xref: /linux/net/ipv6/addrconf.c (revision ba6e8564f459211117ce300eae2c7fdd23befe34)
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/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 
65 #include <net/sock.h>
66 #include <net/snmp.h>
67 
68 #include <net/ipv6.h>
69 #include <net/protocol.h>
70 #include <net/ndisc.h>
71 #include <net/ip6_route.h>
72 #include <net/addrconf.h>
73 #include <net/tcp.h>
74 #include <net/ip.h>
75 #include <net/netlink.h>
76 #include <linux/if_tunnel.h>
77 #include <linux/rtnetlink.h>
78 
79 #ifdef CONFIG_IPV6_PRIVACY
80 #include <linux/random.h>
81 #endif
82 
83 #include <asm/uaccess.h>
84 
85 #include <linux/proc_fs.h>
86 #include <linux/seq_file.h>
87 
88 /* Set to 3 to get tracing... */
89 #define ACONF_DEBUG 2
90 
91 #if ACONF_DEBUG >= 3
92 #define ADBG(x) printk x
93 #else
94 #define ADBG(x)
95 #endif
96 
97 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
98 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
99 
100 #ifdef CONFIG_SYSCTL
101 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p);
102 static void addrconf_sysctl_unregister(struct ipv6_devconf *p);
103 #endif
104 
105 #ifdef CONFIG_IPV6_PRIVACY
106 static int __ipv6_regen_rndid(struct inet6_dev *idev);
107 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
108 static void ipv6_regen_rndid(unsigned long data);
109 
110 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
111 #endif
112 
113 static int ipv6_count_addresses(struct inet6_dev *idev);
114 
115 /*
116  *	Configured unicast address hash table
117  */
118 static struct inet6_ifaddr		*inet6_addr_lst[IN6_ADDR_HSIZE];
119 static DEFINE_RWLOCK(addrconf_hash_lock);
120 
121 static void addrconf_verify(unsigned long);
122 
123 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
124 static DEFINE_SPINLOCK(addrconf_verify_lock);
125 
126 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
127 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
128 
129 static int addrconf_ifdown(struct net_device *dev, int how);
130 
131 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
132 static void addrconf_dad_timer(unsigned long data);
133 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
134 static void addrconf_dad_run(struct inet6_dev *idev);
135 static void addrconf_rs_timer(unsigned long data);
136 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
137 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
138 
139 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
140 				struct prefix_info *pinfo);
141 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
142 
143 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
144 
145 struct ipv6_devconf ipv6_devconf __read_mostly = {
146 	.forwarding		= 0,
147 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
148 	.mtu6			= IPV6_MIN_MTU,
149 	.accept_ra		= 1,
150 	.accept_redirects	= 1,
151 	.autoconf		= 1,
152 	.force_mld_version	= 0,
153 	.dad_transmits		= 1,
154 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
155 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
156 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
157 #ifdef CONFIG_IPV6_PRIVACY
158 	.use_tempaddr 		= 0,
159 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
160 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
161 	.regen_max_retry	= REGEN_MAX_RETRY,
162 	.max_desync_factor	= MAX_DESYNC_FACTOR,
163 #endif
164 	.max_addresses		= IPV6_MAX_ADDRESSES,
165 	.accept_ra_defrtr	= 1,
166 	.accept_ra_pinfo	= 1,
167 #ifdef CONFIG_IPV6_ROUTER_PREF
168 	.accept_ra_rtr_pref	= 1,
169 	.rtr_probe_interval	= 60 * HZ,
170 #ifdef CONFIG_IPV6_ROUTE_INFO
171 	.accept_ra_rt_info_max_plen = 0,
172 #endif
173 #endif
174 	.proxy_ndp		= 0,
175 };
176 
177 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
178 	.forwarding		= 0,
179 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
180 	.mtu6			= IPV6_MIN_MTU,
181 	.accept_ra		= 1,
182 	.accept_redirects	= 1,
183 	.autoconf		= 1,
184 	.dad_transmits		= 1,
185 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
186 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
187 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
188 #ifdef CONFIG_IPV6_PRIVACY
189 	.use_tempaddr		= 0,
190 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
191 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
192 	.regen_max_retry	= REGEN_MAX_RETRY,
193 	.max_desync_factor	= MAX_DESYNC_FACTOR,
194 #endif
195 	.max_addresses		= IPV6_MAX_ADDRESSES,
196 	.accept_ra_defrtr	= 1,
197 	.accept_ra_pinfo	= 1,
198 #ifdef CONFIG_IPV6_ROUTER_PREF
199 	.accept_ra_rtr_pref	= 1,
200 	.rtr_probe_interval	= 60 * HZ,
201 #ifdef CONFIG_IPV6_ROUTE_INFO
202 	.accept_ra_rt_info_max_plen = 0,
203 #endif
204 #endif
205 	.proxy_ndp		= 0,
206 };
207 
208 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
209 #if 0
210 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
211 #endif
212 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
213 
214 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
215 {
216 	if (del_timer(&ifp->timer))
217 		__in6_ifa_put(ifp);
218 }
219 
220 enum addrconf_timer_t
221 {
222 	AC_NONE,
223 	AC_DAD,
224 	AC_RS,
225 };
226 
227 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
228 			       enum addrconf_timer_t what,
229 			       unsigned long when)
230 {
231 	if (!del_timer(&ifp->timer))
232 		in6_ifa_hold(ifp);
233 
234 	switch (what) {
235 	case AC_DAD:
236 		ifp->timer.function = addrconf_dad_timer;
237 		break;
238 	case AC_RS:
239 		ifp->timer.function = addrconf_rs_timer;
240 		break;
241 	default:;
242 	}
243 	ifp->timer.expires = jiffies + when;
244 	add_timer(&ifp->timer);
245 }
246 
247 /* Nobody refers to this device, we may destroy it. */
248 
249 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
250 {
251 	struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
252 	kfree(idev);
253 }
254 
255 void in6_dev_finish_destroy(struct inet6_dev *idev)
256 {
257 	struct net_device *dev = idev->dev;
258 	BUG_TRAP(idev->addr_list==NULL);
259 	BUG_TRAP(idev->mc_list==NULL);
260 #ifdef NET_REFCNT_DEBUG
261 	printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
262 #endif
263 	dev_put(dev);
264 	if (!idev->dead) {
265 		printk("Freeing alive inet6 device %p\n", idev);
266 		return;
267 	}
268 	snmp6_free_dev(idev);
269 	call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
270 }
271 
272 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
273 {
274 	struct inet6_dev *ndev;
275 	struct in6_addr maddr;
276 
277 	ASSERT_RTNL();
278 
279 	if (dev->mtu < IPV6_MIN_MTU)
280 		return NULL;
281 
282 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
283 
284 	if (ndev == NULL)
285 		return NULL;
286 
287 	rwlock_init(&ndev->lock);
288 	ndev->dev = dev;
289 	memcpy(&ndev->cnf, &ipv6_devconf_dflt, sizeof(ndev->cnf));
290 	ndev->cnf.mtu6 = dev->mtu;
291 	ndev->cnf.sysctl = NULL;
292 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
293 	if (ndev->nd_parms == NULL) {
294 		kfree(ndev);
295 		return NULL;
296 	}
297 	/* We refer to the device */
298 	dev_hold(dev);
299 
300 	if (snmp6_alloc_dev(ndev) < 0) {
301 		ADBG((KERN_WARNING
302 			"%s(): cannot allocate memory for statistics; dev=%s.\n",
303 			__FUNCTION__, dev->name));
304 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
305 		ndev->dead = 1;
306 		in6_dev_finish_destroy(ndev);
307 		return NULL;
308 	}
309 
310 	if (snmp6_register_dev(ndev) < 0) {
311 		ADBG((KERN_WARNING
312 			"%s(): cannot create /proc/net/dev_snmp6/%s\n",
313 			__FUNCTION__, dev->name));
314 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
315 		ndev->dead = 1;
316 		in6_dev_finish_destroy(ndev);
317 		return NULL;
318 	}
319 
320 	/* One reference from device.  We must do this before
321 	 * we invoke __ipv6_regen_rndid().
322 	 */
323 	in6_dev_hold(ndev);
324 
325 #ifdef CONFIG_IPV6_PRIVACY
326 	init_timer(&ndev->regen_timer);
327 	ndev->regen_timer.function = ipv6_regen_rndid;
328 	ndev->regen_timer.data = (unsigned long) ndev;
329 	if ((dev->flags&IFF_LOOPBACK) ||
330 	    dev->type == ARPHRD_TUNNEL ||
331 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
332 	    dev->type == ARPHRD_SIT ||
333 #endif
334 	    dev->type == ARPHRD_NONE) {
335 		printk(KERN_INFO
336 		       "%s: Disabled Privacy Extensions\n",
337 		       dev->name);
338 		ndev->cnf.use_tempaddr = -1;
339 	} else {
340 		in6_dev_hold(ndev);
341 		ipv6_regen_rndid((unsigned long) ndev);
342 	}
343 #endif
344 
345 	if (netif_running(dev) && netif_carrier_ok(dev))
346 		ndev->if_flags |= IF_READY;
347 
348 	ipv6_mc_init_dev(ndev);
349 	ndev->tstamp = jiffies;
350 #ifdef CONFIG_SYSCTL
351 	neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
352 			      NET_IPV6_NEIGH, "ipv6",
353 			      &ndisc_ifinfo_sysctl_change,
354 			      NULL);
355 	addrconf_sysctl_register(ndev, &ndev->cnf);
356 #endif
357 	/* protected by rtnl_lock */
358 	rcu_assign_pointer(dev->ip6_ptr, ndev);
359 
360 	/* Join all-node multicast group */
361 	ipv6_addr_all_nodes(&maddr);
362 	ipv6_dev_mc_inc(dev, &maddr);
363 
364 	return ndev;
365 }
366 
367 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
368 {
369 	struct inet6_dev *idev;
370 
371 	ASSERT_RTNL();
372 
373 	if ((idev = __in6_dev_get(dev)) == NULL) {
374 		if ((idev = ipv6_add_dev(dev)) == NULL)
375 			return NULL;
376 	}
377 
378 	if (dev->flags&IFF_UP)
379 		ipv6_mc_up(idev);
380 	return idev;
381 }
382 
383 #ifdef CONFIG_SYSCTL
384 static void dev_forward_change(struct inet6_dev *idev)
385 {
386 	struct net_device *dev;
387 	struct inet6_ifaddr *ifa;
388 	struct in6_addr addr;
389 
390 	if (!idev)
391 		return;
392 	dev = idev->dev;
393 	if (dev && (dev->flags & IFF_MULTICAST)) {
394 		ipv6_addr_all_routers(&addr);
395 
396 		if (idev->cnf.forwarding)
397 			ipv6_dev_mc_inc(dev, &addr);
398 		else
399 			ipv6_dev_mc_dec(dev, &addr);
400 	}
401 	for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
402 		if (ifa->flags&IFA_F_TENTATIVE)
403 			continue;
404 		if (idev->cnf.forwarding)
405 			addrconf_join_anycast(ifa);
406 		else
407 			addrconf_leave_anycast(ifa);
408 	}
409 }
410 
411 
412 static void addrconf_forward_change(void)
413 {
414 	struct net_device *dev;
415 	struct inet6_dev *idev;
416 
417 	read_lock(&dev_base_lock);
418 	for (dev=dev_base; dev; dev=dev->next) {
419 		rcu_read_lock();
420 		idev = __in6_dev_get(dev);
421 		if (idev) {
422 			int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
423 			idev->cnf.forwarding = ipv6_devconf.forwarding;
424 			if (changed)
425 				dev_forward_change(idev);
426 		}
427 		rcu_read_unlock();
428 	}
429 	read_unlock(&dev_base_lock);
430 }
431 #endif
432 
433 /* Nobody refers to this ifaddr, destroy it */
434 
435 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
436 {
437 	BUG_TRAP(ifp->if_next==NULL);
438 	BUG_TRAP(ifp->lst_next==NULL);
439 #ifdef NET_REFCNT_DEBUG
440 	printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
441 #endif
442 
443 	in6_dev_put(ifp->idev);
444 
445 	if (del_timer(&ifp->timer))
446 		printk("Timer is still running, when freeing ifa=%p\n", ifp);
447 
448 	if (!ifp->dead) {
449 		printk("Freeing alive inet6 address %p\n", ifp);
450 		return;
451 	}
452 	dst_release(&ifp->rt->u.dst);
453 
454 	kfree(ifp);
455 }
456 
457 static void
458 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
459 {
460 	struct inet6_ifaddr *ifa, **ifap;
461 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
462 
463 	/*
464 	 * Each device address list is sorted in order of scope -
465 	 * global before linklocal.
466 	 */
467 	for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
468 	     ifap = &ifa->if_next) {
469 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
470 			break;
471 	}
472 
473 	ifp->if_next = *ifap;
474 	*ifap = ifp;
475 }
476 
477 /* On success it returns ifp with increased reference count */
478 
479 static struct inet6_ifaddr *
480 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
481 	      int scope, u32 flags)
482 {
483 	struct inet6_ifaddr *ifa = NULL;
484 	struct rt6_info *rt;
485 	int hash;
486 	int err = 0;
487 
488 	rcu_read_lock_bh();
489 	if (idev->dead) {
490 		err = -ENODEV;			/*XXX*/
491 		goto out2;
492 	}
493 
494 	write_lock(&addrconf_hash_lock);
495 
496 	/* Ignore adding duplicate addresses on an interface */
497 	if (ipv6_chk_same_addr(addr, idev->dev)) {
498 		ADBG(("ipv6_add_addr: already assigned\n"));
499 		err = -EEXIST;
500 		goto out;
501 	}
502 
503 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
504 
505 	if (ifa == NULL) {
506 		ADBG(("ipv6_add_addr: malloc failed\n"));
507 		err = -ENOBUFS;
508 		goto out;
509 	}
510 
511 	rt = addrconf_dst_alloc(idev, addr, 0);
512 	if (IS_ERR(rt)) {
513 		err = PTR_ERR(rt);
514 		goto out;
515 	}
516 
517 	ipv6_addr_copy(&ifa->addr, addr);
518 
519 	spin_lock_init(&ifa->lock);
520 	init_timer(&ifa->timer);
521 	ifa->timer.data = (unsigned long) ifa;
522 	ifa->scope = scope;
523 	ifa->prefix_len = pfxlen;
524 	ifa->flags = flags | IFA_F_TENTATIVE;
525 	ifa->cstamp = ifa->tstamp = jiffies;
526 
527 	ifa->rt = rt;
528 
529 	ifa->idev = idev;
530 	in6_dev_hold(idev);
531 	/* For caller */
532 	in6_ifa_hold(ifa);
533 
534 	/* Add to big hash table */
535 	hash = ipv6_addr_hash(addr);
536 
537 	ifa->lst_next = inet6_addr_lst[hash];
538 	inet6_addr_lst[hash] = ifa;
539 	in6_ifa_hold(ifa);
540 	write_unlock(&addrconf_hash_lock);
541 
542 	write_lock(&idev->lock);
543 	/* Add to inet6_dev unicast addr list. */
544 	ipv6_link_dev_addr(idev, ifa);
545 
546 #ifdef CONFIG_IPV6_PRIVACY
547 	if (ifa->flags&IFA_F_TEMPORARY) {
548 		ifa->tmp_next = idev->tempaddr_list;
549 		idev->tempaddr_list = ifa;
550 		in6_ifa_hold(ifa);
551 	}
552 #endif
553 
554 	in6_ifa_hold(ifa);
555 	write_unlock(&idev->lock);
556 out2:
557 	rcu_read_unlock_bh();
558 
559 	if (likely(err == 0))
560 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
561 	else {
562 		kfree(ifa);
563 		ifa = ERR_PTR(err);
564 	}
565 
566 	return ifa;
567 out:
568 	write_unlock(&addrconf_hash_lock);
569 	goto out2;
570 }
571 
572 /* This function wants to get referenced ifp and releases it before return */
573 
574 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
575 {
576 	struct inet6_ifaddr *ifa, **ifap;
577 	struct inet6_dev *idev = ifp->idev;
578 	int hash;
579 	int deleted = 0, onlink = 0;
580 	unsigned long expires = jiffies;
581 
582 	hash = ipv6_addr_hash(&ifp->addr);
583 
584 	ifp->dead = 1;
585 
586 	write_lock_bh(&addrconf_hash_lock);
587 	for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
588 	     ifap = &ifa->lst_next) {
589 		if (ifa == ifp) {
590 			*ifap = ifa->lst_next;
591 			__in6_ifa_put(ifp);
592 			ifa->lst_next = NULL;
593 			break;
594 		}
595 	}
596 	write_unlock_bh(&addrconf_hash_lock);
597 
598 	write_lock_bh(&idev->lock);
599 #ifdef CONFIG_IPV6_PRIVACY
600 	if (ifp->flags&IFA_F_TEMPORARY) {
601 		for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
602 		     ifap = &ifa->tmp_next) {
603 			if (ifa == ifp) {
604 				*ifap = ifa->tmp_next;
605 				if (ifp->ifpub) {
606 					in6_ifa_put(ifp->ifpub);
607 					ifp->ifpub = NULL;
608 				}
609 				__in6_ifa_put(ifp);
610 				ifa->tmp_next = NULL;
611 				break;
612 			}
613 		}
614 	}
615 #endif
616 
617 	for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
618 		if (ifa == ifp) {
619 			*ifap = ifa->if_next;
620 			__in6_ifa_put(ifp);
621 			ifa->if_next = NULL;
622 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
623 				break;
624 			deleted = 1;
625 			continue;
626 		} else if (ifp->flags & IFA_F_PERMANENT) {
627 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
628 					      ifp->prefix_len)) {
629 				if (ifa->flags & IFA_F_PERMANENT) {
630 					onlink = 1;
631 					if (deleted)
632 						break;
633 				} else {
634 					unsigned long lifetime;
635 
636 					if (!onlink)
637 						onlink = -1;
638 
639 					spin_lock(&ifa->lock);
640 					lifetime = min_t(unsigned long,
641 							 ifa->valid_lft, 0x7fffffffUL/HZ);
642 					if (time_before(expires,
643 							ifa->tstamp + lifetime * HZ))
644 						expires = ifa->tstamp + lifetime * HZ;
645 					spin_unlock(&ifa->lock);
646 				}
647 			}
648 		}
649 		ifap = &ifa->if_next;
650 	}
651 	write_unlock_bh(&idev->lock);
652 
653 	ipv6_ifa_notify(RTM_DELADDR, ifp);
654 
655 	atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
656 
657 	addrconf_del_timer(ifp);
658 
659 	/*
660 	 * Purge or update corresponding prefix
661 	 *
662 	 * 1) we don't purge prefix here if address was not permanent.
663 	 *    prefix is managed by its own lifetime.
664 	 * 2) if there're no addresses, delete prefix.
665 	 * 3) if there're still other permanent address(es),
666 	 *    corresponding prefix is still permanent.
667 	 * 4) otherwise, update prefix lifetime to the
668 	 *    longest valid lifetime among the corresponding
669 	 *    addresses on the device.
670 	 *    Note: subsequent RA will update lifetime.
671 	 *
672 	 * --yoshfuji
673 	 */
674 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
675 		struct in6_addr prefix;
676 		struct rt6_info *rt;
677 
678 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
679 		rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
680 
681 		if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
682 			if (onlink == 0) {
683 				ip6_del_rt(rt);
684 				rt = NULL;
685 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
686 				rt->rt6i_expires = expires;
687 				rt->rt6i_flags |= RTF_EXPIRES;
688 			}
689 		}
690 		dst_release(&rt->u.dst);
691 	}
692 
693 	in6_ifa_put(ifp);
694 }
695 
696 #ifdef CONFIG_IPV6_PRIVACY
697 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
698 {
699 	struct inet6_dev *idev = ifp->idev;
700 	struct in6_addr addr, *tmpaddr;
701 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
702 	int tmp_plen;
703 	int ret = 0;
704 	int max_addresses;
705 
706 	write_lock(&idev->lock);
707 	if (ift) {
708 		spin_lock_bh(&ift->lock);
709 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
710 		spin_unlock_bh(&ift->lock);
711 		tmpaddr = &addr;
712 	} else {
713 		tmpaddr = NULL;
714 	}
715 retry:
716 	in6_dev_hold(idev);
717 	if (idev->cnf.use_tempaddr <= 0) {
718 		write_unlock(&idev->lock);
719 		printk(KERN_INFO
720 			"ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
721 		in6_dev_put(idev);
722 		ret = -1;
723 		goto out;
724 	}
725 	spin_lock_bh(&ifp->lock);
726 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
727 		idev->cnf.use_tempaddr = -1;	/*XXX*/
728 		spin_unlock_bh(&ifp->lock);
729 		write_unlock(&idev->lock);
730 		printk(KERN_WARNING
731 			"ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
732 		in6_dev_put(idev);
733 		ret = -1;
734 		goto out;
735 	}
736 	in6_ifa_hold(ifp);
737 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
738 	if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
739 		spin_unlock_bh(&ifp->lock);
740 		write_unlock(&idev->lock);
741 		printk(KERN_WARNING
742 			"ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
743 		in6_ifa_put(ifp);
744 		in6_dev_put(idev);
745 		ret = -1;
746 		goto out;
747 	}
748 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
749 	tmp_valid_lft = min_t(__u32,
750 			      ifp->valid_lft,
751 			      idev->cnf.temp_valid_lft);
752 	tmp_prefered_lft = min_t(__u32,
753 				 ifp->prefered_lft,
754 				 idev->cnf.temp_prefered_lft - desync_factor / HZ);
755 	tmp_plen = ifp->prefix_len;
756 	max_addresses = idev->cnf.max_addresses;
757 	tmp_cstamp = ifp->cstamp;
758 	tmp_tstamp = ifp->tstamp;
759 	spin_unlock_bh(&ifp->lock);
760 
761 	write_unlock(&idev->lock);
762 	ift = !max_addresses ||
763 	      ipv6_count_addresses(idev) < max_addresses ?
764 		ipv6_add_addr(idev, &addr, tmp_plen,
765 			      ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK, IFA_F_TEMPORARY) : NULL;
766 	if (!ift || IS_ERR(ift)) {
767 		in6_ifa_put(ifp);
768 		in6_dev_put(idev);
769 		printk(KERN_INFO
770 			"ipv6_create_tempaddr(): retry temporary address regeneration.\n");
771 		tmpaddr = &addr;
772 		write_lock(&idev->lock);
773 		goto retry;
774 	}
775 
776 	spin_lock_bh(&ift->lock);
777 	ift->ifpub = ifp;
778 	ift->valid_lft = tmp_valid_lft;
779 	ift->prefered_lft = tmp_prefered_lft;
780 	ift->cstamp = tmp_cstamp;
781 	ift->tstamp = tmp_tstamp;
782 	spin_unlock_bh(&ift->lock);
783 
784 	addrconf_dad_start(ift, 0);
785 	in6_ifa_put(ift);
786 	in6_dev_put(idev);
787 out:
788 	return ret;
789 }
790 #endif
791 
792 /*
793  *	Choose an appropriate source address (RFC3484)
794  */
795 struct ipv6_saddr_score {
796 	int		addr_type;
797 	unsigned int	attrs;
798 	int		matchlen;
799 	int		scope;
800 	unsigned int	rule;
801 };
802 
803 #define IPV6_SADDR_SCORE_LOCAL		0x0001
804 #define IPV6_SADDR_SCORE_PREFERRED	0x0004
805 #define IPV6_SADDR_SCORE_HOA		0x0008
806 #define IPV6_SADDR_SCORE_OIF		0x0010
807 #define IPV6_SADDR_SCORE_LABEL		0x0020
808 #define IPV6_SADDR_SCORE_PRIVACY	0x0040
809 
810 static inline int ipv6_saddr_preferred(int type)
811 {
812 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
813 		    IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
814 		return 1;
815 	return 0;
816 }
817 
818 /* static matching label */
819 static inline int ipv6_saddr_label(const struct in6_addr *addr, int type)
820 {
821  /*
822   * 	prefix (longest match)	label
823   * 	-----------------------------
824   * 	::1/128			0
825   * 	::/0			1
826   * 	2002::/16		2
827   * 	::/96			3
828   * 	::ffff:0:0/96		4
829   *	fc00::/7		5
830   * 	2001::/32		6
831   */
832 	if (type & IPV6_ADDR_LOOPBACK)
833 		return 0;
834 	else if (type & IPV6_ADDR_COMPATv4)
835 		return 3;
836 	else if (type & IPV6_ADDR_MAPPED)
837 		return 4;
838 	else if (addr->s6_addr32[0] == htonl(0x20010000))
839 		return 6;
840 	else if (addr->s6_addr16[0] == htons(0x2002))
841 		return 2;
842 	else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
843 		return 5;
844 	return 1;
845 }
846 
847 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
848 		       struct in6_addr *daddr, struct in6_addr *saddr)
849 {
850 	struct ipv6_saddr_score hiscore;
851 	struct inet6_ifaddr *ifa_result = NULL;
852 	int daddr_type = __ipv6_addr_type(daddr);
853 	int daddr_scope = __ipv6_addr_src_scope(daddr_type);
854 	u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
855 	struct net_device *dev;
856 
857 	memset(&hiscore, 0, sizeof(hiscore));
858 
859 	read_lock(&dev_base_lock);
860 	rcu_read_lock();
861 
862 	for (dev = dev_base; dev; dev=dev->next) {
863 		struct inet6_dev *idev;
864 		struct inet6_ifaddr *ifa;
865 
866 		/* Rule 0: Candidate Source Address (section 4)
867 		 *  - multicast and link-local destination address,
868 		 *    the set of candidate source address MUST only
869 		 *    include addresses assigned to interfaces
870 		 *    belonging to the same link as the outgoing
871 		 *    interface.
872 		 * (- For site-local destination addresses, the
873 		 *    set of candidate source addresses MUST only
874 		 *    include addresses assigned to interfaces
875 		 *    belonging to the same site as the outgoing
876 		 *    interface.)
877 		 */
878 		if ((daddr_type & IPV6_ADDR_MULTICAST ||
879 		     daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
880 		    daddr_dev && dev != daddr_dev)
881 			continue;
882 
883 		idev = __in6_dev_get(dev);
884 		if (!idev)
885 			continue;
886 
887 		read_lock_bh(&idev->lock);
888 		for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
889 			struct ipv6_saddr_score score;
890 
891 			score.addr_type = __ipv6_addr_type(&ifa->addr);
892 
893 			/* Rule 0:
894 			 * - Tentative Address (RFC2462 section 5.4)
895 			 *  - A tentative address is not considered
896 			 *    "assigned to an interface" in the traditional
897 			 *    sense.
898 			 * - Candidate Source Address (section 4)
899 			 *  - In any case, anycast addresses, multicast
900 			 *    addresses, and the unspecified address MUST
901 			 *    NOT be included in a candidate set.
902 			 */
903 			if (ifa->flags & IFA_F_TENTATIVE)
904 				continue;
905 			if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
906 				     score.addr_type & IPV6_ADDR_MULTICAST)) {
907 				LIMIT_NETDEBUG(KERN_DEBUG
908 					       "ADDRCONF: unspecified / multicast address"
909 					       "assigned as unicast address on %s",
910 					       dev->name);
911 				continue;
912 			}
913 
914 			score.attrs = 0;
915 			score.matchlen = 0;
916 			score.scope = 0;
917 			score.rule = 0;
918 
919 			if (ifa_result == NULL) {
920 				/* record it if the first available entry */
921 				goto record_it;
922 			}
923 
924 			/* Rule 1: Prefer same address */
925 			if (hiscore.rule < 1) {
926 				if (ipv6_addr_equal(&ifa_result->addr, daddr))
927 					hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
928 				hiscore.rule++;
929 			}
930 			if (ipv6_addr_equal(&ifa->addr, daddr)) {
931 				score.attrs |= IPV6_SADDR_SCORE_LOCAL;
932 				if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
933 					score.rule = 1;
934 					goto record_it;
935 				}
936 			} else {
937 				if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
938 					continue;
939 			}
940 
941 			/* Rule 2: Prefer appropriate scope */
942 			if (hiscore.rule < 2) {
943 				hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
944 				hiscore.rule++;
945 			}
946 			score.scope = __ipv6_addr_src_scope(score.addr_type);
947 			if (hiscore.scope < score.scope) {
948 				if (hiscore.scope < daddr_scope) {
949 					score.rule = 2;
950 					goto record_it;
951 				} else
952 					continue;
953 			} else if (score.scope < hiscore.scope) {
954 				if (score.scope < daddr_scope)
955 					break; /* addresses sorted by scope */
956 				else {
957 					score.rule = 2;
958 					goto record_it;
959 				}
960 			}
961 
962 			/* Rule 3: Avoid deprecated address */
963 			if (hiscore.rule < 3) {
964 				if (ipv6_saddr_preferred(hiscore.addr_type) ||
965 				    !(ifa_result->flags & IFA_F_DEPRECATED))
966 					hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
967 				hiscore.rule++;
968 			}
969 			if (ipv6_saddr_preferred(score.addr_type) ||
970 			    !(ifa->flags & IFA_F_DEPRECATED)) {
971 				score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
972 				if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
973 					score.rule = 3;
974 					goto record_it;
975 				}
976 			} else {
977 				if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
978 					continue;
979 			}
980 
981 			/* Rule 4: Prefer home address */
982 #ifdef CONFIG_IPV6_MIP6
983 			if (hiscore.rule < 4) {
984 				if (ifa_result->flags & IFA_F_HOMEADDRESS)
985 					hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
986 				hiscore.rule++;
987 			}
988 			if (ifa->flags & IFA_F_HOMEADDRESS) {
989 				score.attrs |= IPV6_SADDR_SCORE_HOA;
990 				if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
991 					score.rule = 4;
992 					goto record_it;
993 				}
994 			} else {
995 				if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
996 					continue;
997 			}
998 #else
999 			if (hiscore.rule < 4)
1000 				hiscore.rule++;
1001 #endif
1002 
1003 			/* Rule 5: Prefer outgoing interface */
1004 			if (hiscore.rule < 5) {
1005 				if (daddr_dev == NULL ||
1006 				    daddr_dev == ifa_result->idev->dev)
1007 					hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1008 				hiscore.rule++;
1009 			}
1010 			if (daddr_dev == NULL ||
1011 			    daddr_dev == ifa->idev->dev) {
1012 				score.attrs |= IPV6_SADDR_SCORE_OIF;
1013 				if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1014 					score.rule = 5;
1015 					goto record_it;
1016 				}
1017 			} else {
1018 				if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1019 					continue;
1020 			}
1021 
1022 			/* Rule 6: Prefer matching label */
1023 			if (hiscore.rule < 6) {
1024 				if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1025 					hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1026 				hiscore.rule++;
1027 			}
1028 			if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1029 				score.attrs |= IPV6_SADDR_SCORE_LABEL;
1030 				if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1031 					score.rule = 6;
1032 					goto record_it;
1033 				}
1034 			} else {
1035 				if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1036 					continue;
1037 			}
1038 
1039 #ifdef CONFIG_IPV6_PRIVACY
1040 			/* Rule 7: Prefer public address
1041 			 * Note: prefer temprary address if use_tempaddr >= 2
1042 			 */
1043 			if (hiscore.rule < 7) {
1044 				if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1045 				    (ifa_result->idev->cnf.use_tempaddr >= 2))
1046 					hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1047 				hiscore.rule++;
1048 			}
1049 			if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1050 			    (ifa->idev->cnf.use_tempaddr >= 2)) {
1051 				score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1052 				if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1053 					score.rule = 7;
1054 					goto record_it;
1055 				}
1056 			} else {
1057 				if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1058 					continue;
1059 			}
1060 #else
1061 			if (hiscore.rule < 7)
1062 				hiscore.rule++;
1063 #endif
1064 			/* Rule 8: Use longest matching prefix */
1065 			if (hiscore.rule < 8) {
1066 				hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1067 				hiscore.rule++;
1068 			}
1069 			score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1070 			if (score.matchlen > hiscore.matchlen) {
1071 				score.rule = 8;
1072 				goto record_it;
1073 			}
1074 #if 0
1075 			else if (score.matchlen < hiscore.matchlen)
1076 				continue;
1077 #endif
1078 
1079 			/* Final Rule: choose first available one */
1080 			continue;
1081 record_it:
1082 			if (ifa_result)
1083 				in6_ifa_put(ifa_result);
1084 			in6_ifa_hold(ifa);
1085 			ifa_result = ifa;
1086 			hiscore = score;
1087 		}
1088 		read_unlock_bh(&idev->lock);
1089 	}
1090 	rcu_read_unlock();
1091 	read_unlock(&dev_base_lock);
1092 
1093 	if (!ifa_result)
1094 		return -EADDRNOTAVAIL;
1095 
1096 	ipv6_addr_copy(saddr, &ifa_result->addr);
1097 	in6_ifa_put(ifa_result);
1098 	return 0;
1099 }
1100 
1101 
1102 int ipv6_get_saddr(struct dst_entry *dst,
1103 		   struct in6_addr *daddr, struct in6_addr *saddr)
1104 {
1105 	return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1106 }
1107 
1108 
1109 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
1110 {
1111 	struct inet6_dev *idev;
1112 	int err = -EADDRNOTAVAIL;
1113 
1114 	rcu_read_lock();
1115 	if ((idev = __in6_dev_get(dev)) != NULL) {
1116 		struct inet6_ifaddr *ifp;
1117 
1118 		read_lock_bh(&idev->lock);
1119 		for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1120 			if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1121 				ipv6_addr_copy(addr, &ifp->addr);
1122 				err = 0;
1123 				break;
1124 			}
1125 		}
1126 		read_unlock_bh(&idev->lock);
1127 	}
1128 	rcu_read_unlock();
1129 	return err;
1130 }
1131 
1132 static int ipv6_count_addresses(struct inet6_dev *idev)
1133 {
1134 	int cnt = 0;
1135 	struct inet6_ifaddr *ifp;
1136 
1137 	read_lock_bh(&idev->lock);
1138 	for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1139 		cnt++;
1140 	read_unlock_bh(&idev->lock);
1141 	return cnt;
1142 }
1143 
1144 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1145 {
1146 	struct inet6_ifaddr * ifp;
1147 	u8 hash = ipv6_addr_hash(addr);
1148 
1149 	read_lock_bh(&addrconf_hash_lock);
1150 	for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1151 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1152 		    !(ifp->flags&IFA_F_TENTATIVE)) {
1153 			if (dev == NULL || ifp->idev->dev == dev ||
1154 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1155 				break;
1156 		}
1157 	}
1158 	read_unlock_bh(&addrconf_hash_lock);
1159 	return ifp != NULL;
1160 }
1161 
1162 static
1163 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1164 {
1165 	struct inet6_ifaddr * ifp;
1166 	u8 hash = ipv6_addr_hash(addr);
1167 
1168 	for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1169 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1170 			if (dev == NULL || ifp->idev->dev == dev)
1171 				break;
1172 		}
1173 	}
1174 	return ifp != NULL;
1175 }
1176 
1177 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1178 {
1179 	struct inet6_ifaddr * ifp;
1180 	u8 hash = ipv6_addr_hash(addr);
1181 
1182 	read_lock_bh(&addrconf_hash_lock);
1183 	for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1184 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1185 			if (dev == NULL || ifp->idev->dev == dev ||
1186 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1187 				in6_ifa_hold(ifp);
1188 				break;
1189 			}
1190 		}
1191 	}
1192 	read_unlock_bh(&addrconf_hash_lock);
1193 
1194 	return ifp;
1195 }
1196 
1197 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1198 {
1199 	const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1200 	const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1201 	__be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1202 	__be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1203 	int sk_ipv6only = ipv6_only_sock(sk);
1204 	int sk2_ipv6only = inet_v6_ipv6only(sk2);
1205 	int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1206 	int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1207 
1208 	if (!sk2_rcv_saddr && !sk_ipv6only)
1209 		return 1;
1210 
1211 	if (addr_type2 == IPV6_ADDR_ANY &&
1212 	    !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1213 		return 1;
1214 
1215 	if (addr_type == IPV6_ADDR_ANY &&
1216 	    !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1217 		return 1;
1218 
1219 	if (sk2_rcv_saddr6 &&
1220 	    ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1221 		return 1;
1222 
1223 	if (addr_type == IPV6_ADDR_MAPPED &&
1224 	    !sk2_ipv6only &&
1225 	    (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1226 		return 1;
1227 
1228 	return 0;
1229 }
1230 
1231 /* Gets referenced address, destroys ifaddr */
1232 
1233 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1234 {
1235 	if (ifp->flags&IFA_F_PERMANENT) {
1236 		spin_lock_bh(&ifp->lock);
1237 		addrconf_del_timer(ifp);
1238 		ifp->flags |= IFA_F_TENTATIVE;
1239 		spin_unlock_bh(&ifp->lock);
1240 		in6_ifa_put(ifp);
1241 #ifdef CONFIG_IPV6_PRIVACY
1242 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1243 		struct inet6_ifaddr *ifpub;
1244 		spin_lock_bh(&ifp->lock);
1245 		ifpub = ifp->ifpub;
1246 		if (ifpub) {
1247 			in6_ifa_hold(ifpub);
1248 			spin_unlock_bh(&ifp->lock);
1249 			ipv6_create_tempaddr(ifpub, ifp);
1250 			in6_ifa_put(ifpub);
1251 		} else {
1252 			spin_unlock_bh(&ifp->lock);
1253 		}
1254 		ipv6_del_addr(ifp);
1255 #endif
1256 	} else
1257 		ipv6_del_addr(ifp);
1258 }
1259 
1260 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1261 {
1262 	if (net_ratelimit())
1263 		printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1264 	addrconf_dad_stop(ifp);
1265 }
1266 
1267 /* Join to solicited addr multicast group. */
1268 
1269 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1270 {
1271 	struct in6_addr maddr;
1272 
1273 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1274 		return;
1275 
1276 	addrconf_addr_solict_mult(addr, &maddr);
1277 	ipv6_dev_mc_inc(dev, &maddr);
1278 }
1279 
1280 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1281 {
1282 	struct in6_addr maddr;
1283 
1284 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1285 		return;
1286 
1287 	addrconf_addr_solict_mult(addr, &maddr);
1288 	__ipv6_dev_mc_dec(idev, &maddr);
1289 }
1290 
1291 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1292 {
1293 	struct in6_addr addr;
1294 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1295 	if (ipv6_addr_any(&addr))
1296 		return;
1297 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1298 }
1299 
1300 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1301 {
1302 	struct in6_addr addr;
1303 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1304 	if (ipv6_addr_any(&addr))
1305 		return;
1306 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1307 }
1308 
1309 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1310 {
1311 	if (dev->addr_len != ETH_ALEN)
1312 		return -1;
1313 	memcpy(eui, dev->dev_addr, 3);
1314 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1315 
1316 	/*
1317 	 * The zSeries OSA network cards can be shared among various
1318 	 * OS instances, but the OSA cards have only one MAC address.
1319 	 * This leads to duplicate address conflicts in conjunction
1320 	 * with IPv6 if more than one instance uses the same card.
1321 	 *
1322 	 * The driver for these cards can deliver a unique 16-bit
1323 	 * identifier for each instance sharing the same card.  It is
1324 	 * placed instead of 0xFFFE in the interface identifier.  The
1325 	 * "u" bit of the interface identifier is not inverted in this
1326 	 * case.  Hence the resulting interface identifier has local
1327 	 * scope according to RFC2373.
1328 	 */
1329 	if (dev->dev_id) {
1330 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1331 		eui[4] = dev->dev_id & 0xFF;
1332 	} else {
1333 		eui[3] = 0xFF;
1334 		eui[4] = 0xFE;
1335 		eui[0] ^= 2;
1336 	}
1337 	return 0;
1338 }
1339 
1340 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1341 {
1342 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1343 	if (dev->addr_len != ARCNET_ALEN)
1344 		return -1;
1345 	memset(eui, 0, 7);
1346 	eui[7] = *(u8*)dev->dev_addr;
1347 	return 0;
1348 }
1349 
1350 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1351 {
1352 	if (dev->addr_len != INFINIBAND_ALEN)
1353 		return -1;
1354 	memcpy(eui, dev->dev_addr + 12, 8);
1355 	eui[0] |= 2;
1356 	return 0;
1357 }
1358 
1359 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1360 {
1361 	switch (dev->type) {
1362 	case ARPHRD_ETHER:
1363 	case ARPHRD_FDDI:
1364 	case ARPHRD_IEEE802_TR:
1365 		return addrconf_ifid_eui48(eui, dev);
1366 	case ARPHRD_ARCNET:
1367 		return addrconf_ifid_arcnet(eui, dev);
1368 	case ARPHRD_INFINIBAND:
1369 		return addrconf_ifid_infiniband(eui, dev);
1370 	}
1371 	return -1;
1372 }
1373 
1374 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1375 {
1376 	int err = -1;
1377 	struct inet6_ifaddr *ifp;
1378 
1379 	read_lock_bh(&idev->lock);
1380 	for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1381 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1382 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1383 			err = 0;
1384 			break;
1385 		}
1386 	}
1387 	read_unlock_bh(&idev->lock);
1388 	return err;
1389 }
1390 
1391 #ifdef CONFIG_IPV6_PRIVACY
1392 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1393 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1394 {
1395 regen:
1396 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1397 	idev->rndid[0] &= ~0x02;
1398 
1399 	/*
1400 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1401 	 * check if generated address is not inappropriate
1402 	 *
1403 	 *  - Reserved subnet anycast (RFC 2526)
1404 	 *	11111101 11....11 1xxxxxxx
1405 	 *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1406 	 *	00-00-5E-FE-xx-xx-xx-xx
1407 	 *  - value 0
1408 	 *  - XXX: already assigned to an address on the device
1409 	 */
1410 	if (idev->rndid[0] == 0xfd &&
1411 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1412 	    (idev->rndid[7]&0x80))
1413 		goto regen;
1414 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1415 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1416 			goto regen;
1417 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1418 			goto regen;
1419 	}
1420 
1421 	return 0;
1422 }
1423 
1424 static void ipv6_regen_rndid(unsigned long data)
1425 {
1426 	struct inet6_dev *idev = (struct inet6_dev *) data;
1427 	unsigned long expires;
1428 
1429 	rcu_read_lock_bh();
1430 	write_lock_bh(&idev->lock);
1431 
1432 	if (idev->dead)
1433 		goto out;
1434 
1435 	if (__ipv6_regen_rndid(idev) < 0)
1436 		goto out;
1437 
1438 	expires = jiffies +
1439 		idev->cnf.temp_prefered_lft * HZ -
1440 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1441 	if (time_before(expires, jiffies)) {
1442 		printk(KERN_WARNING
1443 			"ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1444 			idev->dev->name);
1445 		goto out;
1446 	}
1447 
1448 	if (!mod_timer(&idev->regen_timer, expires))
1449 		in6_dev_hold(idev);
1450 
1451 out:
1452 	write_unlock_bh(&idev->lock);
1453 	rcu_read_unlock_bh();
1454 	in6_dev_put(idev);
1455 }
1456 
1457 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1458 	int ret = 0;
1459 
1460 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1461 		ret = __ipv6_regen_rndid(idev);
1462 	return ret;
1463 }
1464 #endif
1465 
1466 /*
1467  *	Add prefix route.
1468  */
1469 
1470 static void
1471 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1472 		      unsigned long expires, u32 flags)
1473 {
1474 	struct fib6_config cfg = {
1475 		.fc_table = RT6_TABLE_PREFIX,
1476 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1477 		.fc_ifindex = dev->ifindex,
1478 		.fc_expires = expires,
1479 		.fc_dst_len = plen,
1480 		.fc_flags = RTF_UP | flags,
1481 	};
1482 
1483 	ipv6_addr_copy(&cfg.fc_dst, pfx);
1484 
1485 	/* Prevent useless cloning on PtP SIT.
1486 	   This thing is done here expecting that the whole
1487 	   class of non-broadcast devices need not cloning.
1488 	 */
1489 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1490 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1491 		cfg.fc_flags |= RTF_NONEXTHOP;
1492 #endif
1493 
1494 	ip6_route_add(&cfg);
1495 }
1496 
1497 /* Create "default" multicast route to the interface */
1498 
1499 static void addrconf_add_mroute(struct net_device *dev)
1500 {
1501 	struct fib6_config cfg = {
1502 		.fc_table = RT6_TABLE_LOCAL,
1503 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1504 		.fc_ifindex = dev->ifindex,
1505 		.fc_dst_len = 8,
1506 		.fc_flags = RTF_UP,
1507 	};
1508 
1509 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1510 
1511 	ip6_route_add(&cfg);
1512 }
1513 
1514 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1515 static void sit_route_add(struct net_device *dev)
1516 {
1517 	struct fib6_config cfg = {
1518 		.fc_table = RT6_TABLE_MAIN,
1519 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1520 		.fc_ifindex = dev->ifindex,
1521 		.fc_dst_len = 96,
1522 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
1523 	};
1524 
1525 	/* prefix length - 96 bits "::d.d.d.d" */
1526 	ip6_route_add(&cfg);
1527 }
1528 #endif
1529 
1530 static void addrconf_add_lroute(struct net_device *dev)
1531 {
1532 	struct in6_addr addr;
1533 
1534 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1535 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
1536 }
1537 
1538 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1539 {
1540 	struct inet6_dev *idev;
1541 
1542 	ASSERT_RTNL();
1543 
1544 	if ((idev = ipv6_find_idev(dev)) == NULL)
1545 		return NULL;
1546 
1547 	/* Add default multicast route */
1548 	addrconf_add_mroute(dev);
1549 
1550 	/* Add link local route */
1551 	addrconf_add_lroute(dev);
1552 	return idev;
1553 }
1554 
1555 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1556 {
1557 	struct prefix_info *pinfo;
1558 	__u32 valid_lft;
1559 	__u32 prefered_lft;
1560 	int addr_type;
1561 	unsigned long rt_expires;
1562 	struct inet6_dev *in6_dev;
1563 
1564 	pinfo = (struct prefix_info *) opt;
1565 
1566 	if (len < sizeof(struct prefix_info)) {
1567 		ADBG(("addrconf: prefix option too short\n"));
1568 		return;
1569 	}
1570 
1571 	/*
1572 	 *	Validation checks ([ADDRCONF], page 19)
1573 	 */
1574 
1575 	addr_type = ipv6_addr_type(&pinfo->prefix);
1576 
1577 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1578 		return;
1579 
1580 	valid_lft = ntohl(pinfo->valid);
1581 	prefered_lft = ntohl(pinfo->prefered);
1582 
1583 	if (prefered_lft > valid_lft) {
1584 		if (net_ratelimit())
1585 			printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1586 		return;
1587 	}
1588 
1589 	in6_dev = in6_dev_get(dev);
1590 
1591 	if (in6_dev == NULL) {
1592 		if (net_ratelimit())
1593 			printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1594 		return;
1595 	}
1596 
1597 	/*
1598 	 *	Two things going on here:
1599 	 *	1) Add routes for on-link prefixes
1600 	 *	2) Configure prefixes with the auto flag set
1601 	 */
1602 
1603 	/* Avoid arithmetic overflow. Really, we could
1604 	   save rt_expires in seconds, likely valid_lft,
1605 	   but it would require division in fib gc, that it
1606 	   not good.
1607 	 */
1608 	if (valid_lft >= 0x7FFFFFFF/HZ)
1609 		rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1610 	else
1611 		rt_expires = valid_lft * HZ;
1612 
1613 	/*
1614 	 * We convert this (in jiffies) to clock_t later.
1615 	 * Avoid arithmetic overflow there as well.
1616 	 * Overflow can happen only if HZ < USER_HZ.
1617 	 */
1618 	if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1619 		rt_expires = 0x7FFFFFFF / USER_HZ;
1620 
1621 	if (pinfo->onlink) {
1622 		struct rt6_info *rt;
1623 		rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1624 
1625 		if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1626 			if (rt->rt6i_flags&RTF_EXPIRES) {
1627 				if (valid_lft == 0) {
1628 					ip6_del_rt(rt);
1629 					rt = NULL;
1630 				} else {
1631 					rt->rt6i_expires = jiffies + rt_expires;
1632 				}
1633 			}
1634 		} else if (valid_lft) {
1635 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1636 					      dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1637 		}
1638 		if (rt)
1639 			dst_release(&rt->u.dst);
1640 	}
1641 
1642 	/* Try to figure out our local address for this prefix */
1643 
1644 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1645 		struct inet6_ifaddr * ifp;
1646 		struct in6_addr addr;
1647 		int create = 0, update_lft = 0;
1648 
1649 		if (pinfo->prefix_len == 64) {
1650 			memcpy(&addr, &pinfo->prefix, 8);
1651 			if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1652 			    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1653 				in6_dev_put(in6_dev);
1654 				return;
1655 			}
1656 			goto ok;
1657 		}
1658 		if (net_ratelimit())
1659 			printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1660 			       pinfo->prefix_len);
1661 		in6_dev_put(in6_dev);
1662 		return;
1663 
1664 ok:
1665 
1666 		ifp = ipv6_get_ifaddr(&addr, dev, 1);
1667 
1668 		if (ifp == NULL && valid_lft) {
1669 			int max_addresses = in6_dev->cnf.max_addresses;
1670 
1671 			/* Do not allow to create too much of autoconfigured
1672 			 * addresses; this would be too easy way to crash kernel.
1673 			 */
1674 			if (!max_addresses ||
1675 			    ipv6_count_addresses(in6_dev) < max_addresses)
1676 				ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1677 						    addr_type&IPV6_ADDR_SCOPE_MASK, 0);
1678 
1679 			if (!ifp || IS_ERR(ifp)) {
1680 				in6_dev_put(in6_dev);
1681 				return;
1682 			}
1683 
1684 			update_lft = create = 1;
1685 			ifp->cstamp = jiffies;
1686 			addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1687 		}
1688 
1689 		if (ifp) {
1690 			int flags;
1691 			unsigned long now;
1692 #ifdef CONFIG_IPV6_PRIVACY
1693 			struct inet6_ifaddr *ift;
1694 #endif
1695 			u32 stored_lft;
1696 
1697 			/* update lifetime (RFC2462 5.5.3 e) */
1698 			spin_lock(&ifp->lock);
1699 			now = jiffies;
1700 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1701 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1702 			else
1703 				stored_lft = 0;
1704 			if (!update_lft && stored_lft) {
1705 				if (valid_lft > MIN_VALID_LIFETIME ||
1706 				    valid_lft > stored_lft)
1707 					update_lft = 1;
1708 				else if (stored_lft <= MIN_VALID_LIFETIME) {
1709 					/* valid_lft <= stored_lft is always true */
1710 					/* XXX: IPsec */
1711 					update_lft = 0;
1712 				} else {
1713 					valid_lft = MIN_VALID_LIFETIME;
1714 					if (valid_lft < prefered_lft)
1715 						prefered_lft = valid_lft;
1716 					update_lft = 1;
1717 				}
1718 			}
1719 
1720 			if (update_lft) {
1721 				ifp->valid_lft = valid_lft;
1722 				ifp->prefered_lft = prefered_lft;
1723 				ifp->tstamp = now;
1724 				flags = ifp->flags;
1725 				ifp->flags &= ~IFA_F_DEPRECATED;
1726 				spin_unlock(&ifp->lock);
1727 
1728 				if (!(flags&IFA_F_TENTATIVE))
1729 					ipv6_ifa_notify(0, ifp);
1730 			} else
1731 				spin_unlock(&ifp->lock);
1732 
1733 #ifdef CONFIG_IPV6_PRIVACY
1734 			read_lock_bh(&in6_dev->lock);
1735 			/* update all temporary addresses in the list */
1736 			for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1737 				/*
1738 				 * When adjusting the lifetimes of an existing
1739 				 * temporary address, only lower the lifetimes.
1740 				 * Implementations must not increase the
1741 				 * lifetimes of an existing temporary address
1742 				 * when processing a Prefix Information Option.
1743 				 */
1744 				spin_lock(&ift->lock);
1745 				flags = ift->flags;
1746 				if (ift->valid_lft > valid_lft &&
1747 				    ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1748 					ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1749 				if (ift->prefered_lft > prefered_lft &&
1750 				    ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1751 					ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1752 				spin_unlock(&ift->lock);
1753 				if (!(flags&IFA_F_TENTATIVE))
1754 					ipv6_ifa_notify(0, ift);
1755 			}
1756 
1757 			if (create && in6_dev->cnf.use_tempaddr > 0) {
1758 				/*
1759 				 * When a new public address is created as described in [ADDRCONF],
1760 				 * also create a new temporary address.
1761 				 */
1762 				read_unlock_bh(&in6_dev->lock);
1763 				ipv6_create_tempaddr(ifp, NULL);
1764 			} else {
1765 				read_unlock_bh(&in6_dev->lock);
1766 			}
1767 #endif
1768 			in6_ifa_put(ifp);
1769 			addrconf_verify(0);
1770 		}
1771 	}
1772 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1773 	in6_dev_put(in6_dev);
1774 }
1775 
1776 /*
1777  *	Set destination address.
1778  *	Special case for SIT interfaces where we create a new "virtual"
1779  *	device.
1780  */
1781 int addrconf_set_dstaddr(void __user *arg)
1782 {
1783 	struct in6_ifreq ireq;
1784 	struct net_device *dev;
1785 	int err = -EINVAL;
1786 
1787 	rtnl_lock();
1788 
1789 	err = -EFAULT;
1790 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1791 		goto err_exit;
1792 
1793 	dev = __dev_get_by_index(ireq.ifr6_ifindex);
1794 
1795 	err = -ENODEV;
1796 	if (dev == NULL)
1797 		goto err_exit;
1798 
1799 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1800 	if (dev->type == ARPHRD_SIT) {
1801 		struct ifreq ifr;
1802 		mm_segment_t	oldfs;
1803 		struct ip_tunnel_parm p;
1804 
1805 		err = -EADDRNOTAVAIL;
1806 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1807 			goto err_exit;
1808 
1809 		memset(&p, 0, sizeof(p));
1810 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1811 		p.iph.saddr = 0;
1812 		p.iph.version = 4;
1813 		p.iph.ihl = 5;
1814 		p.iph.protocol = IPPROTO_IPV6;
1815 		p.iph.ttl = 64;
1816 		ifr.ifr_ifru.ifru_data = (void __user *)&p;
1817 
1818 		oldfs = get_fs(); set_fs(KERNEL_DS);
1819 		err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1820 		set_fs(oldfs);
1821 
1822 		if (err == 0) {
1823 			err = -ENOBUFS;
1824 			if ((dev = __dev_get_by_name(p.name)) == NULL)
1825 				goto err_exit;
1826 			err = dev_open(dev);
1827 		}
1828 	}
1829 #endif
1830 
1831 err_exit:
1832 	rtnl_unlock();
1833 	return err;
1834 }
1835 
1836 /*
1837  *	Manual configuration of address on an interface
1838  */
1839 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1840 			  __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1841 {
1842 	struct inet6_ifaddr *ifp;
1843 	struct inet6_dev *idev;
1844 	struct net_device *dev;
1845 	int scope;
1846 	u32 flags = RTF_EXPIRES;
1847 
1848 	ASSERT_RTNL();
1849 
1850 	/* check the lifetime */
1851 	if (!valid_lft || prefered_lft > valid_lft)
1852 		return -EINVAL;
1853 
1854 	if ((dev = __dev_get_by_index(ifindex)) == NULL)
1855 		return -ENODEV;
1856 
1857 	if ((idev = addrconf_add_dev(dev)) == NULL)
1858 		return -ENOBUFS;
1859 
1860 	scope = ipv6_addr_scope(pfx);
1861 
1862 	if (valid_lft == INFINITY_LIFE_TIME) {
1863 		ifa_flags |= IFA_F_PERMANENT;
1864 		flags = 0;
1865 	} else if (valid_lft >= 0x7FFFFFFF/HZ)
1866 		valid_lft = 0x7FFFFFFF/HZ;
1867 
1868 	if (prefered_lft == 0)
1869 		ifa_flags |= IFA_F_DEPRECATED;
1870 	else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1871 		 (prefered_lft != INFINITY_LIFE_TIME))
1872 		prefered_lft = 0x7FFFFFFF/HZ;
1873 
1874 	ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1875 
1876 	if (!IS_ERR(ifp)) {
1877 		spin_lock_bh(&ifp->lock);
1878 		ifp->valid_lft = valid_lft;
1879 		ifp->prefered_lft = prefered_lft;
1880 		ifp->tstamp = jiffies;
1881 		spin_unlock_bh(&ifp->lock);
1882 
1883 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1884 				      jiffies_to_clock_t(valid_lft * HZ), flags);
1885 		addrconf_dad_start(ifp, 0);
1886 		in6_ifa_put(ifp);
1887 		addrconf_verify(0);
1888 		return 0;
1889 	}
1890 
1891 	return PTR_ERR(ifp);
1892 }
1893 
1894 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1895 {
1896 	struct inet6_ifaddr *ifp;
1897 	struct inet6_dev *idev;
1898 	struct net_device *dev;
1899 
1900 	if ((dev = __dev_get_by_index(ifindex)) == NULL)
1901 		return -ENODEV;
1902 
1903 	if ((idev = __in6_dev_get(dev)) == NULL)
1904 		return -ENXIO;
1905 
1906 	read_lock_bh(&idev->lock);
1907 	for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1908 		if (ifp->prefix_len == plen &&
1909 		    ipv6_addr_equal(pfx, &ifp->addr)) {
1910 			in6_ifa_hold(ifp);
1911 			read_unlock_bh(&idev->lock);
1912 
1913 			ipv6_del_addr(ifp);
1914 
1915 			/* If the last address is deleted administratively,
1916 			   disable IPv6 on this interface.
1917 			 */
1918 			if (idev->addr_list == NULL)
1919 				addrconf_ifdown(idev->dev, 1);
1920 			return 0;
1921 		}
1922 	}
1923 	read_unlock_bh(&idev->lock);
1924 	return -EADDRNOTAVAIL;
1925 }
1926 
1927 
1928 int addrconf_add_ifaddr(void __user *arg)
1929 {
1930 	struct in6_ifreq ireq;
1931 	int err;
1932 
1933 	if (!capable(CAP_NET_ADMIN))
1934 		return -EPERM;
1935 
1936 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1937 		return -EFAULT;
1938 
1939 	rtnl_lock();
1940 	err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
1941 			     IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1942 	rtnl_unlock();
1943 	return err;
1944 }
1945 
1946 int addrconf_del_ifaddr(void __user *arg)
1947 {
1948 	struct in6_ifreq ireq;
1949 	int err;
1950 
1951 	if (!capable(CAP_NET_ADMIN))
1952 		return -EPERM;
1953 
1954 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1955 		return -EFAULT;
1956 
1957 	rtnl_lock();
1958 	err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
1959 	rtnl_unlock();
1960 	return err;
1961 }
1962 
1963 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1964 static void sit_add_v4_addrs(struct inet6_dev *idev)
1965 {
1966 	struct inet6_ifaddr * ifp;
1967 	struct in6_addr addr;
1968 	struct net_device *dev;
1969 	int scope;
1970 
1971 	ASSERT_RTNL();
1972 
1973 	memset(&addr, 0, sizeof(struct in6_addr));
1974 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
1975 
1976 	if (idev->dev->flags&IFF_POINTOPOINT) {
1977 		addr.s6_addr32[0] = htonl(0xfe800000);
1978 		scope = IFA_LINK;
1979 	} else {
1980 		scope = IPV6_ADDR_COMPATv4;
1981 	}
1982 
1983 	if (addr.s6_addr32[3]) {
1984 		ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
1985 		if (!IS_ERR(ifp)) {
1986 			spin_lock_bh(&ifp->lock);
1987 			ifp->flags &= ~IFA_F_TENTATIVE;
1988 			spin_unlock_bh(&ifp->lock);
1989 			ipv6_ifa_notify(RTM_NEWADDR, ifp);
1990 			in6_ifa_put(ifp);
1991 		}
1992 		return;
1993 	}
1994 
1995 	for (dev = dev_base; dev != NULL; dev = dev->next) {
1996 		struct in_device * in_dev = __in_dev_get_rtnl(dev);
1997 		if (in_dev && (dev->flags & IFF_UP)) {
1998 			struct in_ifaddr * ifa;
1999 
2000 			int flag = scope;
2001 
2002 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2003 				int plen;
2004 
2005 				addr.s6_addr32[3] = ifa->ifa_local;
2006 
2007 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2008 					continue;
2009 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2010 					if (idev->dev->flags&IFF_POINTOPOINT)
2011 						continue;
2012 					flag |= IFA_HOST;
2013 				}
2014 				if (idev->dev->flags&IFF_POINTOPOINT)
2015 					plen = 64;
2016 				else
2017 					plen = 96;
2018 
2019 				ifp = ipv6_add_addr(idev, &addr, plen, flag,
2020 						    IFA_F_PERMANENT);
2021 				if (!IS_ERR(ifp)) {
2022 					spin_lock_bh(&ifp->lock);
2023 					ifp->flags &= ~IFA_F_TENTATIVE;
2024 					spin_unlock_bh(&ifp->lock);
2025 					ipv6_ifa_notify(RTM_NEWADDR, ifp);
2026 					in6_ifa_put(ifp);
2027 				}
2028 			}
2029 		}
2030 	}
2031 }
2032 #endif
2033 
2034 static void init_loopback(struct net_device *dev)
2035 {
2036 	struct inet6_dev  *idev;
2037 	struct inet6_ifaddr * ifp;
2038 
2039 	/* ::1 */
2040 
2041 	ASSERT_RTNL();
2042 
2043 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2044 		printk(KERN_DEBUG "init loopback: add_dev failed\n");
2045 		return;
2046 	}
2047 
2048 	ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2049 	if (!IS_ERR(ifp)) {
2050 		spin_lock_bh(&ifp->lock);
2051 		ifp->flags &= ~IFA_F_TENTATIVE;
2052 		spin_unlock_bh(&ifp->lock);
2053 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2054 		in6_ifa_put(ifp);
2055 	}
2056 }
2057 
2058 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2059 {
2060 	struct inet6_ifaddr * ifp;
2061 
2062 	ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, IFA_F_PERMANENT);
2063 	if (!IS_ERR(ifp)) {
2064 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2065 		addrconf_dad_start(ifp, 0);
2066 		in6_ifa_put(ifp);
2067 	}
2068 }
2069 
2070 static void addrconf_dev_config(struct net_device *dev)
2071 {
2072 	struct in6_addr addr;
2073 	struct inet6_dev    * idev;
2074 
2075 	ASSERT_RTNL();
2076 
2077 	if ((dev->type != ARPHRD_ETHER) &&
2078 	    (dev->type != ARPHRD_FDDI) &&
2079 	    (dev->type != ARPHRD_IEEE802_TR) &&
2080 	    (dev->type != ARPHRD_ARCNET) &&
2081 	    (dev->type != ARPHRD_INFINIBAND)) {
2082 		/* Alas, we support only Ethernet autoconfiguration. */
2083 		return;
2084 	}
2085 
2086 	idev = addrconf_add_dev(dev);
2087 	if (idev == NULL)
2088 		return;
2089 
2090 	memset(&addr, 0, sizeof(struct in6_addr));
2091 	addr.s6_addr32[0] = htonl(0xFE800000);
2092 
2093 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2094 		addrconf_add_linklocal(idev, &addr);
2095 }
2096 
2097 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2098 static void addrconf_sit_config(struct net_device *dev)
2099 {
2100 	struct inet6_dev *idev;
2101 
2102 	ASSERT_RTNL();
2103 
2104 	/*
2105 	 * Configure the tunnel with one of our IPv4
2106 	 * addresses... we should configure all of
2107 	 * our v4 addrs in the tunnel
2108 	 */
2109 
2110 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2111 		printk(KERN_DEBUG "init sit: add_dev failed\n");
2112 		return;
2113 	}
2114 
2115 	sit_add_v4_addrs(idev);
2116 
2117 	if (dev->flags&IFF_POINTOPOINT) {
2118 		addrconf_add_mroute(dev);
2119 		addrconf_add_lroute(dev);
2120 	} else
2121 		sit_route_add(dev);
2122 }
2123 #endif
2124 
2125 static inline int
2126 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2127 {
2128 	struct in6_addr lladdr;
2129 
2130 	if (!ipv6_get_lladdr(link_dev, &lladdr)) {
2131 		addrconf_add_linklocal(idev, &lladdr);
2132 		return 0;
2133 	}
2134 	return -1;
2135 }
2136 
2137 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2138 {
2139 	struct net_device *link_dev;
2140 
2141 	/* first try to inherit the link-local address from the link device */
2142 	if (idev->dev->iflink &&
2143 	    (link_dev = __dev_get_by_index(idev->dev->iflink))) {
2144 		if (!ipv6_inherit_linklocal(idev, link_dev))
2145 			return;
2146 	}
2147 	/* then try to inherit it from any device */
2148 	for (link_dev = dev_base; link_dev; link_dev = link_dev->next) {
2149 		if (!ipv6_inherit_linklocal(idev, link_dev))
2150 			return;
2151 	}
2152 	printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2153 }
2154 
2155 /*
2156  * Autoconfigure tunnel with a link-local address so routing protocols,
2157  * DHCPv6, MLD etc. can be run over the virtual link
2158  */
2159 
2160 static void addrconf_ip6_tnl_config(struct net_device *dev)
2161 {
2162 	struct inet6_dev *idev;
2163 
2164 	ASSERT_RTNL();
2165 
2166 	if ((idev = addrconf_add_dev(dev)) == NULL) {
2167 		printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2168 		return;
2169 	}
2170 	ip6_tnl_add_linklocal(idev);
2171 }
2172 
2173 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2174 			   void * data)
2175 {
2176 	struct net_device *dev = (struct net_device *) data;
2177 	struct inet6_dev *idev = __in6_dev_get(dev);
2178 	int run_pending = 0;
2179 
2180 	switch(event) {
2181 	case NETDEV_REGISTER:
2182 		if (!idev) {
2183 			idev = ipv6_add_dev(dev);
2184 			if (!idev)
2185 				printk(KERN_WARNING "IPv6: add_dev failed for %s\n",
2186 					dev->name);
2187 		}
2188 		break;
2189 	case NETDEV_UP:
2190 	case NETDEV_CHANGE:
2191 		if (event == NETDEV_UP) {
2192 			if (!netif_carrier_ok(dev)) {
2193 				/* device is not ready yet. */
2194 				printk(KERN_INFO
2195 					"ADDRCONF(NETDEV_UP): %s: "
2196 					"link is not ready\n",
2197 					dev->name);
2198 				break;
2199 			}
2200 
2201 			if (idev)
2202 				idev->if_flags |= IF_READY;
2203 		} else {
2204 			if (!netif_carrier_ok(dev)) {
2205 				/* device is still not ready. */
2206 				break;
2207 			}
2208 
2209 			if (idev) {
2210 				if (idev->if_flags & IF_READY) {
2211 					/* device is already configured. */
2212 					break;
2213 				}
2214 				idev->if_flags |= IF_READY;
2215 			}
2216 
2217 			printk(KERN_INFO
2218 					"ADDRCONF(NETDEV_CHANGE): %s: "
2219 					"link becomes ready\n",
2220 					dev->name);
2221 
2222 			run_pending = 1;
2223 		}
2224 
2225 		switch(dev->type) {
2226 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2227 		case ARPHRD_SIT:
2228 			addrconf_sit_config(dev);
2229 			break;
2230 #endif
2231 		case ARPHRD_TUNNEL6:
2232 			addrconf_ip6_tnl_config(dev);
2233 			break;
2234 		case ARPHRD_LOOPBACK:
2235 			init_loopback(dev);
2236 			break;
2237 
2238 		default:
2239 			addrconf_dev_config(dev);
2240 			break;
2241 		};
2242 		if (idev) {
2243 			if (run_pending)
2244 				addrconf_dad_run(idev);
2245 
2246 			/* If the MTU changed during the interface down, when the
2247 			   interface up, the changed MTU must be reflected in the
2248 			   idev as well as routers.
2249 			 */
2250 			if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2251 				rt6_mtu_change(dev, dev->mtu);
2252 				idev->cnf.mtu6 = dev->mtu;
2253 			}
2254 			idev->tstamp = jiffies;
2255 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2256 			/* If the changed mtu during down is lower than IPV6_MIN_MTU
2257 			   stop IPv6 on this interface.
2258 			 */
2259 			if (dev->mtu < IPV6_MIN_MTU)
2260 				addrconf_ifdown(dev, event != NETDEV_DOWN);
2261 		}
2262 		break;
2263 
2264 	case NETDEV_CHANGEMTU:
2265 		if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2266 			rt6_mtu_change(dev, dev->mtu);
2267 			idev->cnf.mtu6 = dev->mtu;
2268 			break;
2269 		}
2270 
2271 		/* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2272 
2273 	case NETDEV_DOWN:
2274 	case NETDEV_UNREGISTER:
2275 		/*
2276 		 *	Remove all addresses from this interface.
2277 		 */
2278 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2279 		break;
2280 
2281 	case NETDEV_CHANGENAME:
2282 #ifdef CONFIG_SYSCTL
2283 		if (idev) {
2284 			addrconf_sysctl_unregister(&idev->cnf);
2285 			neigh_sysctl_unregister(idev->nd_parms);
2286 			neigh_sysctl_register(dev, idev->nd_parms,
2287 					      NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2288 					      &ndisc_ifinfo_sysctl_change,
2289 					      NULL);
2290 			addrconf_sysctl_register(idev, &idev->cnf);
2291 		}
2292 #endif
2293 		break;
2294 	};
2295 
2296 	return NOTIFY_OK;
2297 }
2298 
2299 /*
2300  *	addrconf module should be notified of a device going up
2301  */
2302 static struct notifier_block ipv6_dev_notf = {
2303 	.notifier_call = addrconf_notify,
2304 	.priority = 0
2305 };
2306 
2307 static int addrconf_ifdown(struct net_device *dev, int how)
2308 {
2309 	struct inet6_dev *idev;
2310 	struct inet6_ifaddr *ifa, **bifa;
2311 	int i;
2312 
2313 	ASSERT_RTNL();
2314 
2315 	if (dev == &loopback_dev && how == 1)
2316 		how = 0;
2317 
2318 	rt6_ifdown(dev);
2319 	neigh_ifdown(&nd_tbl, dev);
2320 
2321 	idev = __in6_dev_get(dev);
2322 	if (idev == NULL)
2323 		return -ENODEV;
2324 
2325 	/* Step 1: remove reference to ipv6 device from parent device.
2326 		   Do not dev_put!
2327 	 */
2328 	if (how == 1) {
2329 		idev->dead = 1;
2330 
2331 		/* protected by rtnl_lock */
2332 		rcu_assign_pointer(dev->ip6_ptr, NULL);
2333 
2334 		/* Step 1.5: remove snmp6 entry */
2335 		snmp6_unregister_dev(idev);
2336 
2337 	}
2338 
2339 	/* Step 2: clear hash table */
2340 	for (i=0; i<IN6_ADDR_HSIZE; i++) {
2341 		bifa = &inet6_addr_lst[i];
2342 
2343 		write_lock_bh(&addrconf_hash_lock);
2344 		while ((ifa = *bifa) != NULL) {
2345 			if (ifa->idev == idev) {
2346 				*bifa = ifa->lst_next;
2347 				ifa->lst_next = NULL;
2348 				addrconf_del_timer(ifa);
2349 				in6_ifa_put(ifa);
2350 				continue;
2351 			}
2352 			bifa = &ifa->lst_next;
2353 		}
2354 		write_unlock_bh(&addrconf_hash_lock);
2355 	}
2356 
2357 	write_lock_bh(&idev->lock);
2358 
2359 	/* Step 3: clear flags for stateless addrconf */
2360 	if (how != 1)
2361 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2362 
2363 	/* Step 4: clear address list */
2364 #ifdef CONFIG_IPV6_PRIVACY
2365 	if (how == 1 && del_timer(&idev->regen_timer))
2366 		in6_dev_put(idev);
2367 
2368 	/* clear tempaddr list */
2369 	while ((ifa = idev->tempaddr_list) != NULL) {
2370 		idev->tempaddr_list = ifa->tmp_next;
2371 		ifa->tmp_next = NULL;
2372 		ifa->dead = 1;
2373 		write_unlock_bh(&idev->lock);
2374 		spin_lock_bh(&ifa->lock);
2375 
2376 		if (ifa->ifpub) {
2377 			in6_ifa_put(ifa->ifpub);
2378 			ifa->ifpub = NULL;
2379 		}
2380 		spin_unlock_bh(&ifa->lock);
2381 		in6_ifa_put(ifa);
2382 		write_lock_bh(&idev->lock);
2383 	}
2384 #endif
2385 	while ((ifa = idev->addr_list) != NULL) {
2386 		idev->addr_list = ifa->if_next;
2387 		ifa->if_next = NULL;
2388 		ifa->dead = 1;
2389 		addrconf_del_timer(ifa);
2390 		write_unlock_bh(&idev->lock);
2391 
2392 		__ipv6_ifa_notify(RTM_DELADDR, ifa);
2393 		in6_ifa_put(ifa);
2394 
2395 		write_lock_bh(&idev->lock);
2396 	}
2397 	write_unlock_bh(&idev->lock);
2398 
2399 	/* Step 5: Discard multicast list */
2400 
2401 	if (how == 1)
2402 		ipv6_mc_destroy_dev(idev);
2403 	else
2404 		ipv6_mc_down(idev);
2405 
2406 	/* Step 5: netlink notification of this interface */
2407 	idev->tstamp = jiffies;
2408 	inet6_ifinfo_notify(RTM_DELLINK, idev);
2409 
2410 	/* Shot the device (if unregistered) */
2411 
2412 	if (how == 1) {
2413 #ifdef CONFIG_SYSCTL
2414 		addrconf_sysctl_unregister(&idev->cnf);
2415 		neigh_sysctl_unregister(idev->nd_parms);
2416 #endif
2417 		neigh_parms_release(&nd_tbl, idev->nd_parms);
2418 		neigh_ifdown(&nd_tbl, dev);
2419 		in6_dev_put(idev);
2420 	}
2421 	return 0;
2422 }
2423 
2424 static void addrconf_rs_timer(unsigned long data)
2425 {
2426 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2427 
2428 	if (ifp->idev->cnf.forwarding)
2429 		goto out;
2430 
2431 	if (ifp->idev->if_flags & IF_RA_RCVD) {
2432 		/*
2433 		 *	Announcement received after solicitation
2434 		 *	was sent
2435 		 */
2436 		goto out;
2437 	}
2438 
2439 	spin_lock(&ifp->lock);
2440 	if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2441 		struct in6_addr all_routers;
2442 
2443 		/* The wait after the last probe can be shorter */
2444 		addrconf_mod_timer(ifp, AC_RS,
2445 				   (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2446 				   ifp->idev->cnf.rtr_solicit_delay :
2447 				   ifp->idev->cnf.rtr_solicit_interval);
2448 		spin_unlock(&ifp->lock);
2449 
2450 		ipv6_addr_all_routers(&all_routers);
2451 
2452 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2453 	} else {
2454 		spin_unlock(&ifp->lock);
2455 		/*
2456 		 * Note: we do not support deprecated "all on-link"
2457 		 * assumption any longer.
2458 		 */
2459 		printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2460 		       ifp->idev->dev->name);
2461 	}
2462 
2463 out:
2464 	in6_ifa_put(ifp);
2465 }
2466 
2467 /*
2468  *	Duplicate Address Detection
2469  */
2470 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2471 {
2472 	unsigned long rand_num;
2473 	struct inet6_dev *idev = ifp->idev;
2474 
2475 	rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2476 	ifp->probes = idev->cnf.dad_transmits;
2477 	addrconf_mod_timer(ifp, AC_DAD, rand_num);
2478 }
2479 
2480 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2481 {
2482 	struct inet6_dev *idev = ifp->idev;
2483 	struct net_device *dev = idev->dev;
2484 
2485 	addrconf_join_solict(dev, &ifp->addr);
2486 
2487 	net_srandom(ifp->addr.s6_addr32[3]);
2488 
2489 	read_lock_bh(&idev->lock);
2490 	if (ifp->dead)
2491 		goto out;
2492 	spin_lock_bh(&ifp->lock);
2493 
2494 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2495 	    !(ifp->flags&IFA_F_TENTATIVE) ||
2496 	    ifp->flags & IFA_F_NODAD) {
2497 		ifp->flags &= ~IFA_F_TENTATIVE;
2498 		spin_unlock_bh(&ifp->lock);
2499 		read_unlock_bh(&idev->lock);
2500 
2501 		addrconf_dad_completed(ifp);
2502 		return;
2503 	}
2504 
2505 	if (!(idev->if_flags & IF_READY)) {
2506 		spin_unlock_bh(&ifp->lock);
2507 		read_unlock_bh(&idev->lock);
2508 		/*
2509 		 * If the defice is not ready:
2510 		 * - keep it tentative if it is a permanent address.
2511 		 * - otherwise, kill it.
2512 		 */
2513 		in6_ifa_hold(ifp);
2514 		addrconf_dad_stop(ifp);
2515 		return;
2516 	}
2517 	addrconf_dad_kick(ifp);
2518 	spin_unlock_bh(&ifp->lock);
2519 out:
2520 	read_unlock_bh(&idev->lock);
2521 }
2522 
2523 static void addrconf_dad_timer(unsigned long data)
2524 {
2525 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2526 	struct inet6_dev *idev = ifp->idev;
2527 	struct in6_addr unspec;
2528 	struct in6_addr mcaddr;
2529 
2530 	read_lock_bh(&idev->lock);
2531 	if (idev->dead) {
2532 		read_unlock_bh(&idev->lock);
2533 		goto out;
2534 	}
2535 	spin_lock_bh(&ifp->lock);
2536 	if (ifp->probes == 0) {
2537 		/*
2538 		 * DAD was successful
2539 		 */
2540 
2541 		ifp->flags &= ~IFA_F_TENTATIVE;
2542 		spin_unlock_bh(&ifp->lock);
2543 		read_unlock_bh(&idev->lock);
2544 
2545 		addrconf_dad_completed(ifp);
2546 
2547 		goto out;
2548 	}
2549 
2550 	ifp->probes--;
2551 	addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2552 	spin_unlock_bh(&ifp->lock);
2553 	read_unlock_bh(&idev->lock);
2554 
2555 	/* send a neighbour solicitation for our addr */
2556 	memset(&unspec, 0, sizeof(unspec));
2557 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2558 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2559 out:
2560 	in6_ifa_put(ifp);
2561 }
2562 
2563 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2564 {
2565 	struct net_device *	dev = ifp->idev->dev;
2566 
2567 	/*
2568 	 *	Configure the address for reception. Now it is valid.
2569 	 */
2570 
2571 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
2572 
2573 	/* If added prefix is link local and forwarding is off,
2574 	   start sending router solicitations.
2575 	 */
2576 
2577 	if (ifp->idev->cnf.forwarding == 0 &&
2578 	    ifp->idev->cnf.rtr_solicits > 0 &&
2579 	    (dev->flags&IFF_LOOPBACK) == 0 &&
2580 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2581 		struct in6_addr all_routers;
2582 
2583 		ipv6_addr_all_routers(&all_routers);
2584 
2585 		/*
2586 		 *	If a host as already performed a random delay
2587 		 *	[...] as part of DAD [...] there is no need
2588 		 *	to delay again before sending the first RS
2589 		 */
2590 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2591 
2592 		spin_lock_bh(&ifp->lock);
2593 		ifp->probes = 1;
2594 		ifp->idev->if_flags |= IF_RS_SENT;
2595 		addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2596 		spin_unlock_bh(&ifp->lock);
2597 	}
2598 }
2599 
2600 static void addrconf_dad_run(struct inet6_dev *idev) {
2601 	struct inet6_ifaddr *ifp;
2602 
2603 	read_lock_bh(&idev->lock);
2604 	for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2605 		spin_lock_bh(&ifp->lock);
2606 		if (!(ifp->flags & IFA_F_TENTATIVE)) {
2607 			spin_unlock_bh(&ifp->lock);
2608 			continue;
2609 		}
2610 		spin_unlock_bh(&ifp->lock);
2611 		addrconf_dad_kick(ifp);
2612 	}
2613 	read_unlock_bh(&idev->lock);
2614 }
2615 
2616 #ifdef CONFIG_PROC_FS
2617 struct if6_iter_state {
2618 	int bucket;
2619 };
2620 
2621 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2622 {
2623 	struct inet6_ifaddr *ifa = NULL;
2624 	struct if6_iter_state *state = seq->private;
2625 
2626 	for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2627 		ifa = inet6_addr_lst[state->bucket];
2628 		if (ifa)
2629 			break;
2630 	}
2631 	return ifa;
2632 }
2633 
2634 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2635 {
2636 	struct if6_iter_state *state = seq->private;
2637 
2638 	ifa = ifa->lst_next;
2639 try_again:
2640 	if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2641 		ifa = inet6_addr_lst[state->bucket];
2642 		goto try_again;
2643 	}
2644 	return ifa;
2645 }
2646 
2647 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2648 {
2649 	struct inet6_ifaddr *ifa = if6_get_first(seq);
2650 
2651 	if (ifa)
2652 		while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2653 			--pos;
2654 	return pos ? NULL : ifa;
2655 }
2656 
2657 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2658 {
2659 	read_lock_bh(&addrconf_hash_lock);
2660 	return if6_get_idx(seq, *pos);
2661 }
2662 
2663 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2664 {
2665 	struct inet6_ifaddr *ifa;
2666 
2667 	ifa = if6_get_next(seq, v);
2668 	++*pos;
2669 	return ifa;
2670 }
2671 
2672 static void if6_seq_stop(struct seq_file *seq, void *v)
2673 {
2674 	read_unlock_bh(&addrconf_hash_lock);
2675 }
2676 
2677 static int if6_seq_show(struct seq_file *seq, void *v)
2678 {
2679 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2680 	seq_printf(seq,
2681 		   NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2682 		   NIP6(ifp->addr),
2683 		   ifp->idev->dev->ifindex,
2684 		   ifp->prefix_len,
2685 		   ifp->scope,
2686 		   ifp->flags,
2687 		   ifp->idev->dev->name);
2688 	return 0;
2689 }
2690 
2691 static struct seq_operations if6_seq_ops = {
2692 	.start	= if6_seq_start,
2693 	.next	= if6_seq_next,
2694 	.show	= if6_seq_show,
2695 	.stop	= if6_seq_stop,
2696 };
2697 
2698 static int if6_seq_open(struct inode *inode, struct file *file)
2699 {
2700 	struct seq_file *seq;
2701 	int rc = -ENOMEM;
2702 	struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2703 
2704 	if (!s)
2705 		goto out;
2706 
2707 	rc = seq_open(file, &if6_seq_ops);
2708 	if (rc)
2709 		goto out_kfree;
2710 
2711 	seq = file->private_data;
2712 	seq->private = s;
2713 out:
2714 	return rc;
2715 out_kfree:
2716 	kfree(s);
2717 	goto out;
2718 }
2719 
2720 static const struct file_operations if6_fops = {
2721 	.owner		= THIS_MODULE,
2722 	.open		= if6_seq_open,
2723 	.read		= seq_read,
2724 	.llseek		= seq_lseek,
2725 	.release	= seq_release_private,
2726 };
2727 
2728 int __init if6_proc_init(void)
2729 {
2730 	if (!proc_net_fops_create("if_inet6", S_IRUGO, &if6_fops))
2731 		return -ENOMEM;
2732 	return 0;
2733 }
2734 
2735 void if6_proc_exit(void)
2736 {
2737 	proc_net_remove("if_inet6");
2738 }
2739 #endif	/* CONFIG_PROC_FS */
2740 
2741 #ifdef CONFIG_IPV6_MIP6
2742 /* Check if address is a home address configured on any interface. */
2743 int ipv6_chk_home_addr(struct in6_addr *addr)
2744 {
2745 	int ret = 0;
2746 	struct inet6_ifaddr * ifp;
2747 	u8 hash = ipv6_addr_hash(addr);
2748 	read_lock_bh(&addrconf_hash_lock);
2749 	for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2750 		if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2751 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
2752 			ret = 1;
2753 			break;
2754 		}
2755 	}
2756 	read_unlock_bh(&addrconf_hash_lock);
2757 	return ret;
2758 }
2759 #endif
2760 
2761 /*
2762  *	Periodic address status verification
2763  */
2764 
2765 static void addrconf_verify(unsigned long foo)
2766 {
2767 	struct inet6_ifaddr *ifp;
2768 	unsigned long now, next;
2769 	int i;
2770 
2771 	spin_lock_bh(&addrconf_verify_lock);
2772 	now = jiffies;
2773 	next = now + ADDR_CHECK_FREQUENCY;
2774 
2775 	del_timer(&addr_chk_timer);
2776 
2777 	for (i=0; i < IN6_ADDR_HSIZE; i++) {
2778 
2779 restart:
2780 		read_lock(&addrconf_hash_lock);
2781 		for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2782 			unsigned long age;
2783 #ifdef CONFIG_IPV6_PRIVACY
2784 			unsigned long regen_advance;
2785 #endif
2786 
2787 			if (ifp->flags & IFA_F_PERMANENT)
2788 				continue;
2789 
2790 			spin_lock(&ifp->lock);
2791 			age = (now - ifp->tstamp) / HZ;
2792 
2793 #ifdef CONFIG_IPV6_PRIVACY
2794 			regen_advance = ifp->idev->cnf.regen_max_retry *
2795 					ifp->idev->cnf.dad_transmits *
2796 					ifp->idev->nd_parms->retrans_time / HZ;
2797 #endif
2798 
2799 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2800 			    age >= ifp->valid_lft) {
2801 				spin_unlock(&ifp->lock);
2802 				in6_ifa_hold(ifp);
2803 				read_unlock(&addrconf_hash_lock);
2804 				ipv6_del_addr(ifp);
2805 				goto restart;
2806 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2807 				spin_unlock(&ifp->lock);
2808 				continue;
2809 			} else if (age >= ifp->prefered_lft) {
2810 				/* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2811 				int deprecate = 0;
2812 
2813 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
2814 					deprecate = 1;
2815 					ifp->flags |= IFA_F_DEPRECATED;
2816 				}
2817 
2818 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2819 					next = ifp->tstamp + ifp->valid_lft * HZ;
2820 
2821 				spin_unlock(&ifp->lock);
2822 
2823 				if (deprecate) {
2824 					in6_ifa_hold(ifp);
2825 					read_unlock(&addrconf_hash_lock);
2826 
2827 					ipv6_ifa_notify(0, ifp);
2828 					in6_ifa_put(ifp);
2829 					goto restart;
2830 				}
2831 #ifdef CONFIG_IPV6_PRIVACY
2832 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
2833 				   !(ifp->flags&IFA_F_TENTATIVE)) {
2834 				if (age >= ifp->prefered_lft - regen_advance) {
2835 					struct inet6_ifaddr *ifpub = ifp->ifpub;
2836 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2837 						next = ifp->tstamp + ifp->prefered_lft * HZ;
2838 					if (!ifp->regen_count && ifpub) {
2839 						ifp->regen_count++;
2840 						in6_ifa_hold(ifp);
2841 						in6_ifa_hold(ifpub);
2842 						spin_unlock(&ifp->lock);
2843 						read_unlock(&addrconf_hash_lock);
2844 						spin_lock(&ifpub->lock);
2845 						ifpub->regen_count = 0;
2846 						spin_unlock(&ifpub->lock);
2847 						ipv6_create_tempaddr(ifpub, ifp);
2848 						in6_ifa_put(ifpub);
2849 						in6_ifa_put(ifp);
2850 						goto restart;
2851 					}
2852 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2853 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2854 				spin_unlock(&ifp->lock);
2855 #endif
2856 			} else {
2857 				/* ifp->prefered_lft <= ifp->valid_lft */
2858 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2859 					next = ifp->tstamp + ifp->prefered_lft * HZ;
2860 				spin_unlock(&ifp->lock);
2861 			}
2862 		}
2863 		read_unlock(&addrconf_hash_lock);
2864 	}
2865 
2866 	addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2867 	add_timer(&addr_chk_timer);
2868 	spin_unlock_bh(&addrconf_verify_lock);
2869 }
2870 
2871 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2872 {
2873 	struct in6_addr *pfx = NULL;
2874 
2875 	if (addr)
2876 		pfx = nla_data(addr);
2877 
2878 	if (local) {
2879 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2880 			pfx = NULL;
2881 		else
2882 			pfx = nla_data(local);
2883 	}
2884 
2885 	return pfx;
2886 }
2887 
2888 static struct nla_policy ifa_ipv6_policy[IFA_MAX+1] __read_mostly = {
2889 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
2890 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
2891 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
2892 };
2893 
2894 static int
2895 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2896 {
2897 	struct ifaddrmsg *ifm;
2898 	struct nlattr *tb[IFA_MAX+1];
2899 	struct in6_addr *pfx;
2900 	int err;
2901 
2902 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2903 	if (err < 0)
2904 		return err;
2905 
2906 	ifm = nlmsg_data(nlh);
2907 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2908 	if (pfx == NULL)
2909 		return -EINVAL;
2910 
2911 	return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
2912 }
2913 
2914 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
2915 			     u32 prefered_lft, u32 valid_lft)
2916 {
2917 	u32 flags = RTF_EXPIRES;
2918 
2919 	if (!valid_lft || (prefered_lft > valid_lft))
2920 		return -EINVAL;
2921 
2922 	if (valid_lft == INFINITY_LIFE_TIME) {
2923 		ifa_flags |= IFA_F_PERMANENT;
2924 		flags = 0;
2925 	} else if (valid_lft >= 0x7FFFFFFF/HZ)
2926 		valid_lft = 0x7FFFFFFF/HZ;
2927 
2928 	if (prefered_lft == 0)
2929 		ifa_flags |= IFA_F_DEPRECATED;
2930 	else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2931 		 (prefered_lft != INFINITY_LIFE_TIME))
2932 		prefered_lft = 0x7FFFFFFF/HZ;
2933 
2934 	spin_lock_bh(&ifp->lock);
2935 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
2936 	ifp->tstamp = jiffies;
2937 	ifp->valid_lft = valid_lft;
2938 	ifp->prefered_lft = prefered_lft;
2939 
2940 	spin_unlock_bh(&ifp->lock);
2941 	if (!(ifp->flags&IFA_F_TENTATIVE))
2942 		ipv6_ifa_notify(0, ifp);
2943 
2944 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
2945 			      jiffies_to_clock_t(valid_lft * HZ), flags);
2946 	addrconf_verify(0);
2947 
2948 	return 0;
2949 }
2950 
2951 static int
2952 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2953 {
2954 	struct ifaddrmsg *ifm;
2955 	struct nlattr *tb[IFA_MAX+1];
2956 	struct in6_addr *pfx;
2957 	struct inet6_ifaddr *ifa;
2958 	struct net_device *dev;
2959 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
2960 	u8 ifa_flags;
2961 	int err;
2962 
2963 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2964 	if (err < 0)
2965 		return err;
2966 
2967 	ifm = nlmsg_data(nlh);
2968 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2969 	if (pfx == NULL)
2970 		return -EINVAL;
2971 
2972 	if (tb[IFA_CACHEINFO]) {
2973 		struct ifa_cacheinfo *ci;
2974 
2975 		ci = nla_data(tb[IFA_CACHEINFO]);
2976 		valid_lft = ci->ifa_valid;
2977 		preferred_lft = ci->ifa_prefered;
2978 	} else {
2979 		preferred_lft = INFINITY_LIFE_TIME;
2980 		valid_lft = INFINITY_LIFE_TIME;
2981 	}
2982 
2983 	dev =  __dev_get_by_index(ifm->ifa_index);
2984 	if (dev == NULL)
2985 		return -ENODEV;
2986 
2987 	/* We ignore other flags so far. */
2988 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
2989 
2990 	ifa = ipv6_get_ifaddr(pfx, dev, 1);
2991 	if (ifa == NULL) {
2992 		/*
2993 		 * It would be best to check for !NLM_F_CREATE here but
2994 		 * userspace alreay relies on not having to provide this.
2995 		 */
2996 		return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
2997 				      ifa_flags, preferred_lft, valid_lft);
2998 	}
2999 
3000 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3001 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3002 		err = -EEXIST;
3003 	else
3004 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3005 
3006 	in6_ifa_put(ifa);
3007 
3008 	return err;
3009 }
3010 
3011 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3012 			  u8 scope, int ifindex)
3013 {
3014 	struct ifaddrmsg *ifm;
3015 
3016 	ifm = nlmsg_data(nlh);
3017 	ifm->ifa_family = AF_INET6;
3018 	ifm->ifa_prefixlen = prefixlen;
3019 	ifm->ifa_flags = flags;
3020 	ifm->ifa_scope = scope;
3021 	ifm->ifa_index = ifindex;
3022 }
3023 
3024 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3025 			 unsigned long tstamp, u32 preferred, u32 valid)
3026 {
3027 	struct ifa_cacheinfo ci;
3028 
3029 	ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3030 			+ TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3031 	ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3032 			+ TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3033 	ci.ifa_prefered = preferred;
3034 	ci.ifa_valid = valid;
3035 
3036 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3037 }
3038 
3039 static inline int rt_scope(int ifa_scope)
3040 {
3041 	if (ifa_scope & IFA_HOST)
3042 		return RT_SCOPE_HOST;
3043 	else if (ifa_scope & IFA_LINK)
3044 		return RT_SCOPE_LINK;
3045 	else if (ifa_scope & IFA_SITE)
3046 		return RT_SCOPE_SITE;
3047 	else
3048 		return RT_SCOPE_UNIVERSE;
3049 }
3050 
3051 static inline int inet6_ifaddr_msgsize(void)
3052 {
3053 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3054 	       + nla_total_size(16) /* IFA_ADDRESS */
3055 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3056 }
3057 
3058 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3059 			     u32 pid, u32 seq, int event, unsigned int flags)
3060 {
3061 	struct nlmsghdr  *nlh;
3062 	u32 preferred, valid;
3063 
3064 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3065 	if (nlh == NULL)
3066 		return -EMSGSIZE;
3067 
3068 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3069 		      ifa->idev->dev->ifindex);
3070 
3071 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3072 		preferred = ifa->prefered_lft;
3073 		valid = ifa->valid_lft;
3074 		if (preferred != INFINITY_LIFE_TIME) {
3075 			long tval = (jiffies - ifa->tstamp)/HZ;
3076 			preferred -= tval;
3077 			if (valid != INFINITY_LIFE_TIME)
3078 				valid -= tval;
3079 		}
3080 	} else {
3081 		preferred = INFINITY_LIFE_TIME;
3082 		valid = INFINITY_LIFE_TIME;
3083 	}
3084 
3085 	if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3086 	    put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3087 		nlmsg_cancel(skb, nlh);
3088 		return -EMSGSIZE;
3089 	}
3090 
3091 	return nlmsg_end(skb, nlh);
3092 }
3093 
3094 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3095 				u32 pid, u32 seq, int event, u16 flags)
3096 {
3097 	struct nlmsghdr  *nlh;
3098 	u8 scope = RT_SCOPE_UNIVERSE;
3099 	int ifindex = ifmca->idev->dev->ifindex;
3100 
3101 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3102 		scope = RT_SCOPE_SITE;
3103 
3104 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3105 	if (nlh == NULL)
3106 		return -EMSGSIZE;
3107 
3108 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3109 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3110 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3111 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3112 		nlmsg_cancel(skb, nlh);
3113 		return -EMSGSIZE;
3114 	}
3115 
3116 	return nlmsg_end(skb, nlh);
3117 }
3118 
3119 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3120 				u32 pid, u32 seq, int event, unsigned int flags)
3121 {
3122 	struct nlmsghdr  *nlh;
3123 	u8 scope = RT_SCOPE_UNIVERSE;
3124 	int ifindex = ifaca->aca_idev->dev->ifindex;
3125 
3126 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3127 		scope = RT_SCOPE_SITE;
3128 
3129 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3130 	if (nlh == NULL)
3131 		return -EMSGSIZE;
3132 
3133 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3134 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3135 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3136 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3137 		nlmsg_cancel(skb, nlh);
3138 		return -EMSGSIZE;
3139 	}
3140 
3141 	return nlmsg_end(skb, nlh);
3142 }
3143 
3144 enum addr_type_t
3145 {
3146 	UNICAST_ADDR,
3147 	MULTICAST_ADDR,
3148 	ANYCAST_ADDR,
3149 };
3150 
3151 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3152 			   enum addr_type_t type)
3153 {
3154 	int idx, ip_idx;
3155 	int s_idx, s_ip_idx;
3156 	int err = 1;
3157 	struct net_device *dev;
3158 	struct inet6_dev *idev = NULL;
3159 	struct inet6_ifaddr *ifa;
3160 	struct ifmcaddr6 *ifmca;
3161 	struct ifacaddr6 *ifaca;
3162 
3163 	s_idx = cb->args[0];
3164 	s_ip_idx = ip_idx = cb->args[1];
3165 	read_lock(&dev_base_lock);
3166 
3167 	for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
3168 		if (idx < s_idx)
3169 			continue;
3170 		if (idx > s_idx)
3171 			s_ip_idx = 0;
3172 		ip_idx = 0;
3173 		if ((idev = in6_dev_get(dev)) == NULL)
3174 			continue;
3175 		read_lock_bh(&idev->lock);
3176 		switch (type) {
3177 		case UNICAST_ADDR:
3178 			/* unicast address incl. temp addr */
3179 			for (ifa = idev->addr_list; ifa;
3180 			     ifa = ifa->if_next, ip_idx++) {
3181 				if (ip_idx < s_ip_idx)
3182 					continue;
3183 				if ((err = inet6_fill_ifaddr(skb, ifa,
3184 				    NETLINK_CB(cb->skb).pid,
3185 				    cb->nlh->nlmsg_seq, RTM_NEWADDR,
3186 				    NLM_F_MULTI)) <= 0)
3187 					goto done;
3188 			}
3189 			break;
3190 		case MULTICAST_ADDR:
3191 			/* multicast address */
3192 			for (ifmca = idev->mc_list; ifmca;
3193 			     ifmca = ifmca->next, ip_idx++) {
3194 				if (ip_idx < s_ip_idx)
3195 					continue;
3196 				if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3197 				    NETLINK_CB(cb->skb).pid,
3198 				    cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3199 				    NLM_F_MULTI)) <= 0)
3200 					goto done;
3201 			}
3202 			break;
3203 		case ANYCAST_ADDR:
3204 			/* anycast address */
3205 			for (ifaca = idev->ac_list; ifaca;
3206 			     ifaca = ifaca->aca_next, ip_idx++) {
3207 				if (ip_idx < s_ip_idx)
3208 					continue;
3209 				if ((err = inet6_fill_ifacaddr(skb, ifaca,
3210 				    NETLINK_CB(cb->skb).pid,
3211 				    cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3212 				    NLM_F_MULTI)) <= 0)
3213 					goto done;
3214 			}
3215 			break;
3216 		default:
3217 			break;
3218 		}
3219 		read_unlock_bh(&idev->lock);
3220 		in6_dev_put(idev);
3221 	}
3222 done:
3223 	if (err <= 0) {
3224 		read_unlock_bh(&idev->lock);
3225 		in6_dev_put(idev);
3226 	}
3227 	read_unlock(&dev_base_lock);
3228 	cb->args[0] = idx;
3229 	cb->args[1] = ip_idx;
3230 	return skb->len;
3231 }
3232 
3233 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3234 {
3235 	enum addr_type_t type = UNICAST_ADDR;
3236 	return inet6_dump_addr(skb, cb, type);
3237 }
3238 
3239 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3240 {
3241 	enum addr_type_t type = MULTICAST_ADDR;
3242 	return inet6_dump_addr(skb, cb, type);
3243 }
3244 
3245 
3246 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3247 {
3248 	enum addr_type_t type = ANYCAST_ADDR;
3249 	return inet6_dump_addr(skb, cb, type);
3250 }
3251 
3252 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3253 			     void *arg)
3254 {
3255 	struct ifaddrmsg *ifm;
3256 	struct nlattr *tb[IFA_MAX+1];
3257 	struct in6_addr *addr = NULL;
3258 	struct net_device *dev = NULL;
3259 	struct inet6_ifaddr *ifa;
3260 	struct sk_buff *skb;
3261 	int err;
3262 
3263 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3264 	if (err < 0)
3265 		goto errout;
3266 
3267 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3268 	if (addr == NULL) {
3269 		err = -EINVAL;
3270 		goto errout;
3271 	}
3272 
3273 	ifm = nlmsg_data(nlh);
3274 	if (ifm->ifa_index)
3275 		dev = __dev_get_by_index(ifm->ifa_index);
3276 
3277 	if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3278 		err = -EADDRNOTAVAIL;
3279 		goto errout;
3280 	}
3281 
3282 	if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3283 		err = -ENOBUFS;
3284 		goto errout_ifa;
3285 	}
3286 
3287 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3288 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
3289 	if (err < 0) {
3290 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3291 		WARN_ON(err == -EMSGSIZE);
3292 		kfree_skb(skb);
3293 		goto errout_ifa;
3294 	}
3295 	err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3296 errout_ifa:
3297 	in6_ifa_put(ifa);
3298 errout:
3299 	return err;
3300 }
3301 
3302 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3303 {
3304 	struct sk_buff *skb;
3305 	int err = -ENOBUFS;
3306 
3307 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3308 	if (skb == NULL)
3309 		goto errout;
3310 
3311 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3312 	if (err < 0) {
3313 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3314 		WARN_ON(err == -EMSGSIZE);
3315 		kfree_skb(skb);
3316 		goto errout;
3317 	}
3318 	err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3319 errout:
3320 	if (err < 0)
3321 		rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3322 }
3323 
3324 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3325 				__s32 *array, int bytes)
3326 {
3327 	BUG_ON(bytes < (DEVCONF_MAX * 4));
3328 
3329 	memset(array, 0, bytes);
3330 	array[DEVCONF_FORWARDING] = cnf->forwarding;
3331 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3332 	array[DEVCONF_MTU6] = cnf->mtu6;
3333 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3334 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3335 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
3336 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3337 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3338 	array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3339 	array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3340 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3341 #ifdef CONFIG_IPV6_PRIVACY
3342 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3343 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3344 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3345 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3346 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3347 #endif
3348 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3349 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3350 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3351 #ifdef CONFIG_IPV6_ROUTER_PREF
3352 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3353 	array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3354 #ifdef CONFIG_IPV6_ROUTE_INFO
3355 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3356 #endif
3357 #endif
3358 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3359 }
3360 
3361 static inline size_t inet6_if_nlmsg_size(void)
3362 {
3363 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3364 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3365 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3366 	       + nla_total_size(4) /* IFLA_MTU */
3367 	       + nla_total_size(4) /* IFLA_LINK */
3368 	       + nla_total_size( /* IFLA_PROTINFO */
3369 			nla_total_size(4) /* IFLA_INET6_FLAGS */
3370 			+ nla_total_size(sizeof(struct ifla_cacheinfo))
3371 			+ nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3372 		 );
3373 }
3374 
3375 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3376 			     u32 pid, u32 seq, int event, unsigned int flags)
3377 {
3378 	struct net_device *dev = idev->dev;
3379 	struct nlattr *conf;
3380 	struct ifinfomsg *hdr;
3381 	struct nlmsghdr *nlh;
3382 	void *protoinfo;
3383 	struct ifla_cacheinfo ci;
3384 
3385 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3386 	if (nlh == NULL)
3387 		return -EMSGSIZE;
3388 
3389 	hdr = nlmsg_data(nlh);
3390 	hdr->ifi_family = AF_INET6;
3391 	hdr->__ifi_pad = 0;
3392 	hdr->ifi_type = dev->type;
3393 	hdr->ifi_index = dev->ifindex;
3394 	hdr->ifi_flags = dev_get_flags(dev);
3395 	hdr->ifi_change = 0;
3396 
3397 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3398 
3399 	if (dev->addr_len)
3400 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3401 
3402 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3403 	if (dev->ifindex != dev->iflink)
3404 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3405 
3406 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3407 	if (protoinfo == NULL)
3408 		goto nla_put_failure;
3409 
3410 	NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3411 
3412 	ci.max_reasm_len = IPV6_MAXPLEN;
3413 	ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3414 		    + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3415 	ci.reachable_time = idev->nd_parms->reachable_time;
3416 	ci.retrans_time = idev->nd_parms->retrans_time;
3417 	NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3418 
3419 	conf = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3420 	if (conf == NULL)
3421 		goto nla_put_failure;
3422 	ipv6_store_devconf(&idev->cnf, nla_data(conf), nla_len(conf));
3423 
3424 	/* XXX - Statistics/MC not implemented */
3425 
3426 	nla_nest_end(skb, protoinfo);
3427 	return nlmsg_end(skb, nlh);
3428 
3429 nla_put_failure:
3430 	nlmsg_cancel(skb, nlh);
3431 	return -EMSGSIZE;
3432 }
3433 
3434 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3435 {
3436 	int idx, err;
3437 	int s_idx = cb->args[0];
3438 	struct net_device *dev;
3439 	struct inet6_dev *idev;
3440 
3441 	read_lock(&dev_base_lock);
3442 	for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
3443 		if (idx < s_idx)
3444 			continue;
3445 		if ((idev = in6_dev_get(dev)) == NULL)
3446 			continue;
3447 		err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3448 				cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3449 		in6_dev_put(idev);
3450 		if (err <= 0)
3451 			break;
3452 	}
3453 	read_unlock(&dev_base_lock);
3454 	cb->args[0] = idx;
3455 
3456 	return skb->len;
3457 }
3458 
3459 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3460 {
3461 	struct sk_buff *skb;
3462 	int err = -ENOBUFS;
3463 
3464 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3465 	if (skb == NULL)
3466 		goto errout;
3467 
3468 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3469 	if (err < 0) {
3470 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3471 		WARN_ON(err == -EMSGSIZE);
3472 		kfree_skb(skb);
3473 		goto errout;
3474 	}
3475 	err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3476 errout:
3477 	if (err < 0)
3478 		rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3479 }
3480 
3481 static inline size_t inet6_prefix_nlmsg_size(void)
3482 {
3483 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
3484 	       + nla_total_size(sizeof(struct in6_addr))
3485 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
3486 }
3487 
3488 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3489 			     struct prefix_info *pinfo, u32 pid, u32 seq,
3490 			     int event, unsigned int flags)
3491 {
3492 	struct prefixmsg *pmsg;
3493 	struct nlmsghdr *nlh;
3494 	struct prefix_cacheinfo	ci;
3495 
3496 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3497 	if (nlh == NULL)
3498 		return -EMSGSIZE;
3499 
3500 	pmsg = nlmsg_data(nlh);
3501 	pmsg->prefix_family = AF_INET6;
3502 	pmsg->prefix_pad1 = 0;
3503 	pmsg->prefix_pad2 = 0;
3504 	pmsg->prefix_ifindex = idev->dev->ifindex;
3505 	pmsg->prefix_len = pinfo->prefix_len;
3506 	pmsg->prefix_type = pinfo->type;
3507 	pmsg->prefix_pad3 = 0;
3508 	pmsg->prefix_flags = 0;
3509 	if (pinfo->onlink)
3510 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3511 	if (pinfo->autoconf)
3512 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3513 
3514 	NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3515 
3516 	ci.preferred_time = ntohl(pinfo->prefered);
3517 	ci.valid_time = ntohl(pinfo->valid);
3518 	NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3519 
3520 	return nlmsg_end(skb, nlh);
3521 
3522 nla_put_failure:
3523 	nlmsg_cancel(skb, nlh);
3524 	return -EMSGSIZE;
3525 }
3526 
3527 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3528 			 struct prefix_info *pinfo)
3529 {
3530 	struct sk_buff *skb;
3531 	int err = -ENOBUFS;
3532 
3533 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3534 	if (skb == NULL)
3535 		goto errout;
3536 
3537 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3538 	if (err < 0) {
3539 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3540 		WARN_ON(err == -EMSGSIZE);
3541 		kfree_skb(skb);
3542 		goto errout;
3543 	}
3544 	err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3545 errout:
3546 	if (err < 0)
3547 		rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3548 }
3549 
3550 static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
3551 	[RTM_GETLINK - RTM_BASE] = { .dumpit	= inet6_dump_ifinfo, },
3552 	[RTM_NEWADDR - RTM_BASE] = { .doit	= inet6_rtm_newaddr, },
3553 	[RTM_DELADDR - RTM_BASE] = { .doit	= inet6_rtm_deladdr, },
3554 	[RTM_GETADDR - RTM_BASE] = { .doit	= inet6_rtm_getaddr,
3555 				     .dumpit	= inet6_dump_ifaddr, },
3556 	[RTM_GETMULTICAST - RTM_BASE] = { .dumpit = inet6_dump_ifmcaddr, },
3557 	[RTM_GETANYCAST - RTM_BASE] = { .dumpit	= inet6_dump_ifacaddr, },
3558 	[RTM_NEWROUTE - RTM_BASE] = { .doit	= inet6_rtm_newroute, },
3559 	[RTM_DELROUTE - RTM_BASE] = { .doit	= inet6_rtm_delroute, },
3560 	[RTM_GETROUTE - RTM_BASE] = { .doit	= inet6_rtm_getroute,
3561 				      .dumpit	= inet6_dump_fib, },
3562 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3563 	[RTM_GETRULE  - RTM_BASE] = { .dumpit   = fib6_rules_dump,   },
3564 #endif
3565 };
3566 
3567 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3568 {
3569 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3570 
3571 	switch (event) {
3572 	case RTM_NEWADDR:
3573 		ip6_ins_rt(ifp->rt);
3574 		if (ifp->idev->cnf.forwarding)
3575 			addrconf_join_anycast(ifp);
3576 		break;
3577 	case RTM_DELADDR:
3578 		if (ifp->idev->cnf.forwarding)
3579 			addrconf_leave_anycast(ifp);
3580 		addrconf_leave_solict(ifp->idev, &ifp->addr);
3581 		dst_hold(&ifp->rt->u.dst);
3582 		if (ip6_del_rt(ifp->rt))
3583 			dst_free(&ifp->rt->u.dst);
3584 		break;
3585 	}
3586 }
3587 
3588 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3589 {
3590 	rcu_read_lock_bh();
3591 	if (likely(ifp->idev->dead == 0))
3592 		__ipv6_ifa_notify(event, ifp);
3593 	rcu_read_unlock_bh();
3594 }
3595 
3596 #ifdef CONFIG_SYSCTL
3597 
3598 static
3599 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3600 			   void __user *buffer, size_t *lenp, loff_t *ppos)
3601 {
3602 	int *valp = ctl->data;
3603 	int val = *valp;
3604 	int ret;
3605 
3606 	ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3607 
3608 	if (write && valp != &ipv6_devconf_dflt.forwarding) {
3609 		if (valp != &ipv6_devconf.forwarding) {
3610 			if ((!*valp) ^ (!val)) {
3611 				struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3612 				if (idev == NULL)
3613 					return ret;
3614 				dev_forward_change(idev);
3615 			}
3616 		} else {
3617 			ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3618 			addrconf_forward_change();
3619 		}
3620 		if (*valp)
3621 			rt6_purge_dflt_routers();
3622 	}
3623 
3624 	return ret;
3625 }
3626 
3627 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3628 					    int __user *name, int nlen,
3629 					    void __user *oldval,
3630 					    size_t __user *oldlenp,
3631 					    void __user *newval, size_t newlen)
3632 {
3633 	int *valp = table->data;
3634 	int new;
3635 
3636 	if (!newval || !newlen)
3637 		return 0;
3638 	if (newlen != sizeof(int))
3639 		return -EINVAL;
3640 	if (get_user(new, (int __user *)newval))
3641 		return -EFAULT;
3642 	if (new == *valp)
3643 		return 0;
3644 	if (oldval && oldlenp) {
3645 		size_t len;
3646 		if (get_user(len, oldlenp))
3647 			return -EFAULT;
3648 		if (len) {
3649 			if (len > table->maxlen)
3650 				len = table->maxlen;
3651 			if (copy_to_user(oldval, valp, len))
3652 				return -EFAULT;
3653 			if (put_user(len, oldlenp))
3654 				return -EFAULT;
3655 		}
3656 	}
3657 
3658 	if (valp != &ipv6_devconf_dflt.forwarding) {
3659 		if (valp != &ipv6_devconf.forwarding) {
3660 			struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3661 			int changed;
3662 			if (unlikely(idev == NULL))
3663 				return -ENODEV;
3664 			changed = (!*valp) ^ (!new);
3665 			*valp = new;
3666 			if (changed)
3667 				dev_forward_change(idev);
3668 		} else {
3669 			*valp = new;
3670 			addrconf_forward_change();
3671 		}
3672 
3673 		if (*valp)
3674 			rt6_purge_dflt_routers();
3675 	} else
3676 		*valp = new;
3677 
3678 	return 1;
3679 }
3680 
3681 static struct addrconf_sysctl_table
3682 {
3683 	struct ctl_table_header *sysctl_header;
3684 	ctl_table addrconf_vars[__NET_IPV6_MAX];
3685 	ctl_table addrconf_dev[2];
3686 	ctl_table addrconf_conf_dir[2];
3687 	ctl_table addrconf_proto_dir[2];
3688 	ctl_table addrconf_root_dir[2];
3689 } addrconf_sysctl __read_mostly = {
3690 	.sysctl_header = NULL,
3691 	.addrconf_vars = {
3692 		{
3693 			.ctl_name	=	NET_IPV6_FORWARDING,
3694 			.procname	=	"forwarding",
3695 			.data		=	&ipv6_devconf.forwarding,
3696 			.maxlen		=	sizeof(int),
3697 			.mode		=	0644,
3698 			.proc_handler	=	&addrconf_sysctl_forward,
3699 			.strategy	=	&addrconf_sysctl_forward_strategy,
3700 		},
3701 		{
3702 			.ctl_name	=	NET_IPV6_HOP_LIMIT,
3703 			.procname	=	"hop_limit",
3704 			.data		=	&ipv6_devconf.hop_limit,
3705 			.maxlen		=	sizeof(int),
3706 			.mode		=	0644,
3707 			.proc_handler	=	proc_dointvec,
3708 		},
3709 		{
3710 			.ctl_name	=	NET_IPV6_MTU,
3711 			.procname	=	"mtu",
3712 			.data		=	&ipv6_devconf.mtu6,
3713 			.maxlen		=	sizeof(int),
3714 			.mode		=	0644,
3715 			.proc_handler	=	&proc_dointvec,
3716 		},
3717 		{
3718 			.ctl_name	=	NET_IPV6_ACCEPT_RA,
3719 			.procname	=	"accept_ra",
3720 			.data		=	&ipv6_devconf.accept_ra,
3721 			.maxlen		=	sizeof(int),
3722 			.mode		=	0644,
3723 			.proc_handler	=	&proc_dointvec,
3724 		},
3725 		{
3726 			.ctl_name	=	NET_IPV6_ACCEPT_REDIRECTS,
3727 			.procname	=	"accept_redirects",
3728 			.data		=	&ipv6_devconf.accept_redirects,
3729 			.maxlen		=	sizeof(int),
3730 			.mode		=	0644,
3731 			.proc_handler	=	&proc_dointvec,
3732 		},
3733 		{
3734 			.ctl_name	=	NET_IPV6_AUTOCONF,
3735 			.procname	=	"autoconf",
3736 			.data		=	&ipv6_devconf.autoconf,
3737 			.maxlen		=	sizeof(int),
3738 			.mode		=	0644,
3739 			.proc_handler	=	&proc_dointvec,
3740 		},
3741 		{
3742 			.ctl_name	=	NET_IPV6_DAD_TRANSMITS,
3743 			.procname	=	"dad_transmits",
3744 			.data		=	&ipv6_devconf.dad_transmits,
3745 			.maxlen		=	sizeof(int),
3746 			.mode		=	0644,
3747 			.proc_handler	=	&proc_dointvec,
3748 		},
3749 		{
3750 			.ctl_name	=	NET_IPV6_RTR_SOLICITS,
3751 			.procname	=	"router_solicitations",
3752 			.data		=	&ipv6_devconf.rtr_solicits,
3753 			.maxlen		=	sizeof(int),
3754 			.mode		=	0644,
3755 			.proc_handler	=	&proc_dointvec,
3756 		},
3757 		{
3758 			.ctl_name	=	NET_IPV6_RTR_SOLICIT_INTERVAL,
3759 			.procname	=	"router_solicitation_interval",
3760 			.data		=	&ipv6_devconf.rtr_solicit_interval,
3761 			.maxlen		=	sizeof(int),
3762 			.mode		=	0644,
3763 			.proc_handler	=	&proc_dointvec_jiffies,
3764 			.strategy	=	&sysctl_jiffies,
3765 		},
3766 		{
3767 			.ctl_name	=	NET_IPV6_RTR_SOLICIT_DELAY,
3768 			.procname	=	"router_solicitation_delay",
3769 			.data		=	&ipv6_devconf.rtr_solicit_delay,
3770 			.maxlen		=	sizeof(int),
3771 			.mode		=	0644,
3772 			.proc_handler	=	&proc_dointvec_jiffies,
3773 			.strategy	=	&sysctl_jiffies,
3774 		},
3775 		{
3776 			.ctl_name	=	NET_IPV6_FORCE_MLD_VERSION,
3777 			.procname	=	"force_mld_version",
3778 			.data		=	&ipv6_devconf.force_mld_version,
3779 			.maxlen		=	sizeof(int),
3780 			.mode		=	0644,
3781 			.proc_handler	=	&proc_dointvec,
3782 		},
3783 #ifdef CONFIG_IPV6_PRIVACY
3784 		{
3785 			.ctl_name	=	NET_IPV6_USE_TEMPADDR,
3786 			.procname	=	"use_tempaddr",
3787 			.data		=	&ipv6_devconf.use_tempaddr,
3788 			.maxlen		=	sizeof(int),
3789 			.mode		=	0644,
3790 			.proc_handler	=	&proc_dointvec,
3791 		},
3792 		{
3793 			.ctl_name	=	NET_IPV6_TEMP_VALID_LFT,
3794 			.procname	=	"temp_valid_lft",
3795 			.data		=	&ipv6_devconf.temp_valid_lft,
3796 			.maxlen		=	sizeof(int),
3797 			.mode		=	0644,
3798 			.proc_handler	=	&proc_dointvec,
3799 		},
3800 		{
3801 			.ctl_name	=	NET_IPV6_TEMP_PREFERED_LFT,
3802 			.procname	=	"temp_prefered_lft",
3803 			.data		=	&ipv6_devconf.temp_prefered_lft,
3804 			.maxlen		=	sizeof(int),
3805 			.mode		=	0644,
3806 			.proc_handler	=	&proc_dointvec,
3807 		},
3808 		{
3809 			.ctl_name	=	NET_IPV6_REGEN_MAX_RETRY,
3810 			.procname	=	"regen_max_retry",
3811 			.data		=	&ipv6_devconf.regen_max_retry,
3812 			.maxlen		=	sizeof(int),
3813 			.mode		=	0644,
3814 			.proc_handler	=	&proc_dointvec,
3815 		},
3816 		{
3817 			.ctl_name	=	NET_IPV6_MAX_DESYNC_FACTOR,
3818 			.procname	=	"max_desync_factor",
3819 			.data		=	&ipv6_devconf.max_desync_factor,
3820 			.maxlen		=	sizeof(int),
3821 			.mode		=	0644,
3822 			.proc_handler	=	&proc_dointvec,
3823 		},
3824 #endif
3825 		{
3826 			.ctl_name	=	NET_IPV6_MAX_ADDRESSES,
3827 			.procname	=	"max_addresses",
3828 			.data		=	&ipv6_devconf.max_addresses,
3829 			.maxlen		=	sizeof(int),
3830 			.mode		=	0644,
3831 			.proc_handler	=	&proc_dointvec,
3832 		},
3833 		{
3834 			.ctl_name	=	NET_IPV6_ACCEPT_RA_DEFRTR,
3835 			.procname	=	"accept_ra_defrtr",
3836 			.data		=	&ipv6_devconf.accept_ra_defrtr,
3837 			.maxlen		=	sizeof(int),
3838 			.mode		=	0644,
3839 			.proc_handler	=	&proc_dointvec,
3840 		},
3841 		{
3842 			.ctl_name	=	NET_IPV6_ACCEPT_RA_PINFO,
3843 			.procname	=	"accept_ra_pinfo",
3844 			.data		=	&ipv6_devconf.accept_ra_pinfo,
3845 			.maxlen		=	sizeof(int),
3846 			.mode		=	0644,
3847 			.proc_handler	=	&proc_dointvec,
3848 		},
3849 #ifdef CONFIG_IPV6_ROUTER_PREF
3850 		{
3851 			.ctl_name	=	NET_IPV6_ACCEPT_RA_RTR_PREF,
3852 			.procname	=	"accept_ra_rtr_pref",
3853 			.data		=	&ipv6_devconf.accept_ra_rtr_pref,
3854 			.maxlen		=	sizeof(int),
3855 			.mode		=	0644,
3856 			.proc_handler	=	&proc_dointvec,
3857 		},
3858 		{
3859 			.ctl_name	=	NET_IPV6_RTR_PROBE_INTERVAL,
3860 			.procname	=	"router_probe_interval",
3861 			.data		=	&ipv6_devconf.rtr_probe_interval,
3862 			.maxlen		=	sizeof(int),
3863 			.mode		=	0644,
3864 			.proc_handler	=	&proc_dointvec_jiffies,
3865 			.strategy	=	&sysctl_jiffies,
3866 		},
3867 #ifdef CONFIG_IPV6_ROUTE_INFO
3868 		{
3869 			.ctl_name	=	NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
3870 			.procname	=	"accept_ra_rt_info_max_plen",
3871 			.data		=	&ipv6_devconf.accept_ra_rt_info_max_plen,
3872 			.maxlen		=	sizeof(int),
3873 			.mode		=	0644,
3874 			.proc_handler	=	&proc_dointvec,
3875 		},
3876 #endif
3877 #endif
3878 		{
3879 			.ctl_name	=	NET_IPV6_PROXY_NDP,
3880 			.procname	=	"proxy_ndp",
3881 			.data		=	&ipv6_devconf.proxy_ndp,
3882 			.maxlen		=	sizeof(int),
3883 			.mode		=	0644,
3884 			.proc_handler	=	&proc_dointvec,
3885 		},
3886 		{
3887 			.ctl_name	=	0,	/* sentinel */
3888 		}
3889 	},
3890 	.addrconf_dev = {
3891 		{
3892 			.ctl_name	=	NET_PROTO_CONF_ALL,
3893 			.procname	=	"all",
3894 			.mode		=	0555,
3895 			.child		=	addrconf_sysctl.addrconf_vars,
3896 		},
3897 		{
3898 			.ctl_name	=	0,	/* sentinel */
3899 		}
3900 	},
3901 	.addrconf_conf_dir = {
3902 		{
3903 			.ctl_name	=	NET_IPV6_CONF,
3904 			.procname	=	"conf",
3905 			.mode		=	0555,
3906 			.child		=	addrconf_sysctl.addrconf_dev,
3907 		},
3908 		{
3909 			.ctl_name	=	0,	/* sentinel */
3910 		}
3911 	},
3912 	.addrconf_proto_dir = {
3913 		{
3914 			.ctl_name	=	NET_IPV6,
3915 			.procname	=	"ipv6",
3916 			.mode		=	0555,
3917 			.child		=	addrconf_sysctl.addrconf_conf_dir,
3918 		},
3919 		{
3920 			.ctl_name	=	0,	/* sentinel */
3921 		}
3922 	},
3923 	.addrconf_root_dir = {
3924 		{
3925 			.ctl_name	=	CTL_NET,
3926 			.procname	=	"net",
3927 			.mode		=	0555,
3928 			.child		=	addrconf_sysctl.addrconf_proto_dir,
3929 		},
3930 		{
3931 			.ctl_name	=	0,	/* sentinel */
3932 		}
3933 	},
3934 };
3935 
3936 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
3937 {
3938 	int i;
3939 	struct net_device *dev = idev ? idev->dev : NULL;
3940 	struct addrconf_sysctl_table *t;
3941 	char *dev_name = NULL;
3942 
3943 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
3944 	if (t == NULL)
3945 		return;
3946 	for (i=0; t->addrconf_vars[i].data; i++) {
3947 		t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
3948 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
3949 	}
3950 	if (dev) {
3951 		dev_name = dev->name;
3952 		t->addrconf_dev[0].ctl_name = dev->ifindex;
3953 	} else {
3954 		dev_name = "default";
3955 		t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
3956 	}
3957 
3958 	/*
3959 	 * Make a copy of dev_name, because '.procname' is regarded as const
3960 	 * by sysctl and we wouldn't want anyone to change it under our feet
3961 	 * (see SIOCSIFNAME).
3962 	 */
3963 	dev_name = kstrdup(dev_name, GFP_KERNEL);
3964 	if (!dev_name)
3965 	    goto free;
3966 
3967 	t->addrconf_dev[0].procname = dev_name;
3968 
3969 	t->addrconf_dev[0].child = t->addrconf_vars;
3970 	t->addrconf_conf_dir[0].child = t->addrconf_dev;
3971 	t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
3972 	t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
3973 
3974 	t->sysctl_header = register_sysctl_table(t->addrconf_root_dir);
3975 	if (t->sysctl_header == NULL)
3976 		goto free_procname;
3977 	else
3978 		p->sysctl = t;
3979 	return;
3980 
3981 	/* error path */
3982  free_procname:
3983 	kfree(dev_name);
3984  free:
3985 	kfree(t);
3986 
3987 	return;
3988 }
3989 
3990 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
3991 {
3992 	if (p->sysctl) {
3993 		struct addrconf_sysctl_table *t = p->sysctl;
3994 		p->sysctl = NULL;
3995 		unregister_sysctl_table(t->sysctl_header);
3996 		kfree(t->addrconf_dev[0].procname);
3997 		kfree(t);
3998 	}
3999 }
4000 
4001 
4002 #endif
4003 
4004 /*
4005  *      Device notifier
4006  */
4007 
4008 int register_inet6addr_notifier(struct notifier_block *nb)
4009 {
4010 	return atomic_notifier_chain_register(&inet6addr_chain, nb);
4011 }
4012 
4013 int unregister_inet6addr_notifier(struct notifier_block *nb)
4014 {
4015 	return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4016 }
4017 
4018 /*
4019  *	Init / cleanup code
4020  */
4021 
4022 int __init addrconf_init(void)
4023 {
4024 	int err = 0;
4025 
4026 	/* The addrconf netdev notifier requires that loopback_dev
4027 	 * has it's ipv6 private information allocated and setup
4028 	 * before it can bring up and give link-local addresses
4029 	 * to other devices which are up.
4030 	 *
4031 	 * Unfortunately, loopback_dev is not necessarily the first
4032 	 * entry in the global dev_base list of net devices.  In fact,
4033 	 * it is likely to be the very last entry on that list.
4034 	 * So this causes the notifier registry below to try and
4035 	 * give link-local addresses to all devices besides loopback_dev
4036 	 * first, then loopback_dev, which cases all the non-loopback_dev
4037 	 * devices to fail to get a link-local address.
4038 	 *
4039 	 * So, as a temporary fix, allocate the ipv6 structure for
4040 	 * loopback_dev first by hand.
4041 	 * Longer term, all of the dependencies ipv6 has upon the loopback
4042 	 * device and it being up should be removed.
4043 	 */
4044 	rtnl_lock();
4045 	if (!ipv6_add_dev(&loopback_dev))
4046 		err = -ENOMEM;
4047 	rtnl_unlock();
4048 	if (err)
4049 		return err;
4050 
4051 	ip6_null_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4052 
4053 	register_netdevice_notifier(&ipv6_dev_notf);
4054 
4055 	addrconf_verify(0);
4056 	rtnetlink_links[PF_INET6] = inet6_rtnetlink_table;
4057 #ifdef CONFIG_SYSCTL
4058 	addrconf_sysctl.sysctl_header =
4059 		register_sysctl_table(addrconf_sysctl.addrconf_root_dir);
4060 	addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4061 #endif
4062 
4063 	return 0;
4064 }
4065 
4066 void __exit addrconf_cleanup(void)
4067 {
4068 	struct net_device *dev;
4069 	struct inet6_dev *idev;
4070 	struct inet6_ifaddr *ifa;
4071 	int i;
4072 
4073 	unregister_netdevice_notifier(&ipv6_dev_notf);
4074 
4075 	rtnetlink_links[PF_INET6] = NULL;
4076 #ifdef CONFIG_SYSCTL
4077 	addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4078 	addrconf_sysctl_unregister(&ipv6_devconf);
4079 #endif
4080 
4081 	rtnl_lock();
4082 
4083 	/*
4084 	 *	clean dev list.
4085 	 */
4086 
4087 	for (dev=dev_base; dev; dev=dev->next) {
4088 		if ((idev = __in6_dev_get(dev)) == NULL)
4089 			continue;
4090 		addrconf_ifdown(dev, 1);
4091 	}
4092 	addrconf_ifdown(&loopback_dev, 2);
4093 
4094 	/*
4095 	 *	Check hash table.
4096 	 */
4097 
4098 	write_lock_bh(&addrconf_hash_lock);
4099 	for (i=0; i < IN6_ADDR_HSIZE; i++) {
4100 		for (ifa=inet6_addr_lst[i]; ifa; ) {
4101 			struct inet6_ifaddr *bifa;
4102 
4103 			bifa = ifa;
4104 			ifa = ifa->lst_next;
4105 			printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4106 			/* Do not free it; something is wrong.
4107 			   Now we can investigate it with debugger.
4108 			 */
4109 		}
4110 	}
4111 	write_unlock_bh(&addrconf_hash_lock);
4112 
4113 	del_timer(&addr_chk_timer);
4114 
4115 	rtnl_unlock();
4116 
4117 #ifdef CONFIG_PROC_FS
4118 	proc_net_remove("if_inet6");
4119 #endif
4120 }
4121