) {
541                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
542                         break;
543         }
544
545         ifp->if_next = *ifap;
546         *ifap = ifp;
547 }
548
549 /* On success it returns ifp with increased reference count */
550
551 static struct inet6_ifaddr *
552 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
553               int scope, u32 flags)
554 {
555         struct inet6_ifaddr *ifa = NULL;
556         struct rt6_info *rt;
557         int hash;
558         int err = 0;
559
560         rcu_read_lock_bh();
561         if (idev->dead) {
562                 err = -ENODEV;                  /*XXX*/
563                 goto out2;
564         }
565
566         write_lock(&addrconf_hash_lock);
567
568         /* Ignore adding duplicate addresses on an interface */
569         if (ipv6_chk_same_addr(addr, idev->dev)) {
570                 ADBG(("ipv6_add_addr: already assigned\n"));
571                 err = -EEXIST;
572                 goto out;
573         }
574
575         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
576
577         if (ifa == NULL) {
578                 ADBG(("ipv6_add_addr: malloc failed\n"));
579                 err = -ENOBUFS;
580                 goto out;
581         }
582
583         rt = addrconf_dst_alloc(idev, addr, 0);
584         if (IS_ERR(rt)) {
585                 err = PTR_ERR(rt);
586                 goto out;
587         }
588
589         ipv6_addr_copy(&ifa->addr, addr);
590
591         spin_lock_init(&ifa->lock);
592         init_timer(&ifa->timer);
593         ifa->timer.data = (unsigned long) ifa;
594         ifa->scope = scope;
595         ifa->prefix_len = pfxlen;
596         ifa->flags = flags | IFA_F_TENTATIVE;
597         ifa->cstamp = ifa->tstamp = jiffies;
598
599         ifa->rt = rt;
600
601         ifa->idev = idev;
602         in6_dev_hold(idev);
603         /* For caller */
604         in6_ifa_hold(ifa);
605
606         /* Add to big hash table */
607         hash = ipv6_addr_hash(addr);
608
609         ifa->lst_next = inet6_addr_lst[hash];
610         inet6_addr_lst[hash] = ifa;
611         in6_ifa_hold(ifa);
612         write_unlock(&addrconf_hash_lock);
613
614         write_lock(&idev->lock);
615         /* Add to inet6_dev unicast addr list. */
616         ipv6_link_dev_addr(idev, ifa);
617
618 #ifdef CONFIG_IPV6_PRIVACY
619         if (ifa->flags&IFA_F_TEMPORARY) {
620                 ifa->tmp_next = idev->tempaddr_list;
621                 idev->tempaddr_list = ifa;
622                 in6_ifa_hold(ifa);
623         }
624 #endif
625
626         in6_ifa_hold(ifa);
627         write_unlock(&idev->lock);
628 out2:
629         rcu_read_unlock_bh();
630
631         if (likely(err == 0))
632                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
633         else {
634                 kfree(ifa);
635                 ifa = ERR_PTR(err);
636         }
637
638         return ifa;
639 out:
640         write_unlock(&addrconf_hash_lock);
641         goto out2;
642 }
643
644 /* This function wants to get referenced ifp and releases it before return */
645
646 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
647 {
648         struct inet6_ifaddr *ifa, **ifap;
649         struct inet6_dev *idev = ifp->idev;
650         int hash;
651         int deleted = 0, onlink = 0;
652         unsigned long expires = jiffies;
653
654         hash = ipv6_addr_hash(&ifp->addr);
655
656         ifp->dead = 1;
657
658         write_lock_bh(&addrconf_hash_lock);
659         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
660              ifap = &ifa->lst_next) {
661                 if (ifa == ifp) {
662                         *ifap = ifa->lst_next;
663                         __in6_ifa_put(ifp);
664                         ifa->lst_next = NULL;
665                         break;
666                 }
667         }
668         write_unlock_bh(&addrconf_hash_lock);
669
670         write_lock_bh(&idev->lock);
671 #ifdef CONFIG_IPV6_PRIVACY
672         if (ifp->flags&IFA_F_TEMPORARY) {
673                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
674                      ifap = &ifa->tmp_next) {
675                         if (ifa == ifp) {
676                                 *ifap = ifa->tmp_next;
677                                 if (ifp->ifpub) {
678                                         in6_ifa_put(ifp->ifpub);
679                                         ifp->ifpub = NULL;
680                                 }
681                                 __in6_ifa_put(ifp);
682                                 ifa->tmp_next = NULL;
683                                 break;
684                         }
685                 }
686         }
687 #endif
688
689         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
690                 if (ifa == ifp) {
691                         *ifap = ifa->if_next;
692                         __in6_ifa_put(ifp);
693                         ifa->if_next = NULL;
694                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
695                                 break;
696                         deleted = 1;
697                         continue;
698                 } else if (ifp->flags & IFA_F_PERMANENT) {
699                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
700                                               ifp->prefix_len)) {
701                                 if (ifa->flags & IFA_F_PERMANENT) {
702                                         onlink = 1;
703                                         if (deleted)
704                                                 break;
705                                 } else {
706                                         unsigned long lifetime;
707
708                                         if (!onlink)
709                                                 onlink = -1;
710
711                                         spin_lock(&ifa->lock);
712                                         lifetime = min_t(unsigned long,
713                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
714                                         if (time_before(expires,
715                                                         ifa->tstamp + lifetime * HZ))
716                                                 expires = ifa->tstamp + lifetime * HZ;
717                                         spin_unlock(&ifa->lock);
718                                 }
719                         }
720                 }
721                 ifap = &ifa->if_next;
722         }
723         write_unlock_bh(&idev->lock);
724
725         ipv6_ifa_notify(RTM_DELADDR, ifp);
726
727         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
728
729         addrconf_del_timer(ifp);
730
731         /*
732          * Purge or update corresponding prefix
733          *
734          * 1) we don't purge prefix here if address was not permanent.
735          *    prefix is managed by its own lifetime.
736          * 2) if there're no addresses, delete prefix.
737          * 3) if there're still other permanent address(es),
738          *    corresponding prefix is still permanent.
739          * 4) otherwise, update prefix lifetime to the
740          *    longest valid lifetime among the corresponding
741          *    addresses on the device.
742          *    Note: subsequent RA will update lifetime.
743          *
744          * --yoshfuji
745          */
746         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
747                 struct in6_addr prefix;
748                 struct rt6_info *rt;
749
750                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
751                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
752
753                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
754                         if (onlink == 0) {
755                                 ip6_del_rt(rt);
756                                 rt = NULL;
757                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
758                                 rt->rt6i_expires = expires;
759                                 rt->rt6i_flags |= RTF_EXPIRES;
760                         }
761                 }
762                 dst_release(&rt->u.dst);
763         }
764
765         in6_ifa_put(ifp);
766 }
767
768 #ifdef CONFIG_IPV6_PRIVACY
769 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
770 {
771         struct inet6_dev *idev = ifp->idev;
772         struct in6_addr addr, *tmpaddr;
773         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
774         int tmp_plen;
775         int ret = 0;
776         int max_addresses;
777
778         write_lock(&idev->lock);
779         if (ift) {
780                 spin_lock_bh(&ift->lock);
781                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
782                 spin_unlock_bh(&ift->lock);
783                 tmpaddr = &addr;
784         } else {
785                 tmpaddr = NULL;
786         }
787 retry:
788         in6_dev_hold(idev);
789         if (idev->cnf.use_tempaddr <= 0) {
790                 write_unlock(&idev->lock);
791                 printk(KERN_INFO
792                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
793                 in6_dev_put(idev);
794                 ret = -1;
795                 goto out;
796         }
797         spin_lock_bh(&ifp->lock);
798         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
799                 idev->cnf.use_tempaddr = -1;    /*XXX*/
800                 spin_unlock_bh(&ifp->lock);
801                 write_unlock(&idev->lock);
802                 printk(KERN_WARNING
803                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
804                 in6_dev_put(idev);
805                 ret = -1;
806                 goto out;
807         }
808         in6_ifa_hold(ifp);
809         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
810         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
811                 spin_unlock_bh(&ifp->lock);
812                 write_unlock(&idev->lock);
813                 printk(KERN_WARNING
814                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
815                 in6_ifa_put(ifp);
816                 in6_dev_put(idev);
817                 ret = -1;
818                 goto out;
819         }
820         memcpy(&addr.s6_addr[8], idev->rndid, 8);
821         tmp_valid_lft = min_t(__u32,
822                               ifp->valid_lft,
823                               idev->cnf.temp_valid_lft);
824         tmp_prefered_lft = min_t(__u32, 
825                                  ifp->prefered_lft, 
826                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
827         tmp_plen = ifp->prefix_len;
828         max_addresses = idev->cnf.max_addresses;
829         tmp_cstamp = ifp->cstamp;
830         tmp_tstamp = ifp->tstamp;
831         spin_unlock_bh(&ifp->lock);
832
833         write_unlock(&idev->lock);
834         ift = !max_addresses ||
835               ipv6_count_addresses(idev) < max_addresses ? 
836                 ipv6_add_addr(idev, &addr, tmp_plen,
837                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK, IFA_F_TEMPORARY) : NULL;
838         if (!ift || IS_ERR(ift)) {
839                 in6_ifa_put(ifp);
840                 in6_dev_put(idev);
841                 printk(KERN_INFO
842                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
843                 tmpaddr = &addr;
844                 write_lock(&idev->lock);
845                 goto retry;
846         }
847
848         spin_lock_bh(&ift->lock);
849         ift->ifpub = ifp;
850         ift->valid_lft = tmp_valid_lft;
851         ift->prefered_lft = tmp_prefered_lft;
852         ift->cstamp = tmp_cstamp;
853         ift->tstamp = tmp_tstamp;
854         spin_unlock_bh(&ift->lock);
855
856         addrconf_dad_start(ift, 0);
857         in6_ifa_put(ift);
858         in6_dev_put(idev);
859 out:
860         return ret;
861 }
862 #endif
863
864 /*
865  *      Choose an appropriate source address (RFC3484)
866  */
867 struct ipv6_saddr_score {
868         int             addr_type;
869         unsigned int    attrs;
870         int             matchlen;
871         int             scope;
872         unsigned int    rule;
873 };
874
875 #define IPV6_SADDR_SCORE_LOCAL          0x0001
876 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
877 #define IPV6_SADDR_SCORE_HOA            0x0008
878 #define IPV6_SADDR_SCORE_OIF            0x0010
879 #define IPV6_SADDR_SCORE_LABEL          0x0020
880 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
881
882 static int inline ipv6_saddr_preferred(int type)
883 {
884         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
885                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
886                 return 1;
887         return 0;
888 }
889
890 /* static matching label */
891 static int inline ipv6_saddr_label(const struct in6_addr *addr, int type)
892 {
893  /*
894   *     prefix (longest match)  label
895   *     -----------------------------
896   *     ::1/128                 0
897   *     ::/0                    1
898   *     2002::/16               2
899   *     ::/96                   3
900   *     ::ffff:0:0/96           4
901   *     fc00::/7                5
902   *     2001::/32               6
903   */
904         if (type & IPV6_ADDR_LOOPBACK)
905                 return 0;
906         else if (type & IPV6_ADDR_COMPATv4)
907                 return 3;
908         else if (type & IPV6_ADDR_MAPPED)
909                 return 4;
910         else if (addr->s6_addr32[0] == htonl(0x20010000))
911                 return 6;
912         else if (addr->s6_addr16[0] == htons(0x2002))
913                 return 2;
914         else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
915                 return 5;
916         return 1;
917 }
918
919 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
920                        struct in6_addr *daddr, struct in6_addr *saddr)
921 {
922         struct ipv6_saddr_score hiscore;
923         struct inet6_ifaddr *ifa_result = NULL;
924         int daddr_type = __ipv6_addr_type(daddr);
925         int daddr_scope = __ipv6_addr_src_scope(daddr_type);
926         u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
927         struct net_device *dev;
928
929         memset(&hiscore, 0, sizeof(hiscore));
930
931         read_lock(&dev_base_lock);
932         rcu_read_lock();
933
934         for (dev = dev_base; dev; dev=dev->next) {
935                 struct inet6_dev *idev;
936                 struct inet6_ifaddr *ifa;
937
938                 /* Rule 0: Candidate Source Address (section 4)
939                  *  - multicast and link-local destination address,
940                  *    the set of candidate source address MUST only
941                  *    include addresses assigned to interfaces
942                  *    belonging to the same link as the outgoing
943                  *    interface.
944                  * (- For site-local destination addresses, the
945                  *    set of candidate source addresses MUST only
946                  *    include addresses assigned to interfaces
947                  *    belonging to the same site as the outgoing
948                  *    interface.)
949                  */
950                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
951                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
952                     daddr_dev && dev != daddr_dev)
953                         continue;
954
955                 idev = __in6_dev_get(dev);
956                 if (!idev)
957                         continue;
958
959                 read_lock_bh(&idev->lock);
960                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
961                         struct ipv6_saddr_score score;
962
963                         score.addr_type = __ipv6_addr_type(&ifa->addr);
964
965                         /* Rule 0:
966                          * - Tentative Address (RFC2462 section 5.4)
967                          *  - A tentative address is not considered
968                          *    "assigned to an interface" in the traditional
969                          *    sense.
970                          * - Candidate Source Address (section 4)
971                          *  - In any case, anycast addresses, multicast
972                          *    addresses, and the unspecified address MUST
973                          *    NOT be included in a candidate set.
974                          */
975                         if (ifa->flags & IFA_F_TENTATIVE)
976                                 continue;
977                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
978                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
979                                 LIMIT_NETDEBUG(KERN_DEBUG
980                                                "ADDRCONF: unspecified / multicast address"
981                                                "assigned as unicast address on %s",
982                                                dev->name);
983                                 continue;
984                         }
985
986                         score.attrs = 0;
987                         score.matchlen = 0;
988                         score.scope = 0;
989                         score.rule = 0;
990
991                         if (ifa_result == NULL) {
992                                 /* record it if the first available entry */
993                                 goto record_it;
994                         }
995
996                         /* Rule 1: Prefer same address */
997                         if (hiscore.rule < 1) {
998                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
999                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
1000                                 hiscore.rule++;
1001                         }
1002                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
1003                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
1004                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
1005                                         score.rule = 1;
1006                                         goto record_it;
1007                                 }
1008                         } else {
1009                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
1010                                         continue;
1011                         }
1012
1013                         /* Rule 2: Prefer appropriate scope */
1014                         if (hiscore.rule < 2) {
1015                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1016                                 hiscore.rule++;
1017                         }
1018                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1019                         if (hiscore.scope < score.scope) {
1020                                 if (hiscore.scope < daddr_scope) {
1021                                         score.rule = 2;
1022                                         goto record_it;
1023                                 } else
1024                                         continue;
1025                         } else if (score.scope < hiscore.scope) {
1026                                 if (score.scope < daddr_scope)
1027                                         break; /* addresses sorted by scope */
1028                                 else {
1029                                         score.rule = 2;
1030                                         goto record_it;
1031                                 }
1032                         }
1033
1034                         /* Rule 3: Avoid deprecated address */
1035                         if (hiscore.rule < 3) {
1036                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1037                                     !(ifa_result->flags & IFA_F_DEPRECATED))
1038                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1039                                 hiscore.rule++;
1040                         }
1041                         if (ipv6_saddr_preferred(score.addr_type) ||
1042                             !(ifa->flags & IFA_F_DEPRECATED)) {
1043                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1044                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1045                                         score.rule = 3;
1046                                         goto record_it;
1047                                 }
1048                         } else {
1049                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1050                                         continue;
1051                         }
1052
1053                         /* Rule 4: Prefer home address */
1054 #ifdef CONFIG_IPV6_MIP6
1055                         if (hiscore.rule < 4) {
1056                                 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1057                                         hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1058                                 hiscore.rule++;
1059                         }
1060                         if (ifa->flags & IFA_F_HOMEADDRESS) {
1061                                 score.attrs |= IPV6_SADDR_SCORE_HOA;
1062                                 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1063                                         score.rule = 4;
1064                                         goto record_it;
1065                                 }
1066                         } else {
1067                                 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1068                                         continue;
1069                         }
1070 #else
1071                         if (hiscore.rule < 4)
1072                                 hiscore.rule++;
1073 #endif
1074
1075                         /* Rule 5: Prefer outgoing interface */
1076                         if (hiscore.rule < 5) {
1077                                 if (daddr_dev == NULL ||
1078                                     daddr_dev == ifa_result->idev->dev)
1079                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1080                                 hiscore.rule++;
1081                         }
1082                         if (daddr_dev == NULL ||
1083                             daddr_dev == ifa->idev->dev) {
1084                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1085                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1086                                         score.rule = 5;
1087                                         goto record_it;
1088                                 }
1089                         } else {
1090                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1091                                         continue;
1092                         }
1093
1094                         /* Rule 6: Prefer matching label */
1095                         if (hiscore.rule < 6) {
1096                                 if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1097                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1098                                 hiscore.rule++;
1099                         }
1100                         if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1101                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1102                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1103                                         score.rule = 6;
1104                                         goto record_it;
1105                                 }
1106                         } else {
1107                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1108                                         continue;
1109                         }
1110
1111 #ifdef CONFIG_IPV6_PRIVACY
1112                         /* Rule 7: Prefer public address
1113                          * Note: prefer temprary address if use_tempaddr >= 2
1114                          */
1115                         if (hiscore.rule < 7) {
1116                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1117                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1118                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1119                                 hiscore.rule++;
1120                         }
1121                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1122                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1123                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1124                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1125                                         score.rule = 7;
1126                                         goto record_it;
1127                                 }
1128                         } else {
1129                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1130                                         continue;
1131                         }
1132 #else
1133                         if (hiscore.rule < 7)
1134                                 hiscore.rule++;
1135 #endif
1136                         /* Rule 8: Use longest matching prefix */
1137                         if (hiscore.rule < 8) {
1138                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1139                                 hiscore.rule++;
1140                         }
1141                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1142                         if (score.matchlen > hiscore.matchlen) {
1143                                 score.rule = 8;
1144                                 goto record_it;
1145                         }
1146 #if 0
1147                         else if (score.matchlen < hiscore.matchlen)
1148                                 continue;
1149 #endif
1150
1151                         /* Final Rule: choose first available one */
1152                         continue;
1153 record_it:
1154                         if (ifa_result)
1155                                 in6_ifa_put(ifa_result);
1156                         in6_ifa_hold(ifa);
1157                         ifa_result = ifa;
1158                         hiscore = score;
1159                 }
1160                 read_unlock_bh(&idev->lock);
1161         }
1162         rcu_read_unlock();
1163         read_unlock(&dev_base_lock);
1164
1165         if (!ifa_result)
1166                 return -EADDRNOTAVAIL;
1167         
1168         ipv6_addr_copy(saddr, &ifa_result->addr);
1169         in6_ifa_put(ifa_result);
1170         return 0;
1171 }
1172
1173
1174 int ipv6_get_saddr(struct dst_entry *dst,
1175                    struct in6_addr *daddr, struct in6_addr *saddr)
1176 {
1177         return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1178 }
1179
1180
1181 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
1182 {
1183         struct inet6_dev *idev;
1184         int err = -EADDRNOTAVAIL;
1185
1186         rcu_read_lock();
1187         if ((idev = __in6_dev_get(dev)) != NULL) {
1188                 struct inet6_ifaddr *ifp;
1189
1190                 read_lock_bh(&idev->lock);
1191                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1192                         if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1193                                 ipv6_addr_copy(addr, &ifp->addr);
1194                                 err = 0;
1195                                 break;
1196                         }
1197                 }
1198                 read_unlock_bh(&idev->lock);
1199         }
1200         rcu_read_unlock();
1201         return err;
1202 }
1203
1204 static int ipv6_count_addresses(struct inet6_dev *idev)
1205 {
1206         int cnt = 0;
1207         struct inet6_ifaddr *ifp;
1208
1209         read_lock_bh(&idev->lock);
1210         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1211                 cnt++;
1212         read_unlock_bh(&idev->lock);
1213         return cnt;
1214 }
1215
1216 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1217 {
1218         struct inet6_ifaddr * ifp;
1219         u8 hash = ipv6_addr_hash(addr);
1220
1221         read_lock_bh(&addrconf_hash_lock);
1222         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1223                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1224                     !(ifp->flags&IFA_F_TENTATIVE)) {
1225                         if (dev == NULL || ifp->idev->dev == dev ||
1226                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1227                                 break;
1228                 }
1229         }
1230         read_unlock_bh(&addrconf_hash_lock);
1231         return ifp != NULL;
1232 }
1233
1234 static
1235 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1236 {
1237         struct inet6_ifaddr * ifp;
1238         u8 hash = ipv6_addr_hash(addr);
1239
1240         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1241                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1242                         if (dev == NULL || ifp->idev->dev == dev)
1243                                 break;
1244                 }
1245         }
1246         return ifp != NULL;
1247 }
1248
1249 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1250 {
1251         struct inet6_ifaddr * ifp;
1252         u8 hash = ipv6_addr_hash(addr);
1253
1254         read_lock_bh(&addrconf_hash_lock);
1255         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1256                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1257                         if (dev == NULL || ifp->idev->dev == dev ||
1258                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1259                                 in6_ifa_hold(ifp);
1260                                 break;
1261                         }
1262                 }
1263         }
1264         read_unlock_bh(&addrconf_hash_lock);
1265
1266         return ifp;
1267 }
1268
1269 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1270 {
1271         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1272         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1273         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1274         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1275         int sk_ipv6only = ipv6_only_sock(sk);
1276         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1277         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1278         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1279
1280         if (!sk2_rcv_saddr && !sk_ipv6only)
1281                 return 1;
1282
1283         if (addr_type2 == IPV6_ADDR_ANY &&
1284             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1285                 return 1;
1286
1287         if (addr_type == IPV6_ADDR_ANY &&
1288             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1289                 return 1;
1290
1291         if (sk2_rcv_saddr6 &&
1292             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1293                 return 1;
1294
1295         if (addr_type == IPV6_ADDR_MAPPED &&
1296             !sk2_ipv6only &&
1297             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1298                 return 1;
1299
1300         return 0;
1301 }
1302
1303 /* Gets referenced address, destroys ifaddr */
1304
1305 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1306 {
1307         if (ifp->flags&IFA_F_PERMANENT) {
1308                 spin_lock_bh(&ifp->lock);
1309                 addrconf_del_timer(ifp);
1310                 ifp->flags |= IFA_F_TENTATIVE;
1311                 spin_unlock_bh(&ifp->lock);
1312                 in6_ifa_put(ifp);
1313 #ifdef CONFIG_IPV6_PRIVACY
1314         } else if (ifp->flags&IFA_F_TEMPORARY) {
1315                 struct inet6_ifaddr *ifpub;
1316                 spin_lock_bh(&ifp->lock);
1317                 ifpub = ifp->ifpub;
1318                 if (ifpub) {
1319                         in6_ifa_hold(ifpub);
1320                         spin_unlock_bh(&ifp->lock);
1321                         ipv6_create_tempaddr(ifpub, ifp);
1322                         in6_ifa_put(ifpub);
1323                 } else {
1324                         spin_unlock_bh(&ifp->lock);
1325                 }
1326                 ipv6_del_addr(ifp);
1327 #endif
1328         } else
1329                 ipv6_del_addr(ifp);
1330 }
1331
1332 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1333 {
1334         if (net_ratelimit())
1335                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1336         addrconf_dad_stop(ifp);
1337 }
1338
1339 /* Join to solicited addr multicast group. */
1340
1341 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1342 {
1343         struct in6_addr maddr;
1344
1345         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1346                 return;
1347
1348         addrconf_addr_solict_mult(addr, &maddr);
1349         ipv6_dev_mc_inc(dev, &maddr);
1350 }
1351
1352 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1353 {
1354         struct in6_addr maddr;
1355
1356         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1357                 return;
1358
1359         addrconf_addr_solict_mult(addr, &maddr);
1360         __ipv6_dev_mc_dec(idev, &maddr);
1361 }
1362
1363 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1364 {
1365         struct in6_addr addr;
1366         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1367         if (ipv6_addr_any(&addr))
1368                 return;
1369         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1370 }
1371
1372 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1373 {
1374         struct in6_addr addr;
1375         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1376         if (ipv6_addr_any(&addr))
1377                 return;
1378         __ipv6_dev_ac_dec(ifp->idev, &addr);
1379 }
1380
1381 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1382 {
1383         if (dev->addr_len != ETH_ALEN)
1384                 return -1;
1385         memcpy(eui, dev->dev_addr, 3);
1386         memcpy(eui + 5, dev->dev_addr + 3, 3);
1387
1388         /*
1389          * The zSeries OSA network cards can be shared among various
1390          * OS instances, but the OSA cards have only one MAC address.
1391          * This leads to duplicate address conflicts in conjunction
1392          * with IPv6 if more than one instance uses the same card.
1393          *
1394          * The driver for these cards can deliver a unique 16-bit
1395          * identifier for each instance sharing the same card.  It is
1396          * placed instead of 0xFFFE in the interface identifier.  The
1397          * "u" bit of the interface identifier is not inverted in this
1398          * case.  Hence the resulting interface identifier has local
1399          * scope according to RFC2373.
1400          */
1401         if (dev->dev_id) {
1402                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1403                 eui[4] = dev->dev_id & 0xFF;
1404         } else {
1405                 eui[3] = 0xFF;
1406                 eui[4] = 0xFE;
1407                 eui[0] ^= 2;
1408         }
1409         return 0;
1410 }
1411
1412 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1413 {
1414         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1415         if (dev->addr_len != ARCNET_ALEN)
1416                 return -1;
1417         memset(eui, 0, 7);
1418         eui[7] = *(u8*)dev->dev_addr;
1419         return 0;
1420 }
1421
1422 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1423 {
1424         if (dev->addr_len != INFINIBAND_ALEN)
1425                 return -1;
1426         memcpy(eui, dev->dev_addr + 12, 8);
1427         eui[0] |= 2;
1428         return 0;
1429 }
1430
1431 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1432 {
1433         switch (dev->type) {
1434         case ARPHRD_ETHER:
1435         case ARPHRD_FDDI:
1436         case ARPHRD_IEEE802_TR:
1437                 return addrconf_ifid_eui48(eui, dev);
1438         case ARPHRD_ARCNET:
1439                 return addrconf_ifid_arcnet(eui, dev);
1440         case ARPHRD_INFINIBAND:
1441                 return addrconf_ifid_infiniband(eui, dev);
1442         }
1443         return -1;
1444 }
1445
1446 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1447 {
1448         int err = -1;
1449         struct inet6_ifaddr *ifp;
1450
1451         read_lock_bh(&idev->lock);
1452         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1453                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1454                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1455                         err = 0;
1456                         break;
1457                 }
1458         }
1459         read_unlock_bh(&idev->lock);
1460         return err;
1461 }
1462
1463 #ifdef CONFIG_IPV6_PRIVACY
1464 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1465 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1466 {
1467 regen:
1468         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1469         idev->rndid[0] &= ~0x02;
1470
1471         /*
1472          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1473          * check if generated address is not inappropriate
1474          *
1475          *  - Reserved subnet anycast (RFC 2526)
1476          *      11111101 11....11 1xxxxxxx
1477          *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1478          *      00-00-5E-FE-xx-xx-xx-xx
1479          *  - value 0
1480          *  - XXX: already assigned to an address on the device
1481          */
1482         if (idev->rndid[0] == 0xfd && 
1483             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1484             (idev->rndid[7]&0x80))
1485                 goto regen;
1486         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1487                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1488                         goto regen;
1489                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1490                         goto regen;
1491         }
1492
1493         return 0;
1494 }
1495
1496 static void ipv6_regen_rndid(unsigned long data)
1497 {
1498         struct inet6_dev *idev = (struct inet6_dev *) data;
1499         unsigned long expires;
1500
1501         rcu_read_lock_bh();
1502         write_lock_bh(&idev->lock);
1503
1504         if (idev->dead)
1505                 goto out;
1506
1507         if (__ipv6_regen_rndid(idev) < 0)
1508                 goto out;
1509         
1510         expires = jiffies +
1511                 idev->cnf.temp_prefered_lft * HZ - 
1512                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1513         if (time_before(expires, jiffies)) {
1514                 printk(KERN_WARNING
1515                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1516                         idev->dev->name);
1517                 goto out;
1518         }
1519
1520         if (!mod_timer(&idev->regen_timer, expires))
1521                 in6_dev_hold(idev);
1522
1523 out:
1524         write_unlock_bh(&idev->lock);
1525         rcu_read_unlock_bh();
1526         in6_dev_put(idev);
1527 }
1528
1529 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1530         int ret = 0;
1531
1532         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1533                 ret = __ipv6_regen_rndid(idev);
1534         return ret;
1535 }
1536 #endif
1537
1538 /*
1539  *      Add prefix route.
1540  */
1541
1542 static void
1543 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1544                       unsigned long expires, u32 flags)
1545 {
1546         struct fib6_config cfg = {
1547                 .fc_table = RT6_TABLE_PREFIX,
1548                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1549                 .fc_ifindex = dev->ifindex,
1550                 .fc_expires = expires,
1551                 .fc_dst_len = plen,
1552                 .fc_flags = RTF_UP | flags,
1553         };
1554
1555         ipv6_addr_copy(&cfg.fc_dst, pfx);
1556
1557         /* Prevent useless cloning on PtP SIT.
1558            This thing is done here expecting that the whole
1559            class of non-broadcast devices need not cloning.
1560          */
1561 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1562         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1563                 cfg.fc_flags |= RTF_NONEXTHOP;
1564 #endif
1565
1566         ip6_route_add(&cfg);
1567 }
1568
1569 /* Create "default" multicast route to the interface */
1570
1571 static void addrconf_add_mroute(struct net_device *dev)
1572 {
1573         struct fib6_config cfg = {
1574                 .fc_table = RT6_TABLE_LOCAL,
1575                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1576                 .fc_ifindex = dev->ifindex,
1577                 .fc_dst_len = 8,
1578                 .fc_flags = RTF_UP,
1579         };
1580
1581         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1582
1583         ip6_route_add(&cfg);
1584 }
1585
1586 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1587 static void sit_route_add(struct net_device *dev)
1588 {
1589         struct fib6_config cfg = {
1590                 .fc_table = RT6_TABLE_MAIN,
1591                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1592                 .fc_ifindex = dev->ifindex,
1593                 .fc_dst_len = 96,
1594                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1595         };
1596
1597         /* prefix length - 96 bits "::d.d.d.d" */
1598         ip6_route_add(&cfg);
1599 }
1600 #endif
1601
1602 static void addrconf_add_lroute(struct net_device *dev)
1603 {
1604         struct in6_addr addr;
1605
1606         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1607         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1608 }
1609
1610 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1611 {
1612         struct inet6_dev *idev;
1613
1614         ASSERT_RTNL();
1615
1616         if ((idev = ipv6_find_idev(dev)) == NULL)
1617                 return NULL;
1618
1619         /* Add default multicast route */
1620         addrconf_add_mroute(dev);
1621
1622         /* Add link local route */
1623         addrconf_add_lroute(dev);
1624         return idev;
1625 }
1626
1627 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1628 {
1629         struct prefix_info *pinfo;
1630         __u32 valid_lft;
1631         __u32 prefered_lft;
1632         int addr_type;
1633         unsigned long rt_expires;
1634         struct inet6_dev *in6_dev;
1635
1636         pinfo = (struct prefix_info *) opt;
1637         
1638         if (len < sizeof(struct prefix_info)) {
1639                 ADBG(("addrconf: prefix option too short\n"));
1640                 return;
1641         }
1642         
1643         /*
1644          *      Validation checks ([ADDRCONF], page 19)
1645          */
1646
1647         addr_type = ipv6_addr_type(&pinfo->prefix);
1648
1649         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1650                 return;
1651
1652         valid_lft = ntohl(pinfo->valid);
1653         prefered_lft = ntohl(pinfo->prefered);
1654
1655         if (prefered_lft > valid_lft) {
1656                 if (net_ratelimit())
1657                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1658                 return;
1659         }
1660
1661         in6_dev = in6_dev_get(dev);
1662
1663         if (in6_dev == NULL) {
1664                 if (net_ratelimit())
1665                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1666                 return;
1667         }
1668
1669         /*
1670          *      Two things going on here:
1671          *      1) Add routes for on-link prefixes
1672          *      2) Configure prefixes with the auto flag set
1673          */
1674
1675         /* Avoid arithmetic overflow. Really, we could
1676            save rt_expires in seconds, likely valid_lft,
1677            but it would require division in fib gc, that it
1678            not good.
1679          */
1680         if (valid_lft >= 0x7FFFFFFF/HZ)
1681                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1682         else
1683                 rt_expires = valid_lft * HZ;
1684
1685         /*
1686          * We convert this (in jiffies) to clock_t later.
1687          * Avoid arithmetic overflow there as well.
1688          * Overflow can happen only if HZ < USER_HZ.
1689          */
1690         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1691                 rt_expires = 0x7FFFFFFF / USER_HZ;
1692
1693         if (pinfo->onlink) {
1694                 struct rt6_info *rt;
1695                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1696
1697                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1698                         if (rt->rt6i_flags&RTF_EXPIRES) {
1699                                 if (valid_lft == 0) {
1700                                         ip6_del_rt(rt);
1701                                         rt = NULL;
1702                                 } else {
1703                                         rt->rt6i_expires = jiffies + rt_expires;
1704                                 }
1705                         }
1706                 } else if (valid_lft) {
1707                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1708                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1709                 }
1710                 if (rt)
1711                         dst_release(&rt->u.dst);
1712         }
1713
1714         /* Try to figure out our local address for this prefix */
1715
1716         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1717                 struct inet6_ifaddr * ifp;
1718                 struct in6_addr addr;
1719                 int create = 0, update_lft = 0;
1720
1721                 if (pinfo->prefix_len == 64) {
1722                         memcpy(&addr, &pinfo->prefix, 8);
1723                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1724                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1725                                 in6_dev_put(in6_dev);
1726                                 return;
1727                         }
1728                         goto ok;
1729                 }
1730                 if (net_ratelimit())
1731                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1732                                pinfo->prefix_len);
1733                 in6_dev_put(in6_dev);
1734                 return;
1735
1736 ok:
1737
1738                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1739
1740                 if (ifp == NULL && valid_lft) {
1741                         int max_addresses = in6_dev->cnf.max_addresses;
1742
1743                         /* Do not allow to create too much of autoconfigured
1744                          * addresses; this would be too easy way to crash kernel.
1745                          */
1746                         if (!max_addresses ||
1747                             ipv6_count_addresses(in6_dev) < max_addresses)
1748                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1749                                                     addr_type&IPV6_ADDR_SCOPE_MASK, 0);
1750
1751                         if (!ifp || IS_ERR(ifp)) {
1752                                 in6_dev_put(in6_dev);
1753                                 return;
1754                         }
1755
1756                         update_lft = create = 1;
1757                         ifp->cstamp = jiffies;
1758                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1759                 }
1760
1761                 if (ifp) {
1762                         int flags;
1763                         unsigned long now;
1764 #ifdef CONFIG_IPV6_PRIVACY
1765                         struct inet6_ifaddr *ift;
1766 #endif
1767                         u32 stored_lft;
1768
1769                         /* update lifetime (RFC2462 5.5.3 e) */
1770                         spin_lock(&ifp->lock);
1771                         now = jiffies;
1772                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1773                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1774                         else
1775                                 stored_lft = 0;
1776                         if (!update_lft && stored_lft) {
1777                                 if (valid_lft > MIN_VALID_LIFETIME ||
1778                                     valid_lft > stored_lft)
1779                                         update_lft = 1;
1780                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1781                                         /* valid_lft <= stored_lft is always true */
1782                                         /* XXX: IPsec */
1783                                         update_lft = 0;
1784                                 } else {
1785                                         valid_lft = MIN_VALID_LIFETIME;
1786                                         if (valid_lft < prefered_lft)
1787                                                 prefered_lft = valid_lft;
1788                                         update_lft = 1;
1789                                 }
1790                         }
1791
1792                         if (update_lft) {
1793                                 ifp->valid_lft = valid_lft;
1794                                 ifp->prefered_lft = prefered_lft;
1795                                 ifp->tstamp = now;
1796                                 flags = ifp->flags;
1797                                 ifp->flags &= ~IFA_F_DEPRECATED;
1798                                 spin_unlock(&ifp->lock);
1799
1800                                 if (!(flags&IFA_F_TENTATIVE))
1801                                         ipv6_ifa_notify(0, ifp);
1802                         } else
1803                                 spin_unlock(&ifp->lock);
1804
1805 #ifdef CONFIG_IPV6_PRIVACY
1806                         read_lock_bh(&in6_dev->lock);
1807                         /* update all temporary addresses in the list */
1808                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1809                                 /*
1810                                  * When adjusting the lifetimes of an existing
1811                                  * temporary address, only lower the lifetimes.
1812                                  * Implementations must not increase the
1813                                  * lifetimes of an existing temporary address
1814                                  * when processing a Prefix Information Option.
1815                                  */
1816                                 spin_lock(&ift->lock);
1817                                 flags = ift->flags;
1818                                 if (ift->valid_lft > valid_lft &&
1819                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1820                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1821                                 if (ift->prefered_lft > prefered_lft &&
1822                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1823                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1824                                 spin_unlock(&ift->lock);
1825                                 if (!(flags&IFA_F_TENTATIVE))
1826                                         ipv6_ifa_notify(0, ift);
1827                         }
1828
1829                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1830                                 /*
1831                                  * When a new public address is created as described in [ADDRCONF],
1832                                  * also create a new temporary address.
1833                                  */
1834                                 read_unlock_bh(&in6_dev->lock); 
1835                                 ipv6_create_tempaddr(ifp, NULL);
1836                         } else {
1837                                 read_unlock_bh(&in6_dev->lock);
1838                         }
1839 #endif
1840                         in6_ifa_put(ifp);
1841                         addrconf_verify(0);
1842                 }
1843         }
1844         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1845         in6_dev_put(in6_dev);
1846 }
1847
1848 /*
1849  *      Set destination address.
1850  *      Special case for SIT interfaces where we create a new "virtual"
1851  *      device.
1852  */
1853 int addrconf_set_dstaddr(void __user *arg)
1854 {
1855         struct in6_ifreq ireq;
1856         struct net_device *dev;
1857         int err = -EINVAL;
1858
1859         rtnl_lock();
1860
1861         err = -EFAULT;
1862         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1863                 goto err_exit;
1864
1865         dev = __dev_get_by_index(ireq.ifr6_ifindex);
1866
1867         err = -ENODEV;
1868         if (dev == NULL)
1869                 goto err_exit;
1870
1871 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1872         if (dev->type == ARPHRD_SIT) {
1873                 struct ifreq ifr;
1874                 mm_segment_t    oldfs;
1875                 struct ip_tunnel_parm p;
1876
1877                 err = -EADDRNOTAVAIL;
1878                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1879                         goto err_exit;
1880
1881                 memset(&p, 0, sizeof(p));
1882                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1883                 p.iph.saddr = 0;
1884                 p.iph.version = 4;
1885                 p.iph.ihl = 5;
1886                 p.iph.protocol = IPPROTO_IPV6;
1887                 p.iph.ttl = 64;
1888                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1889
1890                 oldfs = get_fs(); set_fs(KERNEL_DS);
1891                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1892                 set_fs(oldfs);
1893
1894                 if (err == 0) {
1895                         err = -ENOBUFS;
1896                         if ((dev = __dev_get_by_name(p.name)) == NULL)
1897                                 goto err_exit;
1898                         err = dev_open(dev);
1899                 }
1900         }
1901 #endif
1902
1903 err_exit:
1904         rtnl_unlock();
1905         return err;
1906 }
1907
1908 /*
1909  *      Manual configuration of address on an interface
1910  */
1911 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1912                           __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1913 {
1914         struct inet6_ifaddr *ifp;
1915         struct inet6_dev *idev;
1916         struct net_device *dev;
1917         int scope;
1918
1919         ASSERT_RTNL();
1920         
1921         /* check the lifetime */
1922         if (!valid_lft || prefered_lft > valid_lft)
1923                 return -EINVAL;
1924
1925         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1926                 return -ENODEV;
1927         
1928         if ((idev = addrconf_add_dev(dev)) == NULL)
1929                 return -ENOBUFS;
1930
1931         scope = ipv6_addr_scope(pfx);
1932
1933         if (valid_lft == INFINITY_LIFE_TIME)
1934                 ifa_flags |= IFA_F_PERMANENT;
1935         else if (valid_lft >= 0x7FFFFFFF/HZ)
1936                 valid_lft = 0x7FFFFFFF/HZ;
1937
1938         if (prefered_lft == 0)
1939                 ifa_flags |= IFA_F_DEPRECATED;
1940         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1941                  (prefered_lft != INFINITY_LIFE_TIME))
1942                 prefered_lft = 0x7FFFFFFF/HZ;
1943
1944         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1945
1946         if (!IS_ERR(ifp)) {
1947                 spin_lock_bh(&ifp->lock);
1948                 ifp->valid_lft = valid_lft;
1949                 ifp->prefered_lft = prefered_lft;
1950                 ifp->tstamp = jiffies;
1951                 spin_unlock_bh(&ifp->lock);
1952
1953                 addrconf_dad_start(ifp, 0);
1954                 in6_ifa_put(ifp);
1955                 addrconf_verify(0);
1956                 return 0;
1957         }
1958
1959         return PTR_ERR(ifp);
1960 }
1961
1962 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1963 {
1964         struct inet6_ifaddr *ifp;
1965         struct inet6_dev *idev;
1966         struct net_device *dev;
1967         
1968         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1969                 return -ENODEV;
1970
1971         if ((idev = __in6_dev_get(dev)) == NULL)
1972                 return -ENXIO;
1973
1974         read_lock_bh(&idev->lock);
1975         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1976                 if (ifp->prefix_len == plen &&
1977                     ipv6_addr_equal(pfx, &ifp->addr)) {
1978                         in6_ifa_hold(ifp);
1979                         read_unlock_bh(&idev->lock);
1980                         
1981                         ipv6_del_addr(ifp);
1982
1983                         /* If the last address is deleted administratively,
1984                            disable IPv6 on this interface.
1985                          */
1986                         if (idev->addr_list == NULL)
1987                                 addrconf_ifdown(idev->dev, 1);
1988                         return 0;
1989                 }
1990         }
1991         read_unlock_bh(&idev->lock);
1992         return -EADDRNOTAVAIL;
1993 }
1994
1995
1996 int addrconf_add_ifaddr(void __user *arg)
1997 {
1998         struct in6_ifreq ireq;
1999         int err;
2000         
2001         if (!capable(CAP_NET_ADMIN))
2002                 return -EPERM;
2003         
2004         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2005                 return -EFAULT;
2006
2007         rtnl_lock();
2008         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2009                              IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2010         rtnl_unlock();
2011         return err;
2012 }
2013
2014 int addrconf_del_ifaddr(void __user *arg)
2015 {
2016         struct in6_ifreq ireq;
2017         int err;
2018         
2019         if (!capable(CAP_NET_ADMIN))
2020                 return -EPERM;
2021
2022         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2023                 return -EFAULT;
2024
2025         rtnl_lock();
2026         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2027         rtnl_unlock();
2028         return err;
2029 }
2030
2031 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2032 static void sit_add_v4_addrs(struct inet6_dev *idev)
2033 {
2034         struct inet6_ifaddr * ifp;
2035         struct in6_addr addr;
2036         struct net_device *dev;
2037         int scope;
2038
2039         ASSERT_RTNL();
2040
2041         memset(&addr, 0, sizeof(struct in6_addr));
2042         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2043
2044         if (idev->dev->flags&IFF_POINTOPOINT) {
2045                 addr.s6_addr32[0] = htonl(0xfe800000);
2046                 scope = IFA_LINK;
2047         } else {
2048                 scope = IPV6_ADDR_COMPATv4;
2049         }
2050
2051         if (addr.s6_addr32[3]) {
2052                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2053                 if (!IS_ERR(ifp)) {
2054                         spin_lock_bh(&ifp->lock);
2055                         ifp->flags &= ~IFA_F_TENTATIVE;
2056                         spin_unlock_bh(&ifp->lock);
2057                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2058                         in6_ifa_put(ifp);
2059                 }
2060                 return;
2061         }
2062
2063         for (dev = dev_base; dev != NULL; dev = dev->next) {
2064                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2065                 if (in_dev && (dev->flags & IFF_UP)) {
2066                         struct in_ifaddr * ifa;
2067
2068                         int flag = scope;
2069
2070                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2071                                 int plen;
2072
2073                                 addr.s6_addr32[3] = ifa->ifa_local;
2074
2075                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2076                                         continue;
2077                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2078                                         if (idev->dev->flags&IFF_POINTOPOINT)
2079                                                 continue;
2080                                         flag |= IFA_HOST;
2081                                 }
2082                                 if (idev->dev->flags&IFF_POINTOPOINT)
2083                                         plen = 64;
2084                                 else
2085                                         plen = 96;
2086
2087                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2088                                                     IFA_F_PERMANENT);
2089                                 if (!IS_ERR(ifp)) {
2090                                         spin_lock_bh(&ifp->lock);
2091                                         ifp->flags &= ~IFA_F_TENTATIVE;
2092                                         spin_unlock_bh(&ifp->lock);
2093                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2094                                         in6_ifa_put(ifp);
2095                                 }
2096                         }
2097                 }
2098         }
2099 }
2100 #endif
2101
2102 static void init_loopback(struct net_device *dev)
2103 {
2104         struct inet6_dev  *idev;
2105         struct inet6_ifaddr * ifp;
2106
2107         /* ::1 */
2108
2109         ASSERT_RTNL();
2110
2111         if ((idev = ipv6_find_idev(dev)) == NULL) {
2112                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2113                 return;
2114         }
2115
2116         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2117         if (!IS_ERR(ifp)) {
2118                 spin_lock_bh(&ifp->lock);
2119                 ifp->flags &= ~IFA_F_TENTATIVE;
2120                 spin_unlock_bh(&ifp->lock);
2121                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2122                 in6_ifa_put(ifp);
2123         }
2124 }
2125
2126 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2127 {
2128         struct inet6_ifaddr * ifp;
2129
2130         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, IFA_F_PERMANENT);
2131         if (!IS_ERR(ifp)) {
2132                 addrconf_dad_start(ifp, 0);
2133                 in6_ifa_put(ifp);
2134         }
2135 }
2136
2137 static void addrconf_dev_config(struct net_device *dev)
2138 {
2139         struct in6_addr addr;
2140         struct inet6_dev    * idev;
2141
2142         ASSERT_RTNL();
2143
2144         if ((dev->type != ARPHRD_ETHER) && 
2145             (dev->type != ARPHRD_FDDI) &&
2146             (dev->type != ARPHRD_IEEE802_TR) &&
2147             (dev->type != ARPHRD_ARCNET) &&
2148             (dev->type != ARPHRD_INFINIBAND)) {
2149                 /* Alas, we support only Ethernet autoconfiguration. */
2150                 return;
2151         }
2152
2153         idev = addrconf_add_dev(dev);
2154         if (idev == NULL)
2155                 return;
2156
2157         memset(&addr, 0, sizeof(struct in6_addr));
2158         addr.s6_addr32[0] = htonl(0xFE800000);
2159
2160         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2161                 addrconf_add_linklocal(idev, &addr);
2162 }
2163
2164 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2165 static void addrconf_sit_config(struct net_device *dev)
2166 {
2167         struct inet6_dev *idev;
2168
2169         ASSERT_RTNL();
2170
2171         /* 
2172          * Configure the tunnel with one of our IPv4 
2173          * addresses... we should configure all of 
2174          * our v4 addrs in the tunnel
2175          */
2176
2177         if ((idev = ipv6_find_idev(dev)) == NULL) {
2178                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2179                 return;
2180         }
2181
2182         sit_add_v4_addrs(idev);
2183
2184         if (dev->flags&IFF_POINTOPOINT) {
2185                 addrconf_add_mroute(dev);
2186                 addrconf_add_lroute(dev);
2187         } else
2188                 sit_route_add(dev);
2189 }
2190 #endif
2191
2192 static inline int
2193 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2194 {
2195         struct in6_addr lladdr;
2196
2197         if (!ipv6_get_lladdr(link_dev, &lladdr)) {
2198                 addrconf_add_linklocal(idev, &lladdr);
2199                 return 0;
2200         }
2201         return -1;
2202 }
2203
2204 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2205 {
2206         struct net_device *link_dev;
2207
2208         /* first try to inherit the link-local address from the link device */
2209         if (idev->dev->iflink &&
2210             (link_dev = __dev_get_by_index(idev->dev->iflink))) {
2211                 if (!ipv6_inherit_linklocal(idev, link_dev))
2212                         return;
2213         }
2214         /* then try to inherit it from any device */
2215         for (link_dev = dev_base; link_dev; link_dev = link_dev->next) {
2216                 if (!ipv6_inherit_linklocal(idev, link_dev))
2217                         return;
2218         }
2219         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2220 }
2221
2222 /*
2223  * Autoconfigure tunnel with a link-local address so routing protocols,
2224  * DHCPv6, MLD etc. can be run over the virtual link
2225  */
2226
2227 static void addrconf_ip6_tnl_config(struct net_device *dev)
2228 {
2229         struct inet6_dev *idev;
2230
2231         ASSERT_RTNL();
2232
2233         if ((idev = addrconf_add_dev(dev)) == NULL) {
2234                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2235                 return;
2236         }
2237         ip6_tnl_add_linklocal(idev);
2238 }
2239
2240 static int addrconf_notify(struct notifier_block *this, unsigned long event, 
2241                            void * data)
2242 {
2243         struct net_device *dev = (struct net_device *) data;
2244         struct inet6_dev *idev = __in6_dev_get(dev);
2245         int run_pending = 0;
2246
2247         switch(event) {
2248         case NETDEV_UP:
2249         case NETDEV_CHANGE:
2250                 if (event == NETDEV_UP) {
2251                         if (!netif_carrier_ok(dev)) {
2252                                 /* device is not ready yet. */
2253                                 printk(KERN_INFO
2254                                         "ADDRCONF(NETDEV_UP): %s: "
2255                                         "link is not ready\n",
2256                                         dev->name);
2257                                 break;
2258                         }
2259
2260                         if (idev)
2261                                 idev->if_flags |= IF_READY;
2262                 } else {
2263                         if (!netif_carrier_ok(dev)) {
2264                                 /* device is still not ready. */
2265                                 break;
2266                         }
2267
2268                         if (idev) {
2269                                 if (idev->if_flags & IF_READY) {
2270                                         /* device is already configured. */
2271                                         break;
2272                                 }
2273                                 idev->if_flags |= IF_READY;
2274                         }
2275
2276                         printk(KERN_INFO
2277                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2278                                         "link becomes ready\n",
2279                                         dev->name);
2280
2281                         run_pending = 1;
2282                 }
2283
2284                 switch(dev->type) {
2285 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2286                 case ARPHRD_SIT:
2287                         addrconf_sit_config(dev);
2288                         break;
2289 #endif
2290                 case ARPHRD_TUNNEL6:
2291                         addrconf_ip6_tnl_config(dev);
2292                         break;
2293                 case ARPHRD_LOOPBACK:
2294                         init_loopback(dev);
2295                         break;
2296
2297                 default:
2298                         addrconf_dev_config(dev);
2299                         break;
2300                 };
2301                 if (idev) {
2302                         if (run_pending)
2303                                 addrconf_dad_run(idev);
2304
2305                         /* If the MTU changed during the interface down, when the
2306                            interface up, the changed MTU must be reflected in the
2307                            idev as well as routers.
2308                          */
2309                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2310                                 rt6_mtu_change(dev, dev->mtu);
2311                                 idev->cnf.mtu6 = dev->mtu;
2312                         }
2313                         idev->tstamp = jiffies;
2314                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2315                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2316                            stop IPv6 on this interface.
2317                          */
2318                         if (dev->mtu < IPV6_MIN_MTU)
2319                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2320                 }
2321                 break;
2322
2323         case NETDEV_CHANGEMTU:
2324                 if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2325                         rt6_mtu_change(dev, dev->mtu);
2326                         idev->cnf.mtu6 = dev->mtu;
2327                         break;
2328                 }
2329
2330                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2331
2332         case NETDEV_DOWN:
2333         case NETDEV_UNREGISTER:
2334                 /*
2335                  *      Remove all addresses from this interface.
2336                  */
2337                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2338                 break;
2339
2340         case NETDEV_CHANGENAME:
2341 #ifdef CONFIG_SYSCTL
2342                 if (idev) {
2343                         addrconf_sysctl_unregister(&idev->cnf);
2344                         neigh_sysctl_unregister(idev->nd_parms);
2345                         neigh_sysctl_register(dev, idev->nd_parms,
2346                                               NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2347                                               &ndisc_ifinfo_sysctl_change,
2348                                               NULL);
2349                         addrconf_sysctl_register(idev, &idev->cnf);
2350                 }
2351 #endif
2352                 break;
2353         };
2354
2355         return NOTIFY_OK;
2356 }
2357
2358 /*
2359  *      addrconf module should be notified of a device going up
2360  */
2361 static struct notifier_block ipv6_dev_notf = {
2362         .notifier_call = addrconf_notify,
2363         .priority = 0
2364 };
2365
2366 static int addrconf_ifdown(struct net_device *dev, int how)
2367 {
2368         struct inet6_dev *idev;
2369         struct inet6_ifaddr *ifa, **bifa;
2370         int i;
2371
2372         ASSERT_RTNL();
2373
2374         if (dev == &loopback_dev && how == 1)
2375                 how = 0;
2376
2377         rt6_ifdown(dev);
2378         neigh_ifdown(&nd_tbl, dev);
2379
2380         idev = __in6_dev_get(dev);
2381         if (idev == NULL)
2382                 return -ENODEV;
2383
2384         /* Step 1: remove reference to ipv6 device from parent device.
2385                    Do not dev_put!
2386          */
2387         if (how == 1) {
2388                 idev->dead = 1;
2389
2390                 /* protected by rtnl_lock */
2391                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2392
2393                 /* Step 1.5: remove snmp6 entry */
2394                 snmp6_unregister_dev(idev);
2395
2396         }
2397
2398         /* Step 2: clear hash table */
2399         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2400                 bifa = &inet6_addr_lst[i];
2401
2402                 write_lock_bh(&addrconf_hash_lock);
2403                 while ((ifa = *bifa) != NULL) {
2404                         if (ifa->idev == idev) {
2405                                 *bifa = ifa->lst_next;
2406                                 ifa->lst_next = NULL;
2407                                 addrconf_del_timer(ifa);
2408                                 in6_ifa_put(ifa);
2409                                 continue;
2410                         }
2411                         bifa = &ifa->lst_next;
2412                 }
2413                 write_unlock_bh(&addrconf_hash_lock);
2414         }
2415
2416         write_lock_bh(&idev->lock);
2417
2418         /* Step 3: clear flags for stateless addrconf */
2419         if (how != 1)
2420                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2421
2422         /* Step 4: clear address list */
2423 #ifdef CONFIG_IPV6_PRIVACY
2424         if (how == 1 && del_timer(&idev->regen_timer))
2425                 in6_dev_put(idev);
2426
2427         /* clear tempaddr list */
2428         while ((ifa = idev->tempaddr_list) != NULL) {
2429                 idev->tempaddr_list = ifa->tmp_next;
2430                 ifa->tmp_next = NULL;
2431                 ifa->dead = 1;
2432                 write_unlock_bh(&idev->lock);
2433                 spin_lock_bh(&ifa->lock);
2434
2435                 if (ifa->ifpub) {
2436                         in6_ifa_put(ifa->ifpub);
2437                         ifa->ifpub = NULL;
2438                 }
2439                 spin_unlock_bh(&ifa->lock);
2440                 in6_ifa_put(ifa);
2441                 write_lock_bh(&idev->lock);
2442         }
2443 #endif
2444         while ((ifa = idev->addr_list) != NULL) {
2445                 idev->addr_list = ifa->if_next;
2446                 ifa->if_next = NULL;
2447                 ifa->dead = 1;
2448                 addrconf_del_timer(ifa);
2449                 write_unlock_bh(&idev->lock);
2450
2451                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2452                 in6_ifa_put(ifa);
2453
2454                 write_lock_bh(&idev->lock);
2455         }
2456         write_unlock_bh(&idev->lock);
2457
2458         /* Step 5: Discard multicast list */
2459
2460         if (how == 1)
2461                 ipv6_mc_destroy_dev(idev);
2462         else
2463                 ipv6_mc_down(idev);
2464
2465         /* Step 5: netlink notification of this interface */
2466         idev->tstamp = jiffies;
2467         inet6_ifinfo_notify(RTM_DELLINK, idev);
2468         
2469         /* Shot the device (if unregistered) */
2470
2471         if (how == 1) {
2472 #ifdef CONFIG_SYSCTL
2473                 addrconf_sysctl_unregister(&idev->cnf);
2474                 neigh_sysctl_unregister(idev->nd_parms);
2475 #endif
2476                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2477                 neigh_ifdown(&nd_tbl, dev);
2478                 in6_dev_put(idev);
2479         }
2480         return 0;
2481 }
2482
2483 static void addrconf_rs_timer(unsigned long data)
2484 {
2485         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2486
2487         if (ifp->idev->cnf.forwarding)
2488                 goto out;
2489
2490         if (ifp->idev->if_flags & IF_RA_RCVD) {
2491                 /*
2492                  *      Announcement received after solicitation
2493                  *      was sent
2494                  */
2495                 goto out;
2496         }
2497
2498         spin_lock(&ifp->lock);
2499         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2500                 struct in6_addr all_routers;
2501
2502                 /* The wait after the last probe can be shorter */
2503                 addrconf_mod_timer(ifp, AC_RS,
2504                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2505                                    ifp->idev->cnf.rtr_solicit_delay :
2506                                    ifp->idev->cnf.rtr_solicit_interval);
2507                 spin_unlock(&ifp->lock);
2508
2509                 ipv6_addr_all_routers(&all_routers);
2510
2511                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2512         } else {
2513                 spin_unlock(&ifp->lock);
2514                 /*
2515                  * Note: we do not support deprecated "all on-link"
2516                  * assumption any longer.
2517                  */
2518                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2519                        ifp->idev->dev->name);
2520         }
2521
2522 out:
2523         in6_ifa_put(ifp);
2524 }
2525
2526 /*
2527  *      Duplicate Address Detection
2528  */
2529 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2530 {
2531         unsigned long rand_num;
2532         struct inet6_dev *idev = ifp->idev;
2533
2534         rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2535         ifp->probes = idev->cnf.dad_transmits;
2536         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2537 }
2538
2539 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2540 {
2541         struct inet6_dev *idev = ifp->idev;
2542         struct net_device *dev = idev->dev;
2543
2544         addrconf_join_solict(dev, &ifp->addr);
2545
2546         if (ifp->prefix_len != 128 && (ifp->flags&IFA_F_PERMANENT))
2547                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev, 0,
2548                                         flags);
2549
2550         net_srandom(ifp->addr.s6_addr32[3]);
2551
2552         read_lock_bh(&idev->lock);
2553         if (ifp->dead)
2554                 goto out;
2555         spin_lock_bh(&ifp->lock);
2556
2557         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2558
2559             !(dev->flags&IFF_MULTICAST) ||
2560             !(ifp->flags&IFA_F_TENTATIVE) ||
2561             ifp->flags & IFA_F_NODAD) {
2562                 ifp->flags &= ~IFA_F_TENTATIVE;
2563                 spin_unlock_bh(&ifp->lock);
2564                 read_unlock_bh(&idev->lock);
2565
2566                 addrconf_dad_completed(ifp);
2567                 return;
2568         }
2569
2570         if (!(idev->if_flags & IF_READY)) {
2571                 spin_unlock_bh(&ifp->lock);
2572                 read_unlock_bh(&idev->lock);
2573                 /*
2574                  * If the defice is not ready:
2575                  * - keep it tentative if it is a permanent address.
2576                  * - otherwise, kill it.
2577                  */
2578                 in6_ifa_hold(ifp);
2579                 addrconf_dad_stop(ifp);
2580                 return;
2581         }
2582         addrconf_dad_kick(ifp);
2583         spin_unlock_bh(&ifp->lock);
2584 out:
2585         read_unlock_bh(&idev->lock);
2586 }
2587
2588 static void addrconf_dad_timer(unsigned long data)
2589 {
2590         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2591         struct inet6_dev *idev = ifp->idev;
2592         struct in6_addr unspec;
2593         struct in6_addr mcaddr;
2594
2595         read_lock_bh(&idev->lock);
2596         if (idev->dead) {
2597                 read_unlock_bh(&idev->lock);
2598                 goto out;
2599         }
2600         spin_lock_bh(&ifp->lock);
2601         if (ifp->probes == 0) {
2602                 /*
2603                  * DAD was successful
2604                  */
2605
2606                 ifp->flags &= ~IFA_F_TENTATIVE;
2607                 spin_unlock_bh(&ifp->lock);
2608                 read_unlock_bh(&idev->lock);
2609
2610                 addrconf_dad_completed(ifp);
2611
2612                 goto out;
2613         }
2614
2615         ifp->probes--;
2616         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2617         spin_unlock_bh(&ifp->lock);
2618         read_unlock_bh(&idev->lock);
2619
2620         /* send a neighbour solicitation for our addr */
2621         memset(&unspec, 0, sizeof(unspec));
2622         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2623         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2624 out:
2625         in6_ifa_put(ifp);
2626 }
2627
2628 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2629 {
2630         struct net_device *     dev = ifp->idev->dev;
2631
2632         /*
2633          *      Configure the address for reception. Now it is valid.
2634          */
2635
2636         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2637
2638         /* If added prefix is link local and forwarding is off,
2639            start sending router solicitations.
2640          */
2641
2642         if (ifp->idev->cnf.forwarding == 0 &&
2643             ifp->idev->cnf.rtr_solicits > 0 &&
2644             (dev->flags&IFF_LOOPBACK) == 0 &&
2645             (dev->flags & IFF_MULTICAST) &&
2646             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2647                 struct in6_addr all_routers;
2648
2649                 ipv6_addr_all_routers(&all_routers);
2650
2651                 /*
2652                  *      If a host as already performed a random delay
2653                  *      [...] as part of DAD [...] there is no need
2654                  *      to delay again before sending the first RS
2655                  */
2656                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2657
2658                 spin_lock_bh(&ifp->lock);
2659                 ifp->probes = 1;
2660                 ifp->idev->if_flags |= IF_RS_SENT;
2661                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2662                 spin_unlock_bh(&ifp->lock);
2663         }
2664 }
2665
2666 static void addrconf_dad_run(struct inet6_dev *idev) {
2667         struct inet6_ifaddr *ifp;
2668
2669         read_lock_bh(&idev->lock);
2670         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2671                 spin_lock_bh(&ifp->lock);
2672                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2673                         spin_unlock_bh(&ifp->lock);
2674                         continue;
2675                 }
2676                 spin_unlock_bh(&ifp->lock);
2677                 addrconf_dad_kick(ifp);
2678         }
2679         read_unlock_bh(&idev->lock);
2680 }
2681
2682 #ifdef CONFIG_PROC_FS
2683 struct if6_iter_state {
2684         int bucket;
2685 };
2686
2687 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2688 {
2689         struct inet6_ifaddr *ifa = NULL;
2690         struct if6_iter_state *state = seq->private;
2691
2692         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2693                 ifa = inet6_addr_lst[state->bucket];
2694                 if (ifa)
2695                         break;
2696         }
2697         return ifa;
2698 }
2699
2700 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2701 {
2702         struct if6_iter_state *state = seq->private;
2703
2704         ifa = ifa->lst_next;
2705 try_again:
2706         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2707                 ifa = inet6_addr_lst[state->bucket];
2708                 goto try_again;
2709         }
2710         return ifa;
2711 }
2712
2713 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2714 {
2715         struct inet6_ifaddr *ifa = if6_get_first(seq);
2716
2717         if (ifa)
2718                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2719                         --pos;
2720         return pos ? NULL : ifa;
2721 }
2722
2723 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2724 {
2725         read_lock_bh(&addrconf_hash_lock);
2726         return if6_get_idx(seq, *pos);
2727 }
2728
2729 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2730 {
2731         struct inet6_ifaddr *ifa;
2732
2733         ifa = if6_get_next(seq, v);
2734         ++*pos;
2735         return ifa;
2736 }
2737
2738 static void if6_seq_stop(struct seq_file *seq, void *v)
2739 {
2740         read_unlock_bh(&addrconf_hash_lock);
2741 }
2742
2743 static int if6_seq_show(struct seq_file *seq, void *v)
2744 {
2745         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2746
2747         /* no ipv6 inside a vserver for now */
2748         if (nx_check(0, VS_ADMIN|VS_WATCH))
2749                 seq_printf(seq,
2750                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2751                    NIP6(ifp->addr),
2752                    ifp->idev->dev->ifindex,
2753                    ifp->prefix_len,
2754                    ifp->scope,
2755                    ifp->flags,
2756                    ifp->idev->dev->name);
2757         return 0;
2758 }
2759
2760 static struct seq_operations if6_seq_ops = {
2761         .start  = if6_seq_start,
2762         .next   = if6_seq_next,
2763         .show   = if6_seq_show,
2764         .stop   = if6_seq_stop,
2765 };
2766
2767 static int if6_seq_open(struct inode *inode, struct file *file)
2768 {
2769         struct seq_file *seq;
2770         int rc = -ENOMEM;
2771         struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2772
2773         if (!s)
2774                 goto out;
2775
2776         rc = seq_open(file, &if6_seq_ops);
2777         if (rc)
2778                 goto out_kfree;
2779
2780         seq = file->private_data;
2781         seq->private = s;
2782 out:
2783         return rc;
2784 out_kfree:
2785         kfree(s);
2786         goto out;
2787 }
2788
2789 static struct file_operations if6_fops = {
2790         .owner          = THIS_MODULE,
2791         .open           = if6_seq_open,
2792         .read           = seq_read,
2793         .llseek         = seq_lseek,
2794         .release        = seq_release_private,
2795 };
2796
2797 int __init if6_proc_init(void)
2798 {
2799         if (!proc_net_fops_create("if_inet6", S_IRUGO, &if6_fops))
2800                 return -ENOMEM;
2801         return 0;
2802 }
2803
2804 void if6_proc_exit(void)
2805 {
2806         proc_net_remove("if_inet6");
2807 }
2808 #endif  /* CONFIG_PROC_FS */
2809
2810 #ifdef CONFIG_IPV6_MIP6
2811 /* Check if address is a home address configured on any interface. */
2812 int ipv6_chk_home_addr(struct in6_addr *addr)
2813 {
2814         int ret = 0;
2815         struct inet6_ifaddr * ifp;
2816         u8 hash = ipv6_addr_hash(addr);
2817         read_lock_bh(&addrconf_hash_lock);
2818         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2819                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2820                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2821                         ret = 1;
2822                         break;
2823                 }
2824         }
2825         read_unlock_bh(&addrconf_hash_lock);
2826         return ret;
2827 }
2828 #endif
2829
2830 /*
2831  *      Periodic address status verification
2832  */
2833
2834 static void addrconf_verify(unsigned long foo)
2835 {
2836         struct inet6_ifaddr *ifp;
2837         unsigned long now, next;
2838         int i;
2839
2840         spin_lock_bh(&addrconf_verify_lock);
2841         now = jiffies;
2842         next = now + ADDR_CHECK_FREQUENCY;
2843
2844         del_timer(&addr_chk_timer);
2845
2846         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2847
2848 restart:
2849                 read_lock(&addrconf_hash_lock);
2850                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2851                         unsigned long age;
2852 #ifdef CONFIG_IPV6_PRIVACY
2853                         unsigned long regen_advance;
2854 #endif
2855
2856                         if (ifp->flags & IFA_F_PERMANENT)
2857                                 continue;
2858
2859                         spin_lock(&ifp->lock);
2860                         age = (now - ifp->tstamp) / HZ;
2861
2862 #ifdef CONFIG_IPV6_PRIVACY
2863                         regen_advance = ifp->idev->cnf.regen_max_retry * 
2864                                         ifp->idev->cnf.dad_transmits * 
2865                                         ifp->idev->nd_parms->retrans_time / HZ;
2866 #endif
2867
2868                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2869                             age >= ifp->valid_lft) {
2870                                 spin_unlock(&ifp->lock);
2871                                 in6_ifa_hold(ifp);
2872                                 read_unlock(&addrconf_hash_lock);
2873                                 ipv6_del_addr(ifp);
2874                                 goto restart;
2875                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2876                                 spin_unlock(&ifp->lock);
2877                                 continue;
2878                         } else if (age >= ifp->prefered_lft) {
2879                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2880                                 int deprecate = 0;
2881
2882                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2883                                         deprecate = 1;
2884                                         ifp->flags |= IFA_F_DEPRECATED;
2885                                 }
2886
2887                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2888                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2889
2890                                 spin_unlock(&ifp->lock);
2891
2892                                 if (deprecate) {
2893                                         in6_ifa_hold(ifp);
2894                                         read_unlock(&addrconf_hash_lock);
2895
2896                                         ipv6_ifa_notify(0, ifp);
2897                                         in6_ifa_put(ifp);
2898                                         goto restart;
2899                                 }
2900 #ifdef CONFIG_IPV6_PRIVACY
2901                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2902                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2903                                 if (age >= ifp->prefered_lft - regen_advance) {
2904                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2905                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2906                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2907                                         if (!ifp->regen_count && ifpub) {
2908                                                 ifp->regen_count++;
2909                                                 in6_ifa_hold(ifp);
2910                                                 in6_ifa_hold(ifpub);
2911                                                 spin_unlock(&ifp->lock);
2912                                                 read_unlock(&addrconf_hash_lock);
2913                                                 spin_lock(&ifpub->lock);
2914                                                 ifpub->regen_count = 0;
2915                                                 spin_unlock(&ifpub->lock);
2916                                                 ipv6_create_tempaddr(ifpub, ifp);
2917                                                 in6_ifa_put(ifpub);
2918                                                 in6_ifa_put(ifp);
2919                                                 goto restart;
2920                                         }
2921                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2922                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2923                                 spin_unlock(&ifp->lock);
2924 #endif
2925                         } else {
2926                                 /* ifp->prefered_lft <= ifp->valid_lft */
2927                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2928                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
2929                                 spin_unlock(&ifp->lock);
2930                         }
2931                 }
2932                 read_unlock(&addrconf_hash_lock);
2933         }
2934
2935         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2936         add_timer(&addr_chk_timer);
2937         spin_unlock_bh(&addrconf_verify_lock);
2938 }
2939
2940 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2941 {
2942         struct in6_addr *pfx = NULL;
2943
2944         if (addr)
2945                 pfx = nla_data(addr);
2946
2947         if (local) {
2948                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2949                         pfx = NULL;
2950                 else
2951                         pfx = nla_data(local);
2952         }
2953
2954         return pfx;
2955 }
2956
2957 static struct nla_policy ifa_ipv6_policy[IFA_MAX+1] __read_mostly = {
2958         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
2959         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
2960         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
2961 };
2962
2963 static int
2964 inet6_rtm_deladdr(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         int err;
2970
2971         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2972         if (err < 0)
2973                 return err;
2974
2975         ifm = nlmsg_data(nlh);
2976         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2977         if (pfx == NULL)
2978                 return -EINVAL;
2979
2980         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
2981 }
2982
2983 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
2984                              u32 prefered_lft, u32 valid_lft)
2985 {
2986         if (!valid_lft || (prefered_lft > valid_lft))
2987                 return -EINVAL;
2988
2989         if (valid_lft == INFINITY_LIFE_TIME)
2990                 ifa_flags |= IFA_F_PERMANENT;
2991         else if (valid_lft >= 0x7FFFFFFF/HZ)
2992                 valid_lft = 0x7FFFFFFF/HZ;
2993
2994         if (prefered_lft == 0)
2995                 ifa_flags |= IFA_F_DEPRECATED;
2996         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2997                  (prefered_lft != INFINITY_LIFE_TIME))
2998                 prefered_lft = 0x7FFFFFFF/HZ;
2999
3000         spin_lock_bh(&ifp->lock);
3001         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3002         ifp->tstamp = jiffies;
3003         ifp->valid_lft = valid_lft;
3004         ifp->prefered_lft = prefered_lft;
3005
3006         spin_unlock_bh(&ifp->lock);
3007         if (!(ifp->flags&IFA_F_TENTATIVE))
3008                 ipv6_ifa_notify(0, ifp);
3009
3010         addrconf_verify(0);
3011
3012         return 0;
3013 }
3014
3015 static int
3016 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3017 {
3018         struct ifaddrmsg *ifm;
3019         struct nlattr *tb[IFA_MAX+1];
3020         struct in6_addr *pfx;
3021         struct inet6_ifaddr *ifa;
3022         struct net_device *dev;
3023         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3024         u8 ifa_flags;
3025         int err;
3026
3027         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3028         if (err < 0)
3029                 return err;
3030
3031         ifm = nlmsg_data(nlh);
3032         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3033         if (pfx == NULL)
3034                 return -EINVAL;
3035
3036         if (tb[IFA_CACHEINFO]) {
3037                 struct ifa_cacheinfo *ci;
3038
3039                 ci = nla_data(tb[IFA_CACHEINFO]);
3040                 valid_lft = ci->ifa_valid;
3041                 preferred_lft = ci->ifa_prefered;
3042         } else {
3043                 preferred_lft = INFINITY_LIFE_TIME;
3044                 valid_lft = INFINITY_LIFE_TIME;
3045         }
3046
3047         dev =  __dev_get_by_index(ifm->ifa_index);
3048         if (dev == NULL)
3049                 return -ENODEV;
3050
3051         /* We ignore other flags so far. */
3052         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3053
3054         ifa = ipv6_get_ifaddr(pfx, dev, 1);
3055         if (ifa == NULL) {
3056                 /*
3057                  * It would be best to check for !NLM_F_CREATE here but
3058                  * userspace alreay relies on not having to provide this.
3059                  */
3060                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3061                                       ifa_flags, preferred_lft, valid_lft);
3062         }
3063
3064         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3065             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3066                 err = -EEXIST;
3067         else
3068                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3069
3070         in6_ifa_put(ifa);
3071
3072         return err;
3073 }
3074
3075 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3076                           u8 scope, int ifindex)
3077 {
3078         struct ifaddrmsg *ifm;
3079
3080         ifm = nlmsg_data(nlh);
3081         ifm->ifa_family = AF_INET6;
3082         ifm->ifa_prefixlen = prefixlen;
3083         ifm->ifa_flags = flags;
3084         ifm->ifa_scope = scope;
3085         ifm->ifa_index = ifindex;
3086 }
3087
3088 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3089                          unsigned long tstamp, u32 preferred, u32 valid)
3090 {
3091         struct ifa_cacheinfo ci;
3092
3093         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3094                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3095         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3096                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3097         ci.ifa_prefered = preferred;
3098         ci.ifa_valid = valid;
3099
3100         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3101 }
3102
3103 static inline int rt_scope(int ifa_scope)
3104 {
3105         if (ifa_scope & IFA_HOST)
3106                 return RT_SCOPE_HOST;
3107         else if (ifa_scope & IFA_LINK)
3108                 return RT_SCOPE_LINK;
3109         else if (ifa_scope & IFA_SITE)
3110                 return RT_SCOPE_SITE;
3111         else
3112                 return RT_SCOPE_UNIVERSE;
3113 }
3114
3115 static inline int inet6_ifaddr_msgsize(void)
3116 {
3117         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3118                + nla_total_size(16) /* IFA_ADDRESS */
3119                + nla_total_size(sizeof(struct ifa_cacheinfo));
3120 }
3121
3122 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3123                              u32 pid, u32 seq, int event, unsigned int flags)
3124 {
3125         struct nlmsghdr  *nlh;
3126         u32 preferred, valid;
3127
3128         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3129         if (nlh == NULL)
3130                 return -ENOBUFS;
3131
3132         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3133                       ifa->idev->dev->ifindex);
3134
3135         if (!(ifa->flags&IFA_F_PERMANENT)) {
3136                 preferred = ifa->prefered_lft;
3137                 valid = ifa->valid_lft;
3138                 if (preferred != INFINITY_LIFE_TIME) {
3139                         long tval = (jiffies - ifa->tstamp)/HZ;
3140                         preferred -= tval;
3141                         if (valid != INFINITY_LIFE_TIME)
3142                                 valid -= tval;
3143                 }
3144         } else {
3145                 preferred = INFINITY_LIFE_TIME;
3146                 valid = INFINITY_LIFE_TIME;
3147         }
3148
3149         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3150             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3151                 return nlmsg_cancel(skb, nlh);
3152
3153         return nlmsg_end(skb, nlh);
3154 }
3155
3156 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3157                                 u32 pid, u32 seq, int event, u16 flags)
3158 {
3159         struct nlmsghdr  *nlh;
3160         u8 scope = RT_SCOPE_UNIVERSE;
3161         int ifindex = ifmca->idev->dev->ifindex;
3162
3163         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3164                 scope = RT_SCOPE_SITE;
3165
3166         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3167         if (nlh == NULL)
3168                 return -ENOBUFS;
3169
3170         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3171         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3172             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3173                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3174                 return nlmsg_cancel(skb, nlh);
3175
3176         return nlmsg_end(skb, nlh);
3177 }
3178
3179 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3180                                 u32 pid, u32 seq, int event, unsigned int flags)
3181 {
3182         struct nlmsghdr  *nlh;
3183         u8 scope = RT_SCOPE_UNIVERSE;
3184         int ifindex = ifaca->aca_idev->dev->ifindex;
3185
3186         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3187                 scope = RT_SCOPE_SITE;
3188
3189         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3190         if (nlh == NULL)
3191                 return -ENOBUFS;
3192
3193         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3194         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3195             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3196                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3197                 return nlmsg_cancel(skb, nlh);
3198
3199         return nlmsg_end(skb, nlh);
3200 }
3201
3202 enum addr_type_t
3203 {
3204         UNICAST_ADDR,
3205         MULTICAST_ADDR,
3206         ANYCAST_ADDR,
3207 };
3208
3209 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3210                            enum addr_type_t type)
3211 {
3212         int idx, ip_idx;
3213         int s_idx, s_ip_idx;
3214         int err = 1;
3215         struct net_device *dev;
3216         struct inet6_dev *idev = NULL;
3217         struct inet6_ifaddr *ifa;
3218         struct ifmcaddr6 *ifmca;
3219         struct ifacaddr6 *ifaca;
3220
3221         /* no ipv6 inside a vserver for now */
3222         if (skb->sk && skb->sk->sk_vx_info)
3223                 return skb->len;
3224
3225         s_idx = cb->args[0];
3226         s_ip_idx = ip_idx = cb->args[1];
3227         read_lock(&dev_base_lock);
3228         
3229         for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
3230                 if (idx < s_idx)
3231                         continue;
3232                 if (idx > s_idx)
3233                         s_ip_idx = 0;
3234                 ip_idx = 0;
3235                 if ((idev = in6_dev_get(dev)) == NULL)
3236                         continue;
3237                 read_lock_bh(&idev->lock);
3238                 switch (type) {
3239                 case UNICAST_ADDR:
3240                         /* unicast address incl. temp addr */
3241                         for (ifa = idev->addr_list; ifa;
3242                              ifa = ifa->if_next, ip_idx++) {
3243                                 if (ip_idx < s_ip_idx)
3244                                         continue;
3245                                 if ((err = inet6_fill_ifaddr(skb, ifa, 
3246                                     NETLINK_CB(cb->skb).pid, 
3247                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3248                                     NLM_F_MULTI)) <= 0)
3249                                         goto done;
3250                         }
3251                         break;
3252                 case MULTICAST_ADDR:
3253                         /* multicast address */
3254                         for (ifmca = idev->mc_list; ifmca; 
3255                              ifmca = ifmca->next, ip_idx++) {
3256                                 if (ip_idx < s_ip_idx)
3257                                         continue;
3258                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca, 
3259                                     NETLINK_CB(cb->skb).pid, 
3260                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3261                                     NLM_F_MULTI)) <= 0)
3262                                         goto done;
3263                         }
3264                         break;
3265                 case ANYCAST_ADDR:
3266                         /* anycast address */
3267                         for (ifaca = idev->ac_list; ifaca;
3268                              ifaca = ifaca->aca_next, ip_idx++) {
3269                                 if (ip_idx < s_ip_idx)
3270                                         continue;
3271                                 if ((err = inet6_fill_ifacaddr(skb, ifaca, 
3272                                     NETLINK_CB(cb->skb).pid, 
3273                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3274                                     NLM_F_MULTI)) <= 0) 
3275                                         goto done;
3276                         }
3277                         break;
3278                 default:
3279                         break;
3280                 }
3281                 read_unlock_bh(&idev->lock);
3282                 in6_dev_put(idev);
3283         }
3284 done:
3285         if (err <= 0) {
3286                 read_unlock_bh(&idev->lock);
3287                 in6_dev_put(idev);
3288         }
3289         read_unlock(&dev_base_lock);
3290         cb->args[0] = idx;
3291         cb->args[1] = ip_idx;
3292         return skb->len;
3293 }
3294
3295 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3296 {
3297         enum addr_type_t type = UNICAST_ADDR;
3298         return inet6_dump_addr(skb, cb, type);
3299 }
3300
3301 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3302 {
3303         enum addr_type_t type = MULTICAST_ADDR;
3304         return inet6_dump_addr(skb, cb, type);
3305 }
3306
3307
3308 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3309 {
3310         enum addr_type_t type = ANYCAST_ADDR;
3311         return inet6_dump_addr(skb, cb, type);
3312 }
3313
3314 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3315                              void *arg)
3316 {
3317         struct ifaddrmsg *ifm;
3318         struct nlattr *tb[IFA_MAX+1];
3319         struct in6_addr *addr = NULL;
3320         struct net_device *dev = NULL;
3321         struct inet6_ifaddr *ifa;
3322         struct sk_buff *skb;
3323         int err;
3324
3325         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3326         if (err < 0)
3327                 goto errout;
3328
3329         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3330         if (addr == NULL) {
3331                 err = -EINVAL;
3332                 goto errout;
3333         }
3334
3335         ifm = nlmsg_data(nlh);
3336         if (ifm->ifa_index)
3337                 dev = __dev_get_by_index(ifm->ifa_index);
3338
3339         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3340                 err = -EADDRNOTAVAIL;
3341                 goto errout;
3342         }
3343
3344         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3345                 err = -ENOBUFS;
3346                 goto errout_ifa;
3347         }
3348
3349         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3350                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3351         /* failure implies BUG in inet6_ifaddr_msgsize() */
3352         BUG_ON(err < 0);
3353
3354         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3355 errout_ifa:
3356         in6_ifa_put(ifa);
3357 errout:
3358         return err;
3359 }
3360
3361 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3362 {
3363         struct sk_buff *skb;
3364         int err = -ENOBUFS;
3365
3366         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3367         if (skb == NULL)
3368                 goto errout;
3369
3370         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3371         /* failure implies BUG in inet6_ifaddr_msgsize() */
3372         BUG_ON(err < 0);
3373
3374         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3375 errout:
3376         if (err < 0)
3377                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3378 }
3379
3380 static void inline ipv6_store_devconf(struct ipv6_devconf *cnf,
3381                                 __s32 *array, int bytes)
3382 {
3383         BUG_ON(bytes < (DEVCONF_MAX * 4));
3384
3385         memset(array, 0, bytes);
3386         array[DEVCONF_FORWARDING] = cnf->forwarding;
3387         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3388         array[DEVCONF_MTU6] = cnf->mtu6;
3389         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3390         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3391         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3392         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3393         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3394         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3395         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3396         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3397 #ifdef CONFIG_IPV6_PRIVACY
3398         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3399         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3400         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3401         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3402         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3403 #endif
3404         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3405         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3406         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3407 #ifdef CONFIG_IPV6_ROUTER_PREF
3408         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3409         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3410 #ifdef CONFIG_IPV6_ROUTE_INFO
3411         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3412 #endif
3413 #endif
3414         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3415         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3416 }
3417
3418 static inline size_t inet6_if_nlmsg_size(void)
3419 {
3420         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3421                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3422                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3423                + nla_total_size(4) /* IFLA_MTU */
3424                + nla_total_size(4) /* IFLA_LINK */
3425                + nla_total_size( /* IFLA_PROTINFO */
3426                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3427                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3428                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3429                  );
3430 }
3431
3432 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, 
3433                              u32 pid, u32 seq, int event, unsigned int flags)
3434 {
3435         struct net_device *dev = idev->dev;
3436         struct nlattr *conf;
3437         struct ifinfomsg *hdr;
3438         struct nlmsghdr *nlh;
3439         void *protoinfo;
3440         struct ifla_cacheinfo ci;
3441
3442         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3443         if (nlh == NULL)
3444                 return -ENOBUFS;
3445
3446         hdr = nlmsg_data(nlh);
3447         hdr->ifi_family = AF_INET6;
3448         hdr->__ifi_pad = 0;
3449         hdr->ifi_type = dev->type;
3450         hdr->ifi_index = dev->ifindex;
3451         hdr->ifi_flags = dev_get_flags(dev);
3452         hdr->ifi_change = 0;
3453
3454         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3455
3456         if (dev->addr_len)
3457                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3458
3459         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3460         if (dev->ifindex != dev->iflink)
3461                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3462
3463         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3464         if (protoinfo == NULL)
3465                 goto nla_put_failure;
3466
3467         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3468
3469         ci.max_reasm_len = IPV6_MAXPLEN;
3470         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3471                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3472         ci.reachable_time = idev->nd_parms->reachable_time;
3473         ci.retrans_time = idev->nd_parms->retrans_time;
3474         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3475
3476         conf = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3477         if (conf == NULL)
3478                 goto nla_put_failure;
3479         ipv6_store_devconf(&idev->cnf, nla_data(conf), nla_len(conf));
3480
3481         /* XXX - Statistics/MC not implemented */
3482
3483         nla_nest_end(skb, protoinfo);
3484         return nlmsg_end(skb, nlh);
3485
3486 nla_put_failure:
3487         return nlmsg_cancel(skb, nlh);
3488 }
3489
3490 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3491 {
3492         int idx, err;
3493         int s_idx = cb->args[0];
3494         struct net_device *dev;
3495         struct inet6_dev *idev;
3496
3497         /* no ipv6 inside a vserver for now */
3498         if (skb->sk && skb->sk->sk_vx_info)
3499                 return skb->len;
3500
3501         read_lock(&dev_base_lock);
3502         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
3503                 if (idx < s_idx)
3504                         continue;
3505                 if ((idev = in6_dev_get(dev)) == NULL)
3506                         continue;
3507                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid, 
3508                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3509                 in6_dev_put(idev);
3510                 if (err <= 0)
3511                         break;
3512         }
3513         read_unlock(&dev_base_lock);
3514         cb->args[0] = idx;
3515
3516         return skb->len;
3517 }
3518
3519 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3520 {
3521         struct sk_buff *skb;
3522         int err = -ENOBUFS;
3523         
3524         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3525         if (skb == NULL)
3526                 goto errout;
3527
3528         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3529         /* failure implies BUG in inet6_if_nlmsg_size() */
3530         BUG_ON(err < 0);
3531
3532         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3533 errout:
3534         if (err < 0)
3535                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3536 }
3537
3538 static inline size_t inet6_prefix_nlmsg_size(void)
3539 {
3540         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3541                + nla_total_size(sizeof(struct in6_addr))
3542                + nla_total_size(sizeof(struct prefix_cacheinfo));
3543 }
3544
3545 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3546                              struct prefix_info *pinfo, u32 pid, u32 seq,
3547                              int event, unsigned int flags)
3548 {
3549         struct prefixmsg *pmsg;
3550         struct nlmsghdr *nlh;
3551         struct prefix_cacheinfo ci;
3552
3553         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3554         if (nlh == NULL)
3555                 return -ENOBUFS;
3556
3557         pmsg = nlmsg_data(nlh);
3558         pmsg->prefix_family = AF_INET6;
3559         pmsg->prefix_pad1 = 0;
3560         pmsg->prefix_pad2 = 0;
3561         pmsg->prefix_ifindex = idev->dev->ifindex;
3562         pmsg->prefix_len = pinfo->prefix_len;
3563         pmsg->prefix_type = pinfo->type;
3564         pmsg->prefix_pad3 = 0;
3565         pmsg->prefix_flags = 0;
3566         if (pinfo->onlink)
3567                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3568         if (pinfo->autoconf)
3569                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3570
3571         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3572
3573         ci.preferred_time = ntohl(pinfo->prefered);
3574         ci.valid_time = ntohl(pinfo->valid);
3575         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3576
3577         return nlmsg_end(skb, nlh);
3578
3579 nla_put_failure:
3580         return nlmsg_cancel(skb, nlh);
3581 }
3582
3583 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 
3584                          struct prefix_info *pinfo)
3585 {
3586         struct sk_buff *skb;
3587         int err = -ENOBUFS;
3588
3589         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3590         if (skb == NULL)
3591                 goto errout;
3592
3593         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3594         /* failure implies BUG in inet6_prefix_nlmsg_size() */
3595         BUG_ON(err < 0);
3596
3597         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3598 errout:
3599         if (err < 0)
3600                 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3601 }
3602
3603 static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
3604         [RTM_GETLINK - RTM_BASE] = { .dumpit    = inet6_dump_ifinfo, },
3605         [RTM_NEWADDR - RTM_BASE] = { .doit      = inet6_rtm_newaddr, },
3606         [RTM_DELADDR - RTM_BASE] = { .doit      = inet6_rtm_deladdr, },
3607         [RTM_GETADDR - RTM_BASE] = { .doit      = inet6_rtm_getaddr,
3608                                      .dumpit    = inet6_dump_ifaddr, },
3609         [RTM_GETMULTICAST - RTM_BASE] = { .dumpit = inet6_dump_ifmcaddr, },
3610         [RTM_GETANYCAST - RTM_BASE] = { .dumpit = inet6_dump_ifacaddr, },
3611         [RTM_NEWROUTE - RTM_BASE] = { .doit     = inet6_rtm_newroute, },
3612         [RTM_DELROUTE - RTM_BASE] = { .doit     = inet6_rtm_delroute, },
3613         [RTM_GETROUTE - RTM_BASE] = { .doit     = inet6_rtm_getroute,
3614                                       .dumpit   = inet6_dump_fib, },
3615 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3616         [RTM_GETRULE  - RTM_BASE] = { .dumpit   = fib6_rules_dump,   },
3617 #endif
3618 };
3619
3620 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3621 {
3622         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3623
3624         switch (event) {
3625         case RTM_NEWADDR:
3626                 ip6_ins_rt(ifp->rt);
3627                 if (ifp->idev->cnf.forwarding)
3628                         addrconf_join_anycast(ifp);
3629                 break;
3630         case RTM_DELADDR:
3631                 if (ifp->idev->cnf.forwarding)
3632                         addrconf_leave_anycast(ifp);
3633                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3634                 dst_hold(&ifp->rt->u.dst);
3635                 if (ip6_del_rt(ifp->rt))
3636                         dst_free(&ifp->rt->u.dst);
3637                 break;
3638         }
3639 }
3640
3641 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3642 {
3643         rcu_read_lock_bh();
3644         if (likely(ifp->idev->dead == 0))
3645                 __ipv6_ifa_notify(event, ifp);
3646         rcu_read_unlock_bh();
3647 }
3648
3649 #ifdef CONFIG_SYSCTL
3650
3651 static
3652 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3653                            void __user *buffer, size_t *lenp, loff_t *ppos)
3654 {
3655         int *valp = ctl->data;
3656         int val = *valp;
3657         int ret;
3658
3659         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3660
3661         if (write && valp != &ipv6_devconf_dflt.forwarding) {
3662                 if (valp != &ipv6_devconf.forwarding) {
3663                         if ((!*valp) ^ (!val)) {
3664                                 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3665                                 if (idev == NULL)
3666                                         return ret;
3667                                 dev_forward_change(idev);
3668                         }
3669                 } else {
3670                         ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3671                         addrconf_forward_change();
3672                 }
3673                 if (*valp)
3674                         rt6_purge_dflt_routers();
3675         }
3676
3677         return ret;
3678 }
3679
3680 static int addrconf_sysctl_forward_strategy(ctl_table *table, 
3681                                             int __user *name, int nlen,
3682                                             void __user *oldval,
3683                                             size_t __user *oldlenp,
3684                                             void __user *newval, size_t newlen)
3685 {
3686         int *valp = table->data;
3687         int new;
3688
3689         if (!newval || !newlen)
3690                 return 0;
3691         if (newlen != sizeof(int))
3692                 return -EINVAL;
3693         if (get_user(new, (int __user *)newval))
3694                 return -EFAULT;
3695         if (new == *valp)
3696                 return 0;
3697         if (oldval && oldlenp) {
3698                 size_t len;
3699                 if (get_user(len, oldlenp))
3700                         return -EFAULT;
3701                 if (len) {
3702                         if (len > table->maxlen)
3703                                 len = table->maxlen;
3704                         if (copy_to_user(oldval, valp, len))
3705                                 return -EFAULT;
3706                         if (put_user(len, oldlenp))
3707                                 return -EFAULT;
3708                 }
3709         }
3710
3711         if (valp != &ipv6_devconf_dflt.forwarding) {
3712                 if (valp != &ipv6_devconf.forwarding) {
3713                         struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3714                         int changed;
3715                         if (unlikely(idev == NULL))
3716                                 return -ENODEV;
3717                         changed = (!*valp) ^ (!new);
3718                         *valp = new;
3719                         if (changed)
3720                                 dev_forward_change(idev);
3721                 } else {
3722                         *valp = new;
3723                         addrconf_forward_change();
3724                 }
3725
3726                 if (*valp)
3727                         rt6_purge_dflt_routers();
3728         } else
3729                 *valp = new;
3730
3731         return 1;
3732 }
3733
3734 static struct addrconf_sysctl_table
3735 {
3736         struct ctl_table_header *sysctl_header;
3737         ctl_table addrconf_vars[__NET_IPV6_MAX];
3738         ctl_table addrconf_dev[2];
3739         ctl_table addrconf_conf_dir[2];
3740         ctl_table addrconf_proto_dir[2];
3741         ctl_table addrconf_root_dir[2];
3742 } addrconf_sysctl __read_mostly = {
3743         .sysctl_header = NULL,
3744         .addrconf_vars = {
3745                 {
3746                         .ctl_name       =       NET_IPV6_FORWARDING,
3747                         .procname       =       "forwarding",
3748                         .data           =       &ipv6_devconf.forwarding,
3749                         .maxlen         =       sizeof(int),
3750                         .mode           =       0644,
3751                         .proc_handler   =       &addrconf_sysctl_forward,
3752                         .strategy       =       &addrconf_sysctl_forward_strategy,
3753                 },
3754                 {
3755                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3756                         .procname       =       "hop_limit",
3757                         .data           =       &ipv6_devconf.hop_limit,
3758                         .maxlen         =       sizeof(int),
3759                         .mode           =       0644,
3760                         .proc_handler   =       proc_dointvec,
3761                 },
3762                 {
3763                         .ctl_name       =       NET_IPV6_MTU,
3764                         .procname       =       "mtu",
3765                         .data           =       &ipv6_devconf.mtu6,
3766                         .maxlen         =       sizeof(int),
3767                         .mode           =       0644,
3768                         .proc_handler   =       &proc_dointvec,
3769                 },
3770                 {
3771                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3772                         .procname       =       "accept_ra",
3773                         .data           =       &ipv6_devconf.accept_ra,
3774                         .maxlen         =       sizeof(int),
3775                         .mode           =       0644,
3776                         .proc_handler   =       &proc_dointvec,
3777                 },
3778                 {
3779                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3780                         .procname       =       "accept_redirects",
3781                         .data           =       &ipv6_devconf.accept_redirects,
3782                         .maxlen         =       sizeof(int),
3783                         .mode           =       0644,
3784                         .proc_handler   =       &proc_dointvec,
3785                 },
3786                 {
3787                         .ctl_name       =       NET_IPV6_AUTOCONF,
3788                         .procname       =       "autoconf",
3789                         .data           =       &ipv6_devconf.autoconf,
3790                         .maxlen         =       sizeof(int),
3791                         .mode           =       0644,
3792                         .proc_handler   =       &proc_dointvec,
3793                 },
3794                 {
3795                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3796                         .procname       =       "dad_transmits",
3797                         .data           =       &ipv6_devconf.dad_transmits,
3798                         .maxlen         =       sizeof(int),
3799                         .mode           =       0644,
3800                         .proc_handler   =       &proc_dointvec,
3801                 },
3802                 {
3803                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3804                         .procname       =       "router_solicitations",
3805                         .data           =       &ipv6_devconf.rtr_solicits,
3806                         .maxlen         =       sizeof(int),
3807                         .mode           =       0644,
3808                         .proc_handler   =       &proc_dointvec,
3809                 },
3810                 {
3811                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3812                         .procname       =       "router_solicitation_interval",
3813                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3814                         .maxlen         =       sizeof(int),
3815                         .mode           =       0644,
3816                         .proc_handler   =       &proc_dointvec_jiffies,
3817                         .strategy       =       &sysctl_jiffies,
3818                 },
3819                 {
3820                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3821                         .procname       =       "router_solicitation_delay",
3822                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3823                         .maxlen         =       sizeof(int),
3824                         .mode           =       0644,
3825                         .proc_handler   =       &proc_dointvec_jiffies,
3826                         .strategy       =       &sysctl_jiffies,
3827                 },
3828                 {
3829                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3830                         .procname       =       "force_mld_version",
3831                         .data           =       &ipv6_devconf.force_mld_version,
3832                         .maxlen         =       sizeof(int),
3833                         .mode           =       0644,
3834                         .proc_handler   =       &proc_dointvec,
3835                 },
3836 #ifdef CONFIG_IPV6_PRIVACY
3837                 {
3838                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3839                         .procname       =       "use_tempaddr",
3840                         .data           =       &ipv6_devconf.use_tempaddr,
3841                         .maxlen         =       sizeof(int),
3842                         .mode           =       0644,
3843                         .proc_handler   =       &proc_dointvec,
3844                 },
3845                 {
3846                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3847                         .procname       =       "temp_valid_lft",
3848                         .data           =       &ipv6_devconf.temp_valid_lft,
3849                         .maxlen         =       sizeof(int),
3850                         .mode           =       0644,
3851                         .proc_handler   =       &proc_dointvec,
3852                 },
3853                 {
3854                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3855                         .procname       =       "temp_prefered_lft",
3856                         .data           =       &ipv6_devconf.temp_prefered_lft,
3857                         .maxlen         =       sizeof(int),
3858                         .mode           =       0644,
3859                         .proc_handler   =       &proc_dointvec,
3860                 },
3861                 {
3862                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3863                         .procname       =       "regen_max_retry",
3864                         .data           =       &ipv6_devconf.regen_max_retry,
3865                         .maxlen         =       sizeof(int),
3866                         .mode           =       0644,
3867                         .proc_handler   =       &proc_dointvec,
3868                 },
3869                 {
3870                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3871                         .procname       =       "max_desync_factor",
3872                         .data           =       &ipv6_devconf.max_desync_factor,
3873                         .maxlen         =       sizeof(int),
3874                         .mode           =       0644,
3875                         .proc_handler   =       &proc_dointvec,
3876                 },
3877 #endif
3878                 {
3879                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3880                         .procname       =       "max_addresses",
3881                         .data           =       &ipv6_devconf.max_addresses,
3882                         .maxlen         =       sizeof(int),
3883                         .mode           =       0644,
3884                         .proc_handler   =       &proc_dointvec,
3885                 },
3886                 {
3887                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3888                         .procname       =       "accept_ra_defrtr",
3889                         .data           =       &ipv6_devconf.accept_ra_defrtr,
3890                         .maxlen         =       sizeof(int),
3891                         .mode           =       0644,
3892                         .proc_handler   =       &proc_dointvec,
3893                 },
3894                 {
3895                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
3896                         .procname       =       "accept_ra_pinfo",
3897                         .data           =       &ipv6_devconf.accept_ra_pinfo,
3898                         .maxlen         =       sizeof(int),
3899                         .mode           =       0644,
3900                         .proc_handler   =       &proc_dointvec,
3901                 },
3902 #ifdef CONFIG_IPV6_ROUTER_PREF
3903                 {
3904                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
3905                         .procname       =       "accept_ra_rtr_pref",
3906                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
3907                         .maxlen         =       sizeof(int),
3908                         .mode           =       0644,
3909                         .proc_handler   =       &proc_dointvec,
3910                 },
3911                 {
3912                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
3913                         .procname       =       "router_probe_interval",
3914                         .data           =       &ipv6_devconf.rtr_probe_interval,
3915                         .maxlen         =       sizeof(int),
3916                         .mode           =       0644,
3917                         .proc_handler   =       &proc_dointvec_jiffies,
3918                         .strategy       =       &sysctl_jiffies,
3919                 },
3920 #ifdef CONFIG_IPV6_ROUTE_INFO
3921                 {
3922                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
3923                         .procname       =       "accept_ra_rt_info_max_plen",
3924                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
3925                         .maxlen         =       sizeof(int),
3926                         .mode           =       0644,
3927                         .proc_handler   =       &proc_dointvec,
3928                 },
3929 #endif
3930 #endif
3931                 {
3932                         .ctl_name       =       NET_IPV6_PROXY_NDP,
3933                         .procname       =       "proxy_ndp",
3934                         .data           =       &ipv6_devconf.proxy_ndp,
3935                         .maxlen         =       sizeof(int),
3936                         .mode           =       0644,
3937                         .proc_handler   =       &proc_dointvec,
3938                 },
3939                 {
3940                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
3941                         .procname       =       "accept_source_route",
3942                         .data           =       &ipv6_devconf.accept_source_route,
3943                         .maxlen         =       sizeof(int),
3944                         .mode           =       0644,
3945                         .proc_handler   =       &proc_dointvec,
3946                 },
3947                 {
3948                         .ctl_name       =       0,      /* sentinel */
3949                 }
3950         },
3951         .addrconf_dev = {
3952                 {
3953                         .ctl_name       =       NET_PROTO_CONF_ALL,
3954                         .procname       =       "all",
3955                         .mode           =       0555,
3956                         .child          =       addrconf_sysctl.addrconf_vars,
3957                 },
3958                 {
3959                         .ctl_name       =       0,      /* sentinel */
3960                 }
3961         },
3962         .addrconf_conf_dir = {
3963                 {
3964                         .ctl_name       =       NET_IPV6_CONF,
3965                         .procname       =       "conf",
3966                         .mode           =       0555,
3967                         .child          =       addrconf_sysctl.addrconf_dev,
3968                 },
3969                 {
3970                         .ctl_name       =       0,      /* sentinel */
3971                 }
3972         },
3973         .addrconf_proto_dir = {
3974                 {
3975                         .ctl_name       =       NET_IPV6,
3976                         .procname       =       "ipv6",
3977                         .mode           =       0555,
3978                         .child          =       addrconf_sysctl.addrconf_conf_dir,
3979                 },
3980                 {
3981                         .ctl_name       =       0,      /* sentinel */
3982                 }
3983         },
3984         .addrconf_root_dir = {
3985                 {
3986                         .ctl_name       =       CTL_NET,
3987                         .procname       =       "net",
3988                         .mode           =       0555,
3989                         .child          =       addrconf_sysctl.addrconf_proto_dir,
3990                 },
3991                 {
3992                         .ctl_name       =       0,      /* sentinel */
3993                 }
3994         },
3995 };
3996
3997 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
3998 {
3999         int i;
4000         struct net_device *dev = idev ? idev->dev : NULL;
4001         struct addrconf_sysctl_table *t;
4002         char *dev_name = NULL;
4003
4004         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4005         if (t == NULL)
4006                 return;
4007         for (i=0; t->addrconf_vars[i].data; i++) {
4008                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4009                 t->addrconf_vars[i].de = NULL;
4010                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4011         }
4012         if (dev) {
4013                 dev_name = dev->name; 
4014                 t->addrconf_dev[0].ctl_name = dev->ifindex;
4015         } else {
4016                 dev_name = "default";
4017                 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
4018         }
4019
4020         /* 
4021          * Make a copy of dev_name, because '.procname' is regarded as const 
4022          * by sysctl and we wouldn't want anyone to change it under our feet
4023          * (see SIOCSIFNAME).
4024          */     
4025         dev_name = kstrdup(dev_name, GFP_KERNEL);
4026         if (!dev_name)
4027             goto free;
4028
4029         t->addrconf_dev[0].procname = dev_name;
4030
4031         t->addrconf_dev[0].child = t->addrconf_vars;
4032         t->addrconf_dev[0].de = NULL;
4033         t->addrconf_conf_dir[0].child = t->addrconf_dev;
4034         t->addrconf_conf_dir[0].de = NULL;
4035         t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
4036         t->addrconf_proto_dir[0].de = NULL;
4037         t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
4038         t->addrconf_root_dir[0].de = NULL;
4039
4040         t->sysctl_header = register_sysctl_table(t->addrconf_root_dir, 0);
4041         if (t->sysctl_header == NULL)
4042                 goto free_procname;
4043         else
4044                 p->sysctl = t;
4045         return;
4046
4047         /* error path */
4048  free_procname:
4049         kfree(dev_name);
4050  free:
4051         kfree(t);
4052
4053         return;
4054 }
4055
4056 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
4057 {
4058         if (p->sysctl) {
4059                 struct addrconf_sysctl_table *t = p->sysctl;
4060                 p->sysctl = NULL;
4061                 unregister_sysctl_table(t->sysctl_header);
4062                 kfree(t->addrconf_dev[0].procname);
4063                 kfree(t);
4064         }
4065 }
4066
4067
4068 #endif
4069
4070 /*
4071  *      Device notifier
4072  */
4073
4074 int register_inet6addr_notifier(struct notifier_block *nb)
4075 {
4076         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4077 }
4078
4079 int unregister_inet6addr_notifier(struct notifier_block *nb)
4080 {
4081         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4082 }
4083
4084 /*
4085  *      Init / cleanup code
4086  */
4087
4088 int __init addrconf_init(void)
4089 {
4090         int err = 0;
4091
4092         /* The addrconf netdev notifier requires that loopback_dev
4093          * has it's ipv6 private information allocated and setup
4094          * before it can bring up and give link-local addresses
4095          * to other devices which are up.
4096          *
4097          * Unfortunately, loopback_dev is not necessarily the first
4098          * entry in the global dev_base list of net devices.  In fact,
4099          * it is likely to be the very last entry on that list.
4100          * So this causes the notifier registry below to try and
4101          * give link-local addresses to all devices besides loopback_dev
4102          * first, then loopback_dev, which cases all the non-loopback_dev
4103          * devices to fail to get a link-local address.
4104          *
4105          * So, as a temporary fix, allocate the ipv6 structure for
4106          * loopback_dev first by hand.
4107          * Longer term, all of the dependencies ipv6 has upon the loopback
4108          * device and it being up should be removed.
4109          */
4110         rtnl_lock();
4111         if (!ipv6_add_dev(&loopback_dev))
4112                 err = -ENOMEM;
4113         rtnl_unlock();
4114         if (err)
4115                 return err;
4116
4117         ip6_null_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4118
4119         register_netdevice_notifier(&ipv6_dev_notf);
4120
4121         addrconf_verify(0);
4122         rtnetlink_links[PF_INET6] = inet6_rtnetlink_table;
4123 #ifdef CONFIG_SYSCTL
4124         addrconf_sysctl.sysctl_header =
4125                 register_sysctl_table(addrconf_sysctl.addrconf_root_dir, 0);
4126         addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4127 #endif
4128
4129         return 0;
4130 }
4131
4132 void __exit addrconf_cleanup(void)
4133 {
4134         struct net_device *dev;
4135         struct inet6_dev *idev;
4136         struct inet6_ifaddr *ifa;
4137         int i;
4138
4139         unregister_netdevice_notifier(&ipv6_dev_notf);
4140
4141         rtnetlink_links[PF_INET6] = NULL;
4142 #ifdef CONFIG_SYSCTL
4143         addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4144         addrconf_sysctl_unregister(&ipv6_devconf);
4145 #endif
4146
4147         rtnl_lock();
4148
4149         /*
4150          *      clean dev list.
4151          */
4152
4153         for (dev=dev_base; dev; dev=dev->next) {
4154                 if ((idev = __in6_dev_get(dev)) == NULL)
4155                         continue;
4156                 addrconf_ifdown(dev, 1);
4157         }
4158         addrconf_ifdown(&loopback_dev, 2);
4159
4160         /*
4161          *      Check hash table.
4162          */
4163
4164         write_lock_bh(&addrconf_hash_lock);
4165         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4166                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4167                         struct inet6_ifaddr *bifa;
4168
4169                         bifa = ifa;
4170                         ifa = ifa->lst_next;
4171                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4172                         /* Do not free it; something is wrong.
4173                            Now we can investigate it with debugger.
4174                          */
4175                 }
4176         }
4177         write_unlock_bh(&addrconf_hash_lock);
4178
4179         del_timer(&addr_chk_timer);
4180
4181         rtnl_unlock();
4182
4183 #ifdef CONFIG_PROC_FS
4184         proc_net_remove("if_inet6");
4185 #endif
4186 }