xref: /linux/net/ipv6/addrconf.c (revision d229807f669ba3dea9f64467ee965051c4366aed)
1 /*
2  *	IPv6 Address [auto]configuration
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14 
15 /*
16  *	Changes:
17  *
18  *	Janos Farkas			:	delete timer on ifdown
19  *	<chexum@bankinf.banki.hu>
20  *	Andi Kleen			:	kill double kfree on module
21  *						unload.
22  *	Maciej W. Rozycki		:	FDDI support
23  *	sekiya@USAGI			:	Don't send too many RS
24  *						packets.
25  *	yoshfuji@USAGI			:       Fixed interval between DAD
26  *						packets.
27  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
28  *						address validation timer.
29  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
30  *						support.
31  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
32  *						address on a same interface.
33  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
34  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
35  *						seq_file.
36  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
37  *						selection; consider scope,
38  *						status etc.
39  */
40 
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68 
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80 
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84 
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
87 
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90 #include <linux/export.h>
91 
92 /* Set to 3 to get tracing... */
93 #define ACONF_DEBUG 2
94 
95 #if ACONF_DEBUG >= 3
96 #define ADBG(x) printk x
97 #else
98 #define ADBG(x)
99 #endif
100 
101 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
102 
103 static inline u32 cstamp_delta(unsigned long cstamp)
104 {
105 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
106 }
107 
108 #define ADDRCONF_TIMER_FUZZ_MINUS	(HZ > 50 ? HZ/50 : 1)
109 #define ADDRCONF_TIMER_FUZZ		(HZ / 4)
110 #define ADDRCONF_TIMER_FUZZ_MAX		(HZ)
111 
112 #ifdef CONFIG_SYSCTL
113 static void addrconf_sysctl_register(struct inet6_dev *idev);
114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 #else
116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
117 {
118 }
119 
120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
121 {
122 }
123 #endif
124 
125 #ifdef CONFIG_IPV6_PRIVACY
126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
128 static void ipv6_regen_rndid(unsigned long data);
129 #endif
130 
131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
132 static int ipv6_count_addresses(struct inet6_dev *idev);
133 
134 /*
135  *	Configured unicast address hash table
136  */
137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
138 static DEFINE_SPINLOCK(addrconf_hash_lock);
139 
140 static void addrconf_verify(unsigned long);
141 
142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
143 static DEFINE_SPINLOCK(addrconf_verify_lock);
144 
145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
147 
148 static void addrconf_type_change(struct net_device *dev,
149 				 unsigned long event);
150 static int addrconf_ifdown(struct net_device *dev, int how);
151 
152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
153 static void addrconf_dad_timer(unsigned long data);
154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
155 static void addrconf_dad_run(struct inet6_dev *idev);
156 static void addrconf_rs_timer(unsigned long data);
157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
159 
160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
161 				struct prefix_info *pinfo);
162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
163 			       struct net_device *dev);
164 
165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
166 
167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168 	.forwarding		= 0,
169 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
170 	.mtu6			= IPV6_MIN_MTU,
171 	.accept_ra		= 1,
172 	.accept_redirects	= 1,
173 	.autoconf		= 1,
174 	.force_mld_version	= 0,
175 	.dad_transmits		= 1,
176 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
177 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
178 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
179 #ifdef CONFIG_IPV6_PRIVACY
180 	.use_tempaddr 		= 0,
181 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
182 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
183 	.regen_max_retry	= REGEN_MAX_RETRY,
184 	.max_desync_factor	= MAX_DESYNC_FACTOR,
185 #endif
186 	.max_addresses		= IPV6_MAX_ADDRESSES,
187 	.accept_ra_defrtr	= 1,
188 	.accept_ra_pinfo	= 1,
189 #ifdef CONFIG_IPV6_ROUTER_PREF
190 	.accept_ra_rtr_pref	= 1,
191 	.rtr_probe_interval	= 60 * HZ,
192 #ifdef CONFIG_IPV6_ROUTE_INFO
193 	.accept_ra_rt_info_max_plen = 0,
194 #endif
195 #endif
196 	.proxy_ndp		= 0,
197 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
198 	.disable_ipv6		= 0,
199 	.accept_dad		= 1,
200 };
201 
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 	.forwarding		= 0,
204 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
205 	.mtu6			= IPV6_MIN_MTU,
206 	.accept_ra		= 1,
207 	.accept_redirects	= 1,
208 	.autoconf		= 1,
209 	.dad_transmits		= 1,
210 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
211 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
212 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
213 #ifdef CONFIG_IPV6_PRIVACY
214 	.use_tempaddr		= 0,
215 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
216 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
217 	.regen_max_retry	= REGEN_MAX_RETRY,
218 	.max_desync_factor	= MAX_DESYNC_FACTOR,
219 #endif
220 	.max_addresses		= IPV6_MAX_ADDRESSES,
221 	.accept_ra_defrtr	= 1,
222 	.accept_ra_pinfo	= 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 	.accept_ra_rtr_pref	= 1,
225 	.rtr_probe_interval	= 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 	.accept_ra_rt_info_max_plen = 0,
228 #endif
229 #endif
230 	.proxy_ndp		= 0,
231 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
232 	.disable_ipv6		= 0,
233 	.accept_dad		= 1,
234 };
235 
236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
241 
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
244 {
245 	return !qdisc_tx_is_noop(dev);
246 }
247 
248 /* Check if a route is valid prefix route */
249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
250 {
251 	return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
252 }
253 
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
255 {
256 	if (del_timer(&ifp->timer))
257 		__in6_ifa_put(ifp);
258 }
259 
260 enum addrconf_timer_t {
261 	AC_NONE,
262 	AC_DAD,
263 	AC_RS,
264 };
265 
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267 			       enum addrconf_timer_t what,
268 			       unsigned long when)
269 {
270 	if (!del_timer(&ifp->timer))
271 		in6_ifa_hold(ifp);
272 
273 	switch (what) {
274 	case AC_DAD:
275 		ifp->timer.function = addrconf_dad_timer;
276 		break;
277 	case AC_RS:
278 		ifp->timer.function = addrconf_rs_timer;
279 		break;
280 	default:
281 		break;
282 	}
283 	ifp->timer.expires = jiffies + when;
284 	add_timer(&ifp->timer);
285 }
286 
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
288 {
289 	if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290 			  sizeof(struct ipstats_mib),
291 			  __alignof__(struct ipstats_mib)) < 0)
292 		goto err_ip;
293 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294 					GFP_KERNEL);
295 	if (!idev->stats.icmpv6dev)
296 		goto err_icmp;
297 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298 					   GFP_KERNEL);
299 	if (!idev->stats.icmpv6msgdev)
300 		goto err_icmpmsg;
301 
302 	return 0;
303 
304 err_icmpmsg:
305 	kfree(idev->stats.icmpv6dev);
306 err_icmp:
307 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 err_ip:
309 	return -ENOMEM;
310 }
311 
312 static void snmp6_free_dev(struct inet6_dev *idev)
313 {
314 	kfree(idev->stats.icmpv6msgdev);
315 	kfree(idev->stats.icmpv6dev);
316 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
317 }
318 
319 /* Nobody refers to this device, we may destroy it. */
320 
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
322 {
323 	struct net_device *dev = idev->dev;
324 
325 	WARN_ON(!list_empty(&idev->addr_list));
326 	WARN_ON(idev->mc_list != NULL);
327 
328 #ifdef NET_REFCNT_DEBUG
329 	printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
330 #endif
331 	dev_put(dev);
332 	if (!idev->dead) {
333 		pr_warning("Freeing alive inet6 device %p\n", idev);
334 		return;
335 	}
336 	snmp6_free_dev(idev);
337 	kfree_rcu(idev, rcu);
338 }
339 
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
341 
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
343 {
344 	struct inet6_dev *ndev;
345 
346 	ASSERT_RTNL();
347 
348 	if (dev->mtu < IPV6_MIN_MTU)
349 		return NULL;
350 
351 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
352 
353 	if (ndev == NULL)
354 		return NULL;
355 
356 	rwlock_init(&ndev->lock);
357 	ndev->dev = dev;
358 	INIT_LIST_HEAD(&ndev->addr_list);
359 
360 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361 	ndev->cnf.mtu6 = dev->mtu;
362 	ndev->cnf.sysctl = NULL;
363 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 	if (ndev->nd_parms == NULL) {
365 		kfree(ndev);
366 		return NULL;
367 	}
368 	if (ndev->cnf.forwarding)
369 		dev_disable_lro(dev);
370 	/* We refer to the device */
371 	dev_hold(dev);
372 
373 	if (snmp6_alloc_dev(ndev) < 0) {
374 		ADBG((KERN_WARNING
375 			"%s(): cannot allocate memory for statistics; dev=%s.\n",
376 			__func__, dev->name));
377 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 		dev_put(dev);
379 		kfree(ndev);
380 		return NULL;
381 	}
382 
383 	if (snmp6_register_dev(ndev) < 0) {
384 		ADBG((KERN_WARNING
385 			"%s(): cannot create /proc/net/dev_snmp6/%s\n",
386 			__func__, dev->name));
387 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
388 		ndev->dead = 1;
389 		in6_dev_finish_destroy(ndev);
390 		return NULL;
391 	}
392 
393 	/* One reference from device.  We must do this before
394 	 * we invoke __ipv6_regen_rndid().
395 	 */
396 	in6_dev_hold(ndev);
397 
398 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399 		ndev->cnf.accept_dad = -1;
400 
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
403 		printk(KERN_INFO
404 		       "%s: Disabled Multicast RS\n",
405 		       dev->name);
406 		ndev->cnf.rtr_solicits = 0;
407 	}
408 #endif
409 
410 #ifdef CONFIG_IPV6_PRIVACY
411 	INIT_LIST_HEAD(&ndev->tempaddr_list);
412 	setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 	if ((dev->flags&IFF_LOOPBACK) ||
414 	    dev->type == ARPHRD_TUNNEL ||
415 	    dev->type == ARPHRD_TUNNEL6 ||
416 	    dev->type == ARPHRD_SIT ||
417 	    dev->type == ARPHRD_NONE) {
418 		ndev->cnf.use_tempaddr = -1;
419 	} else {
420 		in6_dev_hold(ndev);
421 		ipv6_regen_rndid((unsigned long) ndev);
422 	}
423 #endif
424 
425 	if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 		ndev->if_flags |= IF_READY;
427 
428 	ipv6_mc_init_dev(ndev);
429 	ndev->tstamp = jiffies;
430 	addrconf_sysctl_register(ndev);
431 	/* protected by rtnl_lock */
432 	RCU_INIT_POINTER(dev->ip6_ptr, ndev);
433 
434 	/* Join all-node multicast group */
435 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
436 
437 	return ndev;
438 }
439 
440 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
441 {
442 	struct inet6_dev *idev;
443 
444 	ASSERT_RTNL();
445 
446 	idev = __in6_dev_get(dev);
447 	if (!idev) {
448 		idev = ipv6_add_dev(dev);
449 		if (!idev)
450 			return NULL;
451 	}
452 
453 	if (dev->flags&IFF_UP)
454 		ipv6_mc_up(idev);
455 	return idev;
456 }
457 
458 #ifdef CONFIG_SYSCTL
459 static void dev_forward_change(struct inet6_dev *idev)
460 {
461 	struct net_device *dev;
462 	struct inet6_ifaddr *ifa;
463 
464 	if (!idev)
465 		return;
466 	dev = idev->dev;
467 	if (idev->cnf.forwarding)
468 		dev_disable_lro(dev);
469 	if (dev && (dev->flags & IFF_MULTICAST)) {
470 		if (idev->cnf.forwarding)
471 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
472 		else
473 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
474 	}
475 
476 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
477 		if (ifa->flags&IFA_F_TENTATIVE)
478 			continue;
479 		if (idev->cnf.forwarding)
480 			addrconf_join_anycast(ifa);
481 		else
482 			addrconf_leave_anycast(ifa);
483 	}
484 }
485 
486 
487 static void addrconf_forward_change(struct net *net, __s32 newf)
488 {
489 	struct net_device *dev;
490 	struct inet6_dev *idev;
491 
492 	rcu_read_lock();
493 	for_each_netdev_rcu(net, dev) {
494 		idev = __in6_dev_get(dev);
495 		if (idev) {
496 			int changed = (!idev->cnf.forwarding) ^ (!newf);
497 			idev->cnf.forwarding = newf;
498 			if (changed)
499 				dev_forward_change(idev);
500 		}
501 	}
502 	rcu_read_unlock();
503 }
504 
505 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
506 {
507 	struct net *net;
508 
509 	net = (struct net *)table->extra2;
510 	if (p == &net->ipv6.devconf_dflt->forwarding)
511 		return 0;
512 
513 	if (!rtnl_trylock()) {
514 		/* Restore the original values before restarting */
515 		*p = old;
516 		return restart_syscall();
517 	}
518 
519 	if (p == &net->ipv6.devconf_all->forwarding) {
520 		__s32 newf = net->ipv6.devconf_all->forwarding;
521 		net->ipv6.devconf_dflt->forwarding = newf;
522 		addrconf_forward_change(net, newf);
523 	} else if ((!*p) ^ (!old))
524 		dev_forward_change((struct inet6_dev *)table->extra1);
525 	rtnl_unlock();
526 
527 	if (*p)
528 		rt6_purge_dflt_routers(net);
529 	return 1;
530 }
531 #endif
532 
533 /* Nobody refers to this ifaddr, destroy it */
534 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
535 {
536 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
537 
538 #ifdef NET_REFCNT_DEBUG
539 	printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
540 #endif
541 
542 	in6_dev_put(ifp->idev);
543 
544 	if (del_timer(&ifp->timer))
545 		pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
546 
547 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
548 		pr_warning("Freeing alive inet6 address %p\n", ifp);
549 		return;
550 	}
551 	dst_release(&ifp->rt->dst);
552 
553 	kfree_rcu(ifp, rcu);
554 }
555 
556 static void
557 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
558 {
559 	struct list_head *p;
560 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
561 
562 	/*
563 	 * Each device address list is sorted in order of scope -
564 	 * global before linklocal.
565 	 */
566 	list_for_each(p, &idev->addr_list) {
567 		struct inet6_ifaddr *ifa
568 			= list_entry(p, struct inet6_ifaddr, if_list);
569 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
570 			break;
571 	}
572 
573 	list_add_tail(&ifp->if_list, p);
574 }
575 
576 static u32 ipv6_addr_hash(const struct in6_addr *addr)
577 {
578 	/*
579 	 * We perform the hash function over the last 64 bits of the address
580 	 * This will include the IEEE address token on links that support it.
581 	 */
582 	return jhash_2words((__force u32)addr->s6_addr32[2],
583 			    (__force u32)addr->s6_addr32[3], 0)
584 		& (IN6_ADDR_HSIZE - 1);
585 }
586 
587 /* On success it returns ifp with increased reference count */
588 
589 static struct inet6_ifaddr *
590 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
591 	      int scope, u32 flags)
592 {
593 	struct inet6_ifaddr *ifa = NULL;
594 	struct rt6_info *rt;
595 	unsigned int hash;
596 	int err = 0;
597 	int addr_type = ipv6_addr_type(addr);
598 
599 	if (addr_type == IPV6_ADDR_ANY ||
600 	    addr_type & IPV6_ADDR_MULTICAST ||
601 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
602 	     addr_type & IPV6_ADDR_LOOPBACK))
603 		return ERR_PTR(-EADDRNOTAVAIL);
604 
605 	rcu_read_lock_bh();
606 	if (idev->dead) {
607 		err = -ENODEV;			/*XXX*/
608 		goto out2;
609 	}
610 
611 	if (idev->cnf.disable_ipv6) {
612 		err = -EACCES;
613 		goto out2;
614 	}
615 
616 	spin_lock(&addrconf_hash_lock);
617 
618 	/* Ignore adding duplicate addresses on an interface */
619 	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
620 		ADBG(("ipv6_add_addr: already assigned\n"));
621 		err = -EEXIST;
622 		goto out;
623 	}
624 
625 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
626 
627 	if (ifa == NULL) {
628 		ADBG(("ipv6_add_addr: malloc failed\n"));
629 		err = -ENOBUFS;
630 		goto out;
631 	}
632 
633 	rt = addrconf_dst_alloc(idev, addr, 0);
634 	if (IS_ERR(rt)) {
635 		err = PTR_ERR(rt);
636 		goto out;
637 	}
638 
639 	ipv6_addr_copy(&ifa->addr, addr);
640 
641 	spin_lock_init(&ifa->lock);
642 	spin_lock_init(&ifa->state_lock);
643 	init_timer(&ifa->timer);
644 	INIT_HLIST_NODE(&ifa->addr_lst);
645 	ifa->timer.data = (unsigned long) ifa;
646 	ifa->scope = scope;
647 	ifa->prefix_len = pfxlen;
648 	ifa->flags = flags | IFA_F_TENTATIVE;
649 	ifa->cstamp = ifa->tstamp = jiffies;
650 
651 	ifa->rt = rt;
652 
653 	/*
654 	 * part one of RFC 4429, section 3.3
655 	 * We should not configure an address as
656 	 * optimistic if we do not yet know the link
657 	 * layer address of our nexhop router
658 	 */
659 
660 	if (dst_get_neighbour_raw(&rt->dst) == NULL)
661 		ifa->flags &= ~IFA_F_OPTIMISTIC;
662 
663 	ifa->idev = idev;
664 	in6_dev_hold(idev);
665 	/* For caller */
666 	in6_ifa_hold(ifa);
667 
668 	/* Add to big hash table */
669 	hash = ipv6_addr_hash(addr);
670 
671 	hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
672 	spin_unlock(&addrconf_hash_lock);
673 
674 	write_lock(&idev->lock);
675 	/* Add to inet6_dev unicast addr list. */
676 	ipv6_link_dev_addr(idev, ifa);
677 
678 #ifdef CONFIG_IPV6_PRIVACY
679 	if (ifa->flags&IFA_F_TEMPORARY) {
680 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
681 		in6_ifa_hold(ifa);
682 	}
683 #endif
684 
685 	in6_ifa_hold(ifa);
686 	write_unlock(&idev->lock);
687 out2:
688 	rcu_read_unlock_bh();
689 
690 	if (likely(err == 0))
691 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
692 	else {
693 		kfree(ifa);
694 		ifa = ERR_PTR(err);
695 	}
696 
697 	return ifa;
698 out:
699 	spin_unlock(&addrconf_hash_lock);
700 	goto out2;
701 }
702 
703 /* This function wants to get referenced ifp and releases it before return */
704 
705 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
706 {
707 	struct inet6_ifaddr *ifa, *ifn;
708 	struct inet6_dev *idev = ifp->idev;
709 	int state;
710 	int deleted = 0, onlink = 0;
711 	unsigned long expires = jiffies;
712 
713 	spin_lock_bh(&ifp->state_lock);
714 	state = ifp->state;
715 	ifp->state = INET6_IFADDR_STATE_DEAD;
716 	spin_unlock_bh(&ifp->state_lock);
717 
718 	if (state == INET6_IFADDR_STATE_DEAD)
719 		goto out;
720 
721 	spin_lock_bh(&addrconf_hash_lock);
722 	hlist_del_init_rcu(&ifp->addr_lst);
723 	spin_unlock_bh(&addrconf_hash_lock);
724 
725 	write_lock_bh(&idev->lock);
726 #ifdef CONFIG_IPV6_PRIVACY
727 	if (ifp->flags&IFA_F_TEMPORARY) {
728 		list_del(&ifp->tmp_list);
729 		if (ifp->ifpub) {
730 			in6_ifa_put(ifp->ifpub);
731 			ifp->ifpub = NULL;
732 		}
733 		__in6_ifa_put(ifp);
734 	}
735 #endif
736 
737 	list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
738 		if (ifa == ifp) {
739 			list_del_init(&ifp->if_list);
740 			__in6_ifa_put(ifp);
741 
742 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
743 				break;
744 			deleted = 1;
745 			continue;
746 		} else if (ifp->flags & IFA_F_PERMANENT) {
747 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
748 					      ifp->prefix_len)) {
749 				if (ifa->flags & IFA_F_PERMANENT) {
750 					onlink = 1;
751 					if (deleted)
752 						break;
753 				} else {
754 					unsigned long lifetime;
755 
756 					if (!onlink)
757 						onlink = -1;
758 
759 					spin_lock(&ifa->lock);
760 
761 					lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
762 					/*
763 					 * Note: Because this address is
764 					 * not permanent, lifetime <
765 					 * LONG_MAX / HZ here.
766 					 */
767 					if (time_before(expires,
768 							ifa->tstamp + lifetime * HZ))
769 						expires = ifa->tstamp + lifetime * HZ;
770 					spin_unlock(&ifa->lock);
771 				}
772 			}
773 		}
774 	}
775 	write_unlock_bh(&idev->lock);
776 
777 	addrconf_del_timer(ifp);
778 
779 	ipv6_ifa_notify(RTM_DELADDR, ifp);
780 
781 	atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
782 
783 	/*
784 	 * Purge or update corresponding prefix
785 	 *
786 	 * 1) we don't purge prefix here if address was not permanent.
787 	 *    prefix is managed by its own lifetime.
788 	 * 2) if there're no addresses, delete prefix.
789 	 * 3) if there're still other permanent address(es),
790 	 *    corresponding prefix is still permanent.
791 	 * 4) otherwise, update prefix lifetime to the
792 	 *    longest valid lifetime among the corresponding
793 	 *    addresses on the device.
794 	 *    Note: subsequent RA will update lifetime.
795 	 *
796 	 * --yoshfuji
797 	 */
798 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
799 		struct in6_addr prefix;
800 		struct rt6_info *rt;
801 		struct net *net = dev_net(ifp->idev->dev);
802 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
803 		rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
804 
805 		if (rt && addrconf_is_prefix_route(rt)) {
806 			if (onlink == 0) {
807 				ip6_del_rt(rt);
808 				rt = NULL;
809 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
810 				rt->rt6i_expires = expires;
811 				rt->rt6i_flags |= RTF_EXPIRES;
812 			}
813 		}
814 		dst_release(&rt->dst);
815 	}
816 
817 	/* clean up prefsrc entries */
818 	rt6_remove_prefsrc(ifp);
819 out:
820 	in6_ifa_put(ifp);
821 }
822 
823 #ifdef CONFIG_IPV6_PRIVACY
824 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
825 {
826 	struct inet6_dev *idev = ifp->idev;
827 	struct in6_addr addr, *tmpaddr;
828 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
829 	unsigned long regen_advance;
830 	int tmp_plen;
831 	int ret = 0;
832 	int max_addresses;
833 	u32 addr_flags;
834 	unsigned long now = jiffies;
835 
836 	write_lock(&idev->lock);
837 	if (ift) {
838 		spin_lock_bh(&ift->lock);
839 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
840 		spin_unlock_bh(&ift->lock);
841 		tmpaddr = &addr;
842 	} else {
843 		tmpaddr = NULL;
844 	}
845 retry:
846 	in6_dev_hold(idev);
847 	if (idev->cnf.use_tempaddr <= 0) {
848 		write_unlock(&idev->lock);
849 		printk(KERN_INFO
850 			"ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
851 		in6_dev_put(idev);
852 		ret = -1;
853 		goto out;
854 	}
855 	spin_lock_bh(&ifp->lock);
856 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
857 		idev->cnf.use_tempaddr = -1;	/*XXX*/
858 		spin_unlock_bh(&ifp->lock);
859 		write_unlock(&idev->lock);
860 		printk(KERN_WARNING
861 			"ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
862 		in6_dev_put(idev);
863 		ret = -1;
864 		goto out;
865 	}
866 	in6_ifa_hold(ifp);
867 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
868 	if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
869 		spin_unlock_bh(&ifp->lock);
870 		write_unlock(&idev->lock);
871 		printk(KERN_WARNING
872 			"ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
873 		in6_ifa_put(ifp);
874 		in6_dev_put(idev);
875 		ret = -1;
876 		goto out;
877 	}
878 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
879 	age = (now - ifp->tstamp) / HZ;
880 	tmp_valid_lft = min_t(__u32,
881 			      ifp->valid_lft,
882 			      idev->cnf.temp_valid_lft + age);
883 	tmp_prefered_lft = min_t(__u32,
884 				 ifp->prefered_lft,
885 				 idev->cnf.temp_prefered_lft + age -
886 				 idev->cnf.max_desync_factor);
887 	tmp_plen = ifp->prefix_len;
888 	max_addresses = idev->cnf.max_addresses;
889 	tmp_tstamp = ifp->tstamp;
890 	spin_unlock_bh(&ifp->lock);
891 
892 	regen_advance = idev->cnf.regen_max_retry *
893 	                idev->cnf.dad_transmits *
894 	                idev->nd_parms->retrans_time / HZ;
895 	write_unlock(&idev->lock);
896 
897 	/* A temporary address is created only if this calculated Preferred
898 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
899 	 * an implementation must not create a temporary address with a zero
900 	 * Preferred Lifetime.
901 	 */
902 	if (tmp_prefered_lft <= regen_advance) {
903 		in6_ifa_put(ifp);
904 		in6_dev_put(idev);
905 		ret = -1;
906 		goto out;
907 	}
908 
909 	addr_flags = IFA_F_TEMPORARY;
910 	/* set in addrconf_prefix_rcv() */
911 	if (ifp->flags & IFA_F_OPTIMISTIC)
912 		addr_flags |= IFA_F_OPTIMISTIC;
913 
914 	ift = !max_addresses ||
915 	      ipv6_count_addresses(idev) < max_addresses ?
916 		ipv6_add_addr(idev, &addr, tmp_plen,
917 			      ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
918 			      addr_flags) : NULL;
919 	if (!ift || IS_ERR(ift)) {
920 		in6_ifa_put(ifp);
921 		in6_dev_put(idev);
922 		printk(KERN_INFO
923 			"ipv6_create_tempaddr(): retry temporary address regeneration.\n");
924 		tmpaddr = &addr;
925 		write_lock(&idev->lock);
926 		goto retry;
927 	}
928 
929 	spin_lock_bh(&ift->lock);
930 	ift->ifpub = ifp;
931 	ift->valid_lft = tmp_valid_lft;
932 	ift->prefered_lft = tmp_prefered_lft;
933 	ift->cstamp = now;
934 	ift->tstamp = tmp_tstamp;
935 	spin_unlock_bh(&ift->lock);
936 
937 	addrconf_dad_start(ift, 0);
938 	in6_ifa_put(ift);
939 	in6_dev_put(idev);
940 out:
941 	return ret;
942 }
943 #endif
944 
945 /*
946  *	Choose an appropriate source address (RFC3484)
947  */
948 enum {
949 	IPV6_SADDR_RULE_INIT = 0,
950 	IPV6_SADDR_RULE_LOCAL,
951 	IPV6_SADDR_RULE_SCOPE,
952 	IPV6_SADDR_RULE_PREFERRED,
953 #ifdef CONFIG_IPV6_MIP6
954 	IPV6_SADDR_RULE_HOA,
955 #endif
956 	IPV6_SADDR_RULE_OIF,
957 	IPV6_SADDR_RULE_LABEL,
958 #ifdef CONFIG_IPV6_PRIVACY
959 	IPV6_SADDR_RULE_PRIVACY,
960 #endif
961 	IPV6_SADDR_RULE_ORCHID,
962 	IPV6_SADDR_RULE_PREFIX,
963 	IPV6_SADDR_RULE_MAX
964 };
965 
966 struct ipv6_saddr_score {
967 	int			rule;
968 	int			addr_type;
969 	struct inet6_ifaddr	*ifa;
970 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
971 	int			scopedist;
972 	int			matchlen;
973 };
974 
975 struct ipv6_saddr_dst {
976 	const struct in6_addr *addr;
977 	int ifindex;
978 	int scope;
979 	int label;
980 	unsigned int prefs;
981 };
982 
983 static inline int ipv6_saddr_preferred(int type)
984 {
985 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
986 		return 1;
987 	return 0;
988 }
989 
990 static int ipv6_get_saddr_eval(struct net *net,
991 			       struct ipv6_saddr_score *score,
992 			       struct ipv6_saddr_dst *dst,
993 			       int i)
994 {
995 	int ret;
996 
997 	if (i <= score->rule) {
998 		switch (i) {
999 		case IPV6_SADDR_RULE_SCOPE:
1000 			ret = score->scopedist;
1001 			break;
1002 		case IPV6_SADDR_RULE_PREFIX:
1003 			ret = score->matchlen;
1004 			break;
1005 		default:
1006 			ret = !!test_bit(i, score->scorebits);
1007 		}
1008 		goto out;
1009 	}
1010 
1011 	switch (i) {
1012 	case IPV6_SADDR_RULE_INIT:
1013 		/* Rule 0: remember if hiscore is not ready yet */
1014 		ret = !!score->ifa;
1015 		break;
1016 	case IPV6_SADDR_RULE_LOCAL:
1017 		/* Rule 1: Prefer same address */
1018 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1019 		break;
1020 	case IPV6_SADDR_RULE_SCOPE:
1021 		/* Rule 2: Prefer appropriate scope
1022 		 *
1023 		 *      ret
1024 		 *       ^
1025 		 *    -1 |  d 15
1026 		 *    ---+--+-+---> scope
1027 		 *       |
1028 		 *       |             d is scope of the destination.
1029 		 *  B-d  |  \
1030 		 *       |   \      <- smaller scope is better if
1031 		 *  B-15 |    \        if scope is enough for destinaion.
1032 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1033 		 * d-C-1 | /
1034 		 *       |/         <- greater is better
1035 		 *   -C  /             if scope is not enough for destination.
1036 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1037 		 *
1038 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1039 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1040 		 * Assume B = 0 and we get C > 29.
1041 		 */
1042 		ret = __ipv6_addr_src_scope(score->addr_type);
1043 		if (ret >= dst->scope)
1044 			ret = -ret;
1045 		else
1046 			ret -= 128;	/* 30 is enough */
1047 		score->scopedist = ret;
1048 		break;
1049 	case IPV6_SADDR_RULE_PREFERRED:
1050 		/* Rule 3: Avoid deprecated and optimistic addresses */
1051 		ret = ipv6_saddr_preferred(score->addr_type) ||
1052 		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1053 		break;
1054 #ifdef CONFIG_IPV6_MIP6
1055 	case IPV6_SADDR_RULE_HOA:
1056 	    {
1057 		/* Rule 4: Prefer home address */
1058 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1059 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1060 		break;
1061 	    }
1062 #endif
1063 	case IPV6_SADDR_RULE_OIF:
1064 		/* Rule 5: Prefer outgoing interface */
1065 		ret = (!dst->ifindex ||
1066 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1067 		break;
1068 	case IPV6_SADDR_RULE_LABEL:
1069 		/* Rule 6: Prefer matching label */
1070 		ret = ipv6_addr_label(net,
1071 				      &score->ifa->addr, score->addr_type,
1072 				      score->ifa->idev->dev->ifindex) == dst->label;
1073 		break;
1074 #ifdef CONFIG_IPV6_PRIVACY
1075 	case IPV6_SADDR_RULE_PRIVACY:
1076 	    {
1077 		/* Rule 7: Prefer public address
1078 		 * Note: prefer temporary address if use_tempaddr >= 2
1079 		 */
1080 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1081 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1082 				score->ifa->idev->cnf.use_tempaddr >= 2;
1083 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1084 		break;
1085 	    }
1086 #endif
1087 	case IPV6_SADDR_RULE_ORCHID:
1088 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1089 		 *	    non-ORCHID vs non-ORCHID
1090 		 */
1091 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1092 			ipv6_addr_orchid(dst->addr));
1093 		break;
1094 	case IPV6_SADDR_RULE_PREFIX:
1095 		/* Rule 8: Use longest matching prefix */
1096 		score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1097 						       dst->addr);
1098 		break;
1099 	default:
1100 		ret = 0;
1101 	}
1102 
1103 	if (ret)
1104 		__set_bit(i, score->scorebits);
1105 	score->rule = i;
1106 out:
1107 	return ret;
1108 }
1109 
1110 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1111 		       const struct in6_addr *daddr, unsigned int prefs,
1112 		       struct in6_addr *saddr)
1113 {
1114 	struct ipv6_saddr_score scores[2],
1115 				*score = &scores[0], *hiscore = &scores[1];
1116 	struct ipv6_saddr_dst dst;
1117 	struct net_device *dev;
1118 	int dst_type;
1119 
1120 	dst_type = __ipv6_addr_type(daddr);
1121 	dst.addr = daddr;
1122 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1123 	dst.scope = __ipv6_addr_src_scope(dst_type);
1124 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1125 	dst.prefs = prefs;
1126 
1127 	hiscore->rule = -1;
1128 	hiscore->ifa = NULL;
1129 
1130 	rcu_read_lock();
1131 
1132 	for_each_netdev_rcu(net, dev) {
1133 		struct inet6_dev *idev;
1134 
1135 		/* Candidate Source Address (section 4)
1136 		 *  - multicast and link-local destination address,
1137 		 *    the set of candidate source address MUST only
1138 		 *    include addresses assigned to interfaces
1139 		 *    belonging to the same link as the outgoing
1140 		 *    interface.
1141 		 * (- For site-local destination addresses, the
1142 		 *    set of candidate source addresses MUST only
1143 		 *    include addresses assigned to interfaces
1144 		 *    belonging to the same site as the outgoing
1145 		 *    interface.)
1146 		 */
1147 		if (((dst_type & IPV6_ADDR_MULTICAST) ||
1148 		     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1149 		    dst.ifindex && dev->ifindex != dst.ifindex)
1150 			continue;
1151 
1152 		idev = __in6_dev_get(dev);
1153 		if (!idev)
1154 			continue;
1155 
1156 		read_lock_bh(&idev->lock);
1157 		list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1158 			int i;
1159 
1160 			/*
1161 			 * - Tentative Address (RFC2462 section 5.4)
1162 			 *  - A tentative address is not considered
1163 			 *    "assigned to an interface" in the traditional
1164 			 *    sense, unless it is also flagged as optimistic.
1165 			 * - Candidate Source Address (section 4)
1166 			 *  - In any case, anycast addresses, multicast
1167 			 *    addresses, and the unspecified address MUST
1168 			 *    NOT be included in a candidate set.
1169 			 */
1170 			if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1171 			    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1172 				continue;
1173 
1174 			score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1175 
1176 			if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1177 				     score->addr_type & IPV6_ADDR_MULTICAST)) {
1178 				LIMIT_NETDEBUG(KERN_DEBUG
1179 					       "ADDRCONF: unspecified / multicast address "
1180 					       "assigned as unicast address on %s",
1181 					       dev->name);
1182 				continue;
1183 			}
1184 
1185 			score->rule = -1;
1186 			bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1187 
1188 			for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1189 				int minihiscore, miniscore;
1190 
1191 				minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1192 				miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1193 
1194 				if (minihiscore > miniscore) {
1195 					if (i == IPV6_SADDR_RULE_SCOPE &&
1196 					    score->scopedist > 0) {
1197 						/*
1198 						 * special case:
1199 						 * each remaining entry
1200 						 * has too small (not enough)
1201 						 * scope, because ifa entries
1202 						 * are sorted by their scope
1203 						 * values.
1204 						 */
1205 						goto try_nextdev;
1206 					}
1207 					break;
1208 				} else if (minihiscore < miniscore) {
1209 					if (hiscore->ifa)
1210 						in6_ifa_put(hiscore->ifa);
1211 
1212 					in6_ifa_hold(score->ifa);
1213 
1214 					swap(hiscore, score);
1215 
1216 					/* restore our iterator */
1217 					score->ifa = hiscore->ifa;
1218 
1219 					break;
1220 				}
1221 			}
1222 		}
1223 try_nextdev:
1224 		read_unlock_bh(&idev->lock);
1225 	}
1226 	rcu_read_unlock();
1227 
1228 	if (!hiscore->ifa)
1229 		return -EADDRNOTAVAIL;
1230 
1231 	ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1232 	in6_ifa_put(hiscore->ifa);
1233 	return 0;
1234 }
1235 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1236 
1237 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1238 		    unsigned char banned_flags)
1239 {
1240 	struct inet6_dev *idev;
1241 	int err = -EADDRNOTAVAIL;
1242 
1243 	rcu_read_lock();
1244 	idev = __in6_dev_get(dev);
1245 	if (idev) {
1246 		struct inet6_ifaddr *ifp;
1247 
1248 		read_lock_bh(&idev->lock);
1249 		list_for_each_entry(ifp, &idev->addr_list, if_list) {
1250 			if (ifp->scope == IFA_LINK &&
1251 			    !(ifp->flags & banned_flags)) {
1252 				ipv6_addr_copy(addr, &ifp->addr);
1253 				err = 0;
1254 				break;
1255 			}
1256 		}
1257 		read_unlock_bh(&idev->lock);
1258 	}
1259 	rcu_read_unlock();
1260 	return err;
1261 }
1262 
1263 static int ipv6_count_addresses(struct inet6_dev *idev)
1264 {
1265 	int cnt = 0;
1266 	struct inet6_ifaddr *ifp;
1267 
1268 	read_lock_bh(&idev->lock);
1269 	list_for_each_entry(ifp, &idev->addr_list, if_list)
1270 		cnt++;
1271 	read_unlock_bh(&idev->lock);
1272 	return cnt;
1273 }
1274 
1275 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1276 		  struct net_device *dev, int strict)
1277 {
1278 	struct inet6_ifaddr *ifp;
1279 	struct hlist_node *node;
1280 	unsigned int hash = ipv6_addr_hash(addr);
1281 
1282 	rcu_read_lock_bh();
1283 	hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1284 		if (!net_eq(dev_net(ifp->idev->dev), net))
1285 			continue;
1286 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1287 		    !(ifp->flags&IFA_F_TENTATIVE) &&
1288 		    (dev == NULL || ifp->idev->dev == dev ||
1289 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1290 			rcu_read_unlock_bh();
1291 			return 1;
1292 		}
1293 	}
1294 
1295 	rcu_read_unlock_bh();
1296 	return 0;
1297 }
1298 EXPORT_SYMBOL(ipv6_chk_addr);
1299 
1300 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1301 			       struct net_device *dev)
1302 {
1303 	unsigned int hash = ipv6_addr_hash(addr);
1304 	struct inet6_ifaddr *ifp;
1305 	struct hlist_node *node;
1306 
1307 	hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1308 		if (!net_eq(dev_net(ifp->idev->dev), net))
1309 			continue;
1310 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1311 			if (dev == NULL || ifp->idev->dev == dev)
1312 				return true;
1313 		}
1314 	}
1315 	return false;
1316 }
1317 
1318 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1319 {
1320 	struct inet6_dev *idev;
1321 	struct inet6_ifaddr *ifa;
1322 	int	onlink;
1323 
1324 	onlink = 0;
1325 	rcu_read_lock();
1326 	idev = __in6_dev_get(dev);
1327 	if (idev) {
1328 		read_lock_bh(&idev->lock);
1329 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1330 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1331 						   ifa->prefix_len);
1332 			if (onlink)
1333 				break;
1334 		}
1335 		read_unlock_bh(&idev->lock);
1336 	}
1337 	rcu_read_unlock();
1338 	return onlink;
1339 }
1340 
1341 EXPORT_SYMBOL(ipv6_chk_prefix);
1342 
1343 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1344 				     struct net_device *dev, int strict)
1345 {
1346 	struct inet6_ifaddr *ifp, *result = NULL;
1347 	unsigned int hash = ipv6_addr_hash(addr);
1348 	struct hlist_node *node;
1349 
1350 	rcu_read_lock_bh();
1351 	hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1352 		if (!net_eq(dev_net(ifp->idev->dev), net))
1353 			continue;
1354 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1355 			if (dev == NULL || ifp->idev->dev == dev ||
1356 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1357 				result = ifp;
1358 				in6_ifa_hold(ifp);
1359 				break;
1360 			}
1361 		}
1362 	}
1363 	rcu_read_unlock_bh();
1364 
1365 	return result;
1366 }
1367 
1368 /* Gets referenced address, destroys ifaddr */
1369 
1370 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1371 {
1372 	if (ifp->flags&IFA_F_PERMANENT) {
1373 		spin_lock_bh(&ifp->lock);
1374 		addrconf_del_timer(ifp);
1375 		ifp->flags |= IFA_F_TENTATIVE;
1376 		if (dad_failed)
1377 			ifp->flags |= IFA_F_DADFAILED;
1378 		spin_unlock_bh(&ifp->lock);
1379 		if (dad_failed)
1380 			ipv6_ifa_notify(0, ifp);
1381 		in6_ifa_put(ifp);
1382 #ifdef CONFIG_IPV6_PRIVACY
1383 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1384 		struct inet6_ifaddr *ifpub;
1385 		spin_lock_bh(&ifp->lock);
1386 		ifpub = ifp->ifpub;
1387 		if (ifpub) {
1388 			in6_ifa_hold(ifpub);
1389 			spin_unlock_bh(&ifp->lock);
1390 			ipv6_create_tempaddr(ifpub, ifp);
1391 			in6_ifa_put(ifpub);
1392 		} else {
1393 			spin_unlock_bh(&ifp->lock);
1394 		}
1395 		ipv6_del_addr(ifp);
1396 #endif
1397 	} else
1398 		ipv6_del_addr(ifp);
1399 }
1400 
1401 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1402 {
1403 	int err = -ENOENT;
1404 
1405 	spin_lock(&ifp->state_lock);
1406 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
1407 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
1408 		err = 0;
1409 	}
1410 	spin_unlock(&ifp->state_lock);
1411 
1412 	return err;
1413 }
1414 
1415 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1416 {
1417 	struct inet6_dev *idev = ifp->idev;
1418 
1419 	if (addrconf_dad_end(ifp)) {
1420 		in6_ifa_put(ifp);
1421 		return;
1422 	}
1423 
1424 	if (net_ratelimit())
1425 		printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1426 			ifp->idev->dev->name, &ifp->addr);
1427 
1428 	if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1429 		struct in6_addr addr;
1430 
1431 		addr.s6_addr32[0] = htonl(0xfe800000);
1432 		addr.s6_addr32[1] = 0;
1433 
1434 		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1435 		    ipv6_addr_equal(&ifp->addr, &addr)) {
1436 			/* DAD failed for link-local based on MAC address */
1437 			idev->cnf.disable_ipv6 = 1;
1438 
1439 			printk(KERN_INFO "%s: IPv6 being disabled!\n",
1440 				ifp->idev->dev->name);
1441 		}
1442 	}
1443 
1444 	addrconf_dad_stop(ifp, 1);
1445 }
1446 
1447 /* Join to solicited addr multicast group. */
1448 
1449 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1450 {
1451 	struct in6_addr maddr;
1452 
1453 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1454 		return;
1455 
1456 	addrconf_addr_solict_mult(addr, &maddr);
1457 	ipv6_dev_mc_inc(dev, &maddr);
1458 }
1459 
1460 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1461 {
1462 	struct in6_addr maddr;
1463 
1464 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1465 		return;
1466 
1467 	addrconf_addr_solict_mult(addr, &maddr);
1468 	__ipv6_dev_mc_dec(idev, &maddr);
1469 }
1470 
1471 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1472 {
1473 	struct in6_addr addr;
1474 	if (ifp->prefix_len == 127) /* RFC 6164 */
1475 		return;
1476 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1477 	if (ipv6_addr_any(&addr))
1478 		return;
1479 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1480 }
1481 
1482 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1483 {
1484 	struct in6_addr addr;
1485 	if (ifp->prefix_len == 127) /* RFC 6164 */
1486 		return;
1487 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1488 	if (ipv6_addr_any(&addr))
1489 		return;
1490 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1491 }
1492 
1493 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1494 {
1495 	if (dev->addr_len != ETH_ALEN)
1496 		return -1;
1497 	memcpy(eui, dev->dev_addr, 3);
1498 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1499 
1500 	/*
1501 	 * The zSeries OSA network cards can be shared among various
1502 	 * OS instances, but the OSA cards have only one MAC address.
1503 	 * This leads to duplicate address conflicts in conjunction
1504 	 * with IPv6 if more than one instance uses the same card.
1505 	 *
1506 	 * The driver for these cards can deliver a unique 16-bit
1507 	 * identifier for each instance sharing the same card.  It is
1508 	 * placed instead of 0xFFFE in the interface identifier.  The
1509 	 * "u" bit of the interface identifier is not inverted in this
1510 	 * case.  Hence the resulting interface identifier has local
1511 	 * scope according to RFC2373.
1512 	 */
1513 	if (dev->dev_id) {
1514 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1515 		eui[4] = dev->dev_id & 0xFF;
1516 	} else {
1517 		eui[3] = 0xFF;
1518 		eui[4] = 0xFE;
1519 		eui[0] ^= 2;
1520 	}
1521 	return 0;
1522 }
1523 
1524 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1525 {
1526 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1527 	if (dev->addr_len != ARCNET_ALEN)
1528 		return -1;
1529 	memset(eui, 0, 7);
1530 	eui[7] = *(u8*)dev->dev_addr;
1531 	return 0;
1532 }
1533 
1534 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1535 {
1536 	if (dev->addr_len != INFINIBAND_ALEN)
1537 		return -1;
1538 	memcpy(eui, dev->dev_addr + 12, 8);
1539 	eui[0] |= 2;
1540 	return 0;
1541 }
1542 
1543 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1544 {
1545 	if (addr == 0)
1546 		return -1;
1547 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1548 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1549 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1550 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1551 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1552 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1553 	eui[1] = 0;
1554 	eui[2] = 0x5E;
1555 	eui[3] = 0xFE;
1556 	memcpy(eui + 4, &addr, 4);
1557 	return 0;
1558 }
1559 
1560 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1561 {
1562 	if (dev->priv_flags & IFF_ISATAP)
1563 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1564 	return -1;
1565 }
1566 
1567 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1568 {
1569 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1570 }
1571 
1572 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1573 {
1574 	switch (dev->type) {
1575 	case ARPHRD_ETHER:
1576 	case ARPHRD_FDDI:
1577 	case ARPHRD_IEEE802_TR:
1578 		return addrconf_ifid_eui48(eui, dev);
1579 	case ARPHRD_ARCNET:
1580 		return addrconf_ifid_arcnet(eui, dev);
1581 	case ARPHRD_INFINIBAND:
1582 		return addrconf_ifid_infiniband(eui, dev);
1583 	case ARPHRD_SIT:
1584 		return addrconf_ifid_sit(eui, dev);
1585 	case ARPHRD_IPGRE:
1586 		return addrconf_ifid_gre(eui, dev);
1587 	}
1588 	return -1;
1589 }
1590 
1591 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1592 {
1593 	int err = -1;
1594 	struct inet6_ifaddr *ifp;
1595 
1596 	read_lock_bh(&idev->lock);
1597 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
1598 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1599 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1600 			err = 0;
1601 			break;
1602 		}
1603 	}
1604 	read_unlock_bh(&idev->lock);
1605 	return err;
1606 }
1607 
1608 #ifdef CONFIG_IPV6_PRIVACY
1609 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1610 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1611 {
1612 regen:
1613 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1614 	idev->rndid[0] &= ~0x02;
1615 
1616 	/*
1617 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1618 	 * check if generated address is not inappropriate
1619 	 *
1620 	 *  - Reserved subnet anycast (RFC 2526)
1621 	 *	11111101 11....11 1xxxxxxx
1622 	 *  - ISATAP (RFC4214) 6.1
1623 	 *	00-00-5E-FE-xx-xx-xx-xx
1624 	 *  - value 0
1625 	 *  - XXX: already assigned to an address on the device
1626 	 */
1627 	if (idev->rndid[0] == 0xfd &&
1628 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1629 	    (idev->rndid[7]&0x80))
1630 		goto regen;
1631 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1632 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1633 			goto regen;
1634 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1635 			goto regen;
1636 	}
1637 
1638 	return 0;
1639 }
1640 
1641 static void ipv6_regen_rndid(unsigned long data)
1642 {
1643 	struct inet6_dev *idev = (struct inet6_dev *) data;
1644 	unsigned long expires;
1645 
1646 	rcu_read_lock_bh();
1647 	write_lock_bh(&idev->lock);
1648 
1649 	if (idev->dead)
1650 		goto out;
1651 
1652 	if (__ipv6_regen_rndid(idev) < 0)
1653 		goto out;
1654 
1655 	expires = jiffies +
1656 		idev->cnf.temp_prefered_lft * HZ -
1657 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1658 		idev->cnf.max_desync_factor * HZ;
1659 	if (time_before(expires, jiffies)) {
1660 		printk(KERN_WARNING
1661 			"ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1662 			idev->dev->name);
1663 		goto out;
1664 	}
1665 
1666 	if (!mod_timer(&idev->regen_timer, expires))
1667 		in6_dev_hold(idev);
1668 
1669 out:
1670 	write_unlock_bh(&idev->lock);
1671 	rcu_read_unlock_bh();
1672 	in6_dev_put(idev);
1673 }
1674 
1675 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1676 	int ret = 0;
1677 
1678 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1679 		ret = __ipv6_regen_rndid(idev);
1680 	return ret;
1681 }
1682 #endif
1683 
1684 /*
1685  *	Add prefix route.
1686  */
1687 
1688 static void
1689 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1690 		      unsigned long expires, u32 flags)
1691 {
1692 	struct fib6_config cfg = {
1693 		.fc_table = RT6_TABLE_PREFIX,
1694 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1695 		.fc_ifindex = dev->ifindex,
1696 		.fc_expires = expires,
1697 		.fc_dst_len = plen,
1698 		.fc_flags = RTF_UP | flags,
1699 		.fc_nlinfo.nl_net = dev_net(dev),
1700 		.fc_protocol = RTPROT_KERNEL,
1701 	};
1702 
1703 	ipv6_addr_copy(&cfg.fc_dst, pfx);
1704 
1705 	/* Prevent useless cloning on PtP SIT.
1706 	   This thing is done here expecting that the whole
1707 	   class of non-broadcast devices need not cloning.
1708 	 */
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1711 		cfg.fc_flags |= RTF_NONEXTHOP;
1712 #endif
1713 
1714 	ip6_route_add(&cfg);
1715 }
1716 
1717 /* Create "default" multicast route to the interface */
1718 
1719 static void addrconf_add_mroute(struct net_device *dev)
1720 {
1721 	struct fib6_config cfg = {
1722 		.fc_table = RT6_TABLE_LOCAL,
1723 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1724 		.fc_ifindex = dev->ifindex,
1725 		.fc_dst_len = 8,
1726 		.fc_flags = RTF_UP,
1727 		.fc_nlinfo.nl_net = dev_net(dev),
1728 	};
1729 
1730 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1731 
1732 	ip6_route_add(&cfg);
1733 }
1734 
1735 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1736 static void sit_route_add(struct net_device *dev)
1737 {
1738 	struct fib6_config cfg = {
1739 		.fc_table = RT6_TABLE_MAIN,
1740 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1741 		.fc_ifindex = dev->ifindex,
1742 		.fc_dst_len = 96,
1743 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
1744 		.fc_nlinfo.nl_net = dev_net(dev),
1745 	};
1746 
1747 	/* prefix length - 96 bits "::d.d.d.d" */
1748 	ip6_route_add(&cfg);
1749 }
1750 #endif
1751 
1752 static void addrconf_add_lroute(struct net_device *dev)
1753 {
1754 	struct in6_addr addr;
1755 
1756 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1757 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
1758 }
1759 
1760 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1761 {
1762 	struct inet6_dev *idev;
1763 
1764 	ASSERT_RTNL();
1765 
1766 	idev = ipv6_find_idev(dev);
1767 	if (!idev)
1768 		return ERR_PTR(-ENOBUFS);
1769 
1770 	if (idev->cnf.disable_ipv6)
1771 		return ERR_PTR(-EACCES);
1772 
1773 	/* Add default multicast route */
1774 	addrconf_add_mroute(dev);
1775 
1776 	/* Add link local route */
1777 	addrconf_add_lroute(dev);
1778 	return idev;
1779 }
1780 
1781 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1782 {
1783 	struct prefix_info *pinfo;
1784 	__u32 valid_lft;
1785 	__u32 prefered_lft;
1786 	int addr_type;
1787 	struct inet6_dev *in6_dev;
1788 	struct net *net = dev_net(dev);
1789 
1790 	pinfo = (struct prefix_info *) opt;
1791 
1792 	if (len < sizeof(struct prefix_info)) {
1793 		ADBG(("addrconf: prefix option too short\n"));
1794 		return;
1795 	}
1796 
1797 	/*
1798 	 *	Validation checks ([ADDRCONF], page 19)
1799 	 */
1800 
1801 	addr_type = ipv6_addr_type(&pinfo->prefix);
1802 
1803 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1804 		return;
1805 
1806 	valid_lft = ntohl(pinfo->valid);
1807 	prefered_lft = ntohl(pinfo->prefered);
1808 
1809 	if (prefered_lft > valid_lft) {
1810 		if (net_ratelimit())
1811 			printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1812 		return;
1813 	}
1814 
1815 	in6_dev = in6_dev_get(dev);
1816 
1817 	if (in6_dev == NULL) {
1818 		if (net_ratelimit())
1819 			printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1820 		return;
1821 	}
1822 
1823 	/*
1824 	 *	Two things going on here:
1825 	 *	1) Add routes for on-link prefixes
1826 	 *	2) Configure prefixes with the auto flag set
1827 	 */
1828 
1829 	if (pinfo->onlink) {
1830 		struct rt6_info *rt;
1831 		unsigned long rt_expires;
1832 
1833 		/* Avoid arithmetic overflow. Really, we could
1834 		 * save rt_expires in seconds, likely valid_lft,
1835 		 * but it would require division in fib gc, that it
1836 		 * not good.
1837 		 */
1838 		if (HZ > USER_HZ)
1839 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1840 		else
1841 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1842 
1843 		if (addrconf_finite_timeout(rt_expires))
1844 			rt_expires *= HZ;
1845 
1846 		rt = rt6_lookup(net, &pinfo->prefix, NULL,
1847 				dev->ifindex, 1);
1848 
1849 		if (rt && addrconf_is_prefix_route(rt)) {
1850 			/* Autoconf prefix route */
1851 			if (valid_lft == 0) {
1852 				ip6_del_rt(rt);
1853 				rt = NULL;
1854 			} else if (addrconf_finite_timeout(rt_expires)) {
1855 				/* not infinity */
1856 				rt->rt6i_expires = jiffies + rt_expires;
1857 				rt->rt6i_flags |= RTF_EXPIRES;
1858 			} else {
1859 				rt->rt6i_flags &= ~RTF_EXPIRES;
1860 				rt->rt6i_expires = 0;
1861 			}
1862 		} else if (valid_lft) {
1863 			clock_t expires = 0;
1864 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1865 			if (addrconf_finite_timeout(rt_expires)) {
1866 				/* not infinity */
1867 				flags |= RTF_EXPIRES;
1868 				expires = jiffies_to_clock_t(rt_expires);
1869 			}
1870 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1871 					      dev, expires, flags);
1872 		}
1873 		if (rt)
1874 			dst_release(&rt->dst);
1875 	}
1876 
1877 	/* Try to figure out our local address for this prefix */
1878 
1879 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1880 		struct inet6_ifaddr * ifp;
1881 		struct in6_addr addr;
1882 		int create = 0, update_lft = 0;
1883 
1884 		if (pinfo->prefix_len == 64) {
1885 			memcpy(&addr, &pinfo->prefix, 8);
1886 			if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1887 			    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1888 				in6_dev_put(in6_dev);
1889 				return;
1890 			}
1891 			goto ok;
1892 		}
1893 		if (net_ratelimit())
1894 			printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1895 			       pinfo->prefix_len);
1896 		in6_dev_put(in6_dev);
1897 		return;
1898 
1899 ok:
1900 
1901 		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1902 
1903 		if (ifp == NULL && valid_lft) {
1904 			int max_addresses = in6_dev->cnf.max_addresses;
1905 			u32 addr_flags = 0;
1906 
1907 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1908 			if (in6_dev->cnf.optimistic_dad &&
1909 			    !net->ipv6.devconf_all->forwarding)
1910 				addr_flags = IFA_F_OPTIMISTIC;
1911 #endif
1912 
1913 			/* Do not allow to create too much of autoconfigured
1914 			 * addresses; this would be too easy way to crash kernel.
1915 			 */
1916 			if (!max_addresses ||
1917 			    ipv6_count_addresses(in6_dev) < max_addresses)
1918 				ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1919 						    addr_type&IPV6_ADDR_SCOPE_MASK,
1920 						    addr_flags);
1921 
1922 			if (!ifp || IS_ERR(ifp)) {
1923 				in6_dev_put(in6_dev);
1924 				return;
1925 			}
1926 
1927 			update_lft = create = 1;
1928 			ifp->cstamp = jiffies;
1929 			addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1930 		}
1931 
1932 		if (ifp) {
1933 			int flags;
1934 			unsigned long now;
1935 #ifdef CONFIG_IPV6_PRIVACY
1936 			struct inet6_ifaddr *ift;
1937 #endif
1938 			u32 stored_lft;
1939 
1940 			/* update lifetime (RFC2462 5.5.3 e) */
1941 			spin_lock(&ifp->lock);
1942 			now = jiffies;
1943 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1944 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1945 			else
1946 				stored_lft = 0;
1947 			if (!update_lft && stored_lft) {
1948 				if (valid_lft > MIN_VALID_LIFETIME ||
1949 				    valid_lft > stored_lft)
1950 					update_lft = 1;
1951 				else if (stored_lft <= MIN_VALID_LIFETIME) {
1952 					/* valid_lft <= stored_lft is always true */
1953 					/*
1954 					 * RFC 4862 Section 5.5.3e:
1955 					 * "Note that the preferred lifetime of
1956 					 *  the corresponding address is always
1957 					 *  reset to the Preferred Lifetime in
1958 					 *  the received Prefix Information
1959 					 *  option, regardless of whether the
1960 					 *  valid lifetime is also reset or
1961 					 *  ignored."
1962 					 *
1963 					 *  So if the preferred lifetime in
1964 					 *  this advertisement is different
1965 					 *  than what we have stored, but the
1966 					 *  valid lifetime is invalid, just
1967 					 *  reset prefered_lft.
1968 					 *
1969 					 *  We must set the valid lifetime
1970 					 *  to the stored lifetime since we'll
1971 					 *  be updating the timestamp below,
1972 					 *  else we'll set it back to the
1973 					 *  minimum.
1974 					 */
1975 					if (prefered_lft != ifp->prefered_lft) {
1976 						valid_lft = stored_lft;
1977 						update_lft = 1;
1978 					}
1979 				} else {
1980 					valid_lft = MIN_VALID_LIFETIME;
1981 					if (valid_lft < prefered_lft)
1982 						prefered_lft = valid_lft;
1983 					update_lft = 1;
1984 				}
1985 			}
1986 
1987 			if (update_lft) {
1988 				ifp->valid_lft = valid_lft;
1989 				ifp->prefered_lft = prefered_lft;
1990 				ifp->tstamp = now;
1991 				flags = ifp->flags;
1992 				ifp->flags &= ~IFA_F_DEPRECATED;
1993 				spin_unlock(&ifp->lock);
1994 
1995 				if (!(flags&IFA_F_TENTATIVE))
1996 					ipv6_ifa_notify(0, ifp);
1997 			} else
1998 				spin_unlock(&ifp->lock);
1999 
2000 #ifdef CONFIG_IPV6_PRIVACY
2001 			read_lock_bh(&in6_dev->lock);
2002 			/* update all temporary addresses in the list */
2003 			list_for_each_entry(ift, &in6_dev->tempaddr_list,
2004 					    tmp_list) {
2005 				int age, max_valid, max_prefered;
2006 
2007 				if (ifp != ift->ifpub)
2008 					continue;
2009 
2010 				/*
2011 				 * RFC 4941 section 3.3:
2012 				 * If a received option will extend the lifetime
2013 				 * of a public address, the lifetimes of
2014 				 * temporary addresses should be extended,
2015 				 * subject to the overall constraint that no
2016 				 * temporary addresses should ever remain
2017 				 * "valid" or "preferred" for a time longer than
2018 				 * (TEMP_VALID_LIFETIME) or
2019 				 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2020 				 * respectively.
2021 				 */
2022 				age = (now - ift->cstamp) / HZ;
2023 				max_valid = in6_dev->cnf.temp_valid_lft - age;
2024 				if (max_valid < 0)
2025 					max_valid = 0;
2026 
2027 				max_prefered = in6_dev->cnf.temp_prefered_lft -
2028 					       in6_dev->cnf.max_desync_factor -
2029 					       age;
2030 				if (max_prefered < 0)
2031 					max_prefered = 0;
2032 
2033 				if (valid_lft > max_valid)
2034 					valid_lft = max_valid;
2035 
2036 				if (prefered_lft > max_prefered)
2037 					prefered_lft = max_prefered;
2038 
2039 				spin_lock(&ift->lock);
2040 				flags = ift->flags;
2041 				ift->valid_lft = valid_lft;
2042 				ift->prefered_lft = prefered_lft;
2043 				ift->tstamp = now;
2044 				if (prefered_lft > 0)
2045 					ift->flags &= ~IFA_F_DEPRECATED;
2046 
2047 				spin_unlock(&ift->lock);
2048 				if (!(flags&IFA_F_TENTATIVE))
2049 					ipv6_ifa_notify(0, ift);
2050 			}
2051 
2052 			if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2053 				/*
2054 				 * When a new public address is created as
2055 				 * described in [ADDRCONF], also create a new
2056 				 * temporary address. Also create a temporary
2057 				 * address if it's enabled but no temporary
2058 				 * address currently exists.
2059 				 */
2060 				read_unlock_bh(&in6_dev->lock);
2061 				ipv6_create_tempaddr(ifp, NULL);
2062 			} else {
2063 				read_unlock_bh(&in6_dev->lock);
2064 			}
2065 #endif
2066 			in6_ifa_put(ifp);
2067 			addrconf_verify(0);
2068 		}
2069 	}
2070 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2071 	in6_dev_put(in6_dev);
2072 }
2073 
2074 /*
2075  *	Set destination address.
2076  *	Special case for SIT interfaces where we create a new "virtual"
2077  *	device.
2078  */
2079 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2080 {
2081 	struct in6_ifreq ireq;
2082 	struct net_device *dev;
2083 	int err = -EINVAL;
2084 
2085 	rtnl_lock();
2086 
2087 	err = -EFAULT;
2088 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2089 		goto err_exit;
2090 
2091 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2092 
2093 	err = -ENODEV;
2094 	if (dev == NULL)
2095 		goto err_exit;
2096 
2097 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2098 	if (dev->type == ARPHRD_SIT) {
2099 		const struct net_device_ops *ops = dev->netdev_ops;
2100 		struct ifreq ifr;
2101 		struct ip_tunnel_parm p;
2102 
2103 		err = -EADDRNOTAVAIL;
2104 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2105 			goto err_exit;
2106 
2107 		memset(&p, 0, sizeof(p));
2108 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2109 		p.iph.saddr = 0;
2110 		p.iph.version = 4;
2111 		p.iph.ihl = 5;
2112 		p.iph.protocol = IPPROTO_IPV6;
2113 		p.iph.ttl = 64;
2114 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2115 
2116 		if (ops->ndo_do_ioctl) {
2117 			mm_segment_t oldfs = get_fs();
2118 
2119 			set_fs(KERNEL_DS);
2120 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2121 			set_fs(oldfs);
2122 		} else
2123 			err = -EOPNOTSUPP;
2124 
2125 		if (err == 0) {
2126 			err = -ENOBUFS;
2127 			dev = __dev_get_by_name(net, p.name);
2128 			if (!dev)
2129 				goto err_exit;
2130 			err = dev_open(dev);
2131 		}
2132 	}
2133 #endif
2134 
2135 err_exit:
2136 	rtnl_unlock();
2137 	return err;
2138 }
2139 
2140 /*
2141  *	Manual configuration of address on an interface
2142  */
2143 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2144 			  unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2145 			  __u32 valid_lft)
2146 {
2147 	struct inet6_ifaddr *ifp;
2148 	struct inet6_dev *idev;
2149 	struct net_device *dev;
2150 	int scope;
2151 	u32 flags;
2152 	clock_t expires;
2153 	unsigned long timeout;
2154 
2155 	ASSERT_RTNL();
2156 
2157 	if (plen > 128)
2158 		return -EINVAL;
2159 
2160 	/* check the lifetime */
2161 	if (!valid_lft || prefered_lft > valid_lft)
2162 		return -EINVAL;
2163 
2164 	dev = __dev_get_by_index(net, ifindex);
2165 	if (!dev)
2166 		return -ENODEV;
2167 
2168 	idev = addrconf_add_dev(dev);
2169 	if (IS_ERR(idev))
2170 		return PTR_ERR(idev);
2171 
2172 	scope = ipv6_addr_scope(pfx);
2173 
2174 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2175 	if (addrconf_finite_timeout(timeout)) {
2176 		expires = jiffies_to_clock_t(timeout * HZ);
2177 		valid_lft = timeout;
2178 		flags = RTF_EXPIRES;
2179 	} else {
2180 		expires = 0;
2181 		flags = 0;
2182 		ifa_flags |= IFA_F_PERMANENT;
2183 	}
2184 
2185 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2186 	if (addrconf_finite_timeout(timeout)) {
2187 		if (timeout == 0)
2188 			ifa_flags |= IFA_F_DEPRECATED;
2189 		prefered_lft = timeout;
2190 	}
2191 
2192 	ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2193 
2194 	if (!IS_ERR(ifp)) {
2195 		spin_lock_bh(&ifp->lock);
2196 		ifp->valid_lft = valid_lft;
2197 		ifp->prefered_lft = prefered_lft;
2198 		ifp->tstamp = jiffies;
2199 		spin_unlock_bh(&ifp->lock);
2200 
2201 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2202 				      expires, flags);
2203 		/*
2204 		 * Note that section 3.1 of RFC 4429 indicates
2205 		 * that the Optimistic flag should not be set for
2206 		 * manually configured addresses
2207 		 */
2208 		addrconf_dad_start(ifp, 0);
2209 		in6_ifa_put(ifp);
2210 		addrconf_verify(0);
2211 		return 0;
2212 	}
2213 
2214 	return PTR_ERR(ifp);
2215 }
2216 
2217 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2218 			  unsigned int plen)
2219 {
2220 	struct inet6_ifaddr *ifp;
2221 	struct inet6_dev *idev;
2222 	struct net_device *dev;
2223 
2224 	if (plen > 128)
2225 		return -EINVAL;
2226 
2227 	dev = __dev_get_by_index(net, ifindex);
2228 	if (!dev)
2229 		return -ENODEV;
2230 
2231 	if ((idev = __in6_dev_get(dev)) == NULL)
2232 		return -ENXIO;
2233 
2234 	read_lock_bh(&idev->lock);
2235 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2236 		if (ifp->prefix_len == plen &&
2237 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2238 			in6_ifa_hold(ifp);
2239 			read_unlock_bh(&idev->lock);
2240 
2241 			ipv6_del_addr(ifp);
2242 
2243 			/* If the last address is deleted administratively,
2244 			   disable IPv6 on this interface.
2245 			 */
2246 			if (list_empty(&idev->addr_list))
2247 				addrconf_ifdown(idev->dev, 1);
2248 			return 0;
2249 		}
2250 	}
2251 	read_unlock_bh(&idev->lock);
2252 	return -EADDRNOTAVAIL;
2253 }
2254 
2255 
2256 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2257 {
2258 	struct in6_ifreq ireq;
2259 	int err;
2260 
2261 	if (!capable(CAP_NET_ADMIN))
2262 		return -EPERM;
2263 
2264 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2265 		return -EFAULT;
2266 
2267 	rtnl_lock();
2268 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2269 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2270 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2271 	rtnl_unlock();
2272 	return err;
2273 }
2274 
2275 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2276 {
2277 	struct in6_ifreq ireq;
2278 	int err;
2279 
2280 	if (!capable(CAP_NET_ADMIN))
2281 		return -EPERM;
2282 
2283 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2284 		return -EFAULT;
2285 
2286 	rtnl_lock();
2287 	err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2288 			     ireq.ifr6_prefixlen);
2289 	rtnl_unlock();
2290 	return err;
2291 }
2292 
2293 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2294 		     int plen, int scope)
2295 {
2296 	struct inet6_ifaddr *ifp;
2297 
2298 	ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2299 	if (!IS_ERR(ifp)) {
2300 		spin_lock_bh(&ifp->lock);
2301 		ifp->flags &= ~IFA_F_TENTATIVE;
2302 		spin_unlock_bh(&ifp->lock);
2303 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2304 		in6_ifa_put(ifp);
2305 	}
2306 }
2307 
2308 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2309 static void sit_add_v4_addrs(struct inet6_dev *idev)
2310 {
2311 	struct in6_addr addr;
2312 	struct net_device *dev;
2313 	struct net *net = dev_net(idev->dev);
2314 	int scope;
2315 
2316 	ASSERT_RTNL();
2317 
2318 	memset(&addr, 0, sizeof(struct in6_addr));
2319 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2320 
2321 	if (idev->dev->flags&IFF_POINTOPOINT) {
2322 		addr.s6_addr32[0] = htonl(0xfe800000);
2323 		scope = IFA_LINK;
2324 	} else {
2325 		scope = IPV6_ADDR_COMPATv4;
2326 	}
2327 
2328 	if (addr.s6_addr32[3]) {
2329 		add_addr(idev, &addr, 128, scope);
2330 		return;
2331 	}
2332 
2333 	for_each_netdev(net, dev) {
2334 		struct in_device * in_dev = __in_dev_get_rtnl(dev);
2335 		if (in_dev && (dev->flags & IFF_UP)) {
2336 			struct in_ifaddr * ifa;
2337 
2338 			int flag = scope;
2339 
2340 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2341 				int plen;
2342 
2343 				addr.s6_addr32[3] = ifa->ifa_local;
2344 
2345 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2346 					continue;
2347 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2348 					if (idev->dev->flags&IFF_POINTOPOINT)
2349 						continue;
2350 					flag |= IFA_HOST;
2351 				}
2352 				if (idev->dev->flags&IFF_POINTOPOINT)
2353 					plen = 64;
2354 				else
2355 					plen = 96;
2356 
2357 				add_addr(idev, &addr, plen, flag);
2358 			}
2359 		}
2360 	}
2361 }
2362 #endif
2363 
2364 static void init_loopback(struct net_device *dev)
2365 {
2366 	struct inet6_dev  *idev;
2367 
2368 	/* ::1 */
2369 
2370 	ASSERT_RTNL();
2371 
2372 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2373 		printk(KERN_DEBUG "init loopback: add_dev failed\n");
2374 		return;
2375 	}
2376 
2377 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2378 }
2379 
2380 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2381 {
2382 	struct inet6_ifaddr * ifp;
2383 	u32 addr_flags = IFA_F_PERMANENT;
2384 
2385 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2386 	if (idev->cnf.optimistic_dad &&
2387 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2388 		addr_flags |= IFA_F_OPTIMISTIC;
2389 #endif
2390 
2391 
2392 	ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2393 	if (!IS_ERR(ifp)) {
2394 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2395 		addrconf_dad_start(ifp, 0);
2396 		in6_ifa_put(ifp);
2397 	}
2398 }
2399 
2400 static void addrconf_dev_config(struct net_device *dev)
2401 {
2402 	struct in6_addr addr;
2403 	struct inet6_dev    * idev;
2404 
2405 	ASSERT_RTNL();
2406 
2407 	if ((dev->type != ARPHRD_ETHER) &&
2408 	    (dev->type != ARPHRD_FDDI) &&
2409 	    (dev->type != ARPHRD_IEEE802_TR) &&
2410 	    (dev->type != ARPHRD_ARCNET) &&
2411 	    (dev->type != ARPHRD_INFINIBAND)) {
2412 		/* Alas, we support only Ethernet autoconfiguration. */
2413 		return;
2414 	}
2415 
2416 	idev = addrconf_add_dev(dev);
2417 	if (IS_ERR(idev))
2418 		return;
2419 
2420 	memset(&addr, 0, sizeof(struct in6_addr));
2421 	addr.s6_addr32[0] = htonl(0xFE800000);
2422 
2423 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2424 		addrconf_add_linklocal(idev, &addr);
2425 }
2426 
2427 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2428 static void addrconf_sit_config(struct net_device *dev)
2429 {
2430 	struct inet6_dev *idev;
2431 
2432 	ASSERT_RTNL();
2433 
2434 	/*
2435 	 * Configure the tunnel with one of our IPv4
2436 	 * addresses... we should configure all of
2437 	 * our v4 addrs in the tunnel
2438 	 */
2439 
2440 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2441 		printk(KERN_DEBUG "init sit: add_dev failed\n");
2442 		return;
2443 	}
2444 
2445 	if (dev->priv_flags & IFF_ISATAP) {
2446 		struct in6_addr addr;
2447 
2448 		ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2449 		addrconf_prefix_route(&addr, 64, dev, 0, 0);
2450 		if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2451 			addrconf_add_linklocal(idev, &addr);
2452 		return;
2453 	}
2454 
2455 	sit_add_v4_addrs(idev);
2456 
2457 	if (dev->flags&IFF_POINTOPOINT) {
2458 		addrconf_add_mroute(dev);
2459 		addrconf_add_lroute(dev);
2460 	} else
2461 		sit_route_add(dev);
2462 }
2463 #endif
2464 
2465 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2466 static void addrconf_gre_config(struct net_device *dev)
2467 {
2468 	struct inet6_dev *idev;
2469 	struct in6_addr addr;
2470 
2471 	pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2472 
2473 	ASSERT_RTNL();
2474 
2475 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2476 		printk(KERN_DEBUG "init gre: add_dev failed\n");
2477 		return;
2478 	}
2479 
2480 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2481 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
2482 
2483 	if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2484 		addrconf_add_linklocal(idev, &addr);
2485 }
2486 #endif
2487 
2488 static inline int
2489 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2490 {
2491 	struct in6_addr lladdr;
2492 
2493 	if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2494 		addrconf_add_linklocal(idev, &lladdr);
2495 		return 0;
2496 	}
2497 	return -1;
2498 }
2499 
2500 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2501 {
2502 	struct net_device *link_dev;
2503 	struct net *net = dev_net(idev->dev);
2504 
2505 	/* first try to inherit the link-local address from the link device */
2506 	if (idev->dev->iflink &&
2507 	    (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2508 		if (!ipv6_inherit_linklocal(idev, link_dev))
2509 			return;
2510 	}
2511 	/* then try to inherit it from any device */
2512 	for_each_netdev(net, link_dev) {
2513 		if (!ipv6_inherit_linklocal(idev, link_dev))
2514 			return;
2515 	}
2516 	printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2517 }
2518 
2519 /*
2520  * Autoconfigure tunnel with a link-local address so routing protocols,
2521  * DHCPv6, MLD etc. can be run over the virtual link
2522  */
2523 
2524 static void addrconf_ip6_tnl_config(struct net_device *dev)
2525 {
2526 	struct inet6_dev *idev;
2527 
2528 	ASSERT_RTNL();
2529 
2530 	idev = addrconf_add_dev(dev);
2531 	if (IS_ERR(idev)) {
2532 		printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2533 		return;
2534 	}
2535 	ip6_tnl_add_linklocal(idev);
2536 }
2537 
2538 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2539 			   void * data)
2540 {
2541 	struct net_device *dev = (struct net_device *) data;
2542 	struct inet6_dev *idev = __in6_dev_get(dev);
2543 	int run_pending = 0;
2544 	int err;
2545 
2546 	switch (event) {
2547 	case NETDEV_REGISTER:
2548 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2549 			idev = ipv6_add_dev(dev);
2550 			if (!idev)
2551 				return notifier_from_errno(-ENOMEM);
2552 		}
2553 		break;
2554 
2555 	case NETDEV_UP:
2556 	case NETDEV_CHANGE:
2557 		if (dev->flags & IFF_SLAVE)
2558 			break;
2559 
2560 		if (event == NETDEV_UP) {
2561 			if (!addrconf_qdisc_ok(dev)) {
2562 				/* device is not ready yet. */
2563 				printk(KERN_INFO
2564 					"ADDRCONF(NETDEV_UP): %s: "
2565 					"link is not ready\n",
2566 					dev->name);
2567 				break;
2568 			}
2569 
2570 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
2571 				idev = ipv6_add_dev(dev);
2572 
2573 			if (idev) {
2574 				idev->if_flags |= IF_READY;
2575 				run_pending = 1;
2576 			}
2577 		} else {
2578 			if (!addrconf_qdisc_ok(dev)) {
2579 				/* device is still not ready. */
2580 				break;
2581 			}
2582 
2583 			if (idev) {
2584 				if (idev->if_flags & IF_READY)
2585 					/* device is already configured. */
2586 					break;
2587 				idev->if_flags |= IF_READY;
2588 			}
2589 
2590 			printk(KERN_INFO
2591 					"ADDRCONF(NETDEV_CHANGE): %s: "
2592 					"link becomes ready\n",
2593 					dev->name);
2594 
2595 			run_pending = 1;
2596 		}
2597 
2598 		switch (dev->type) {
2599 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2600 		case ARPHRD_SIT:
2601 			addrconf_sit_config(dev);
2602 			break;
2603 #endif
2604 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2605 		case ARPHRD_IPGRE:
2606 			addrconf_gre_config(dev);
2607 			break;
2608 #endif
2609 		case ARPHRD_TUNNEL6:
2610 			addrconf_ip6_tnl_config(dev);
2611 			break;
2612 		case ARPHRD_LOOPBACK:
2613 			init_loopback(dev);
2614 			break;
2615 
2616 		default:
2617 			addrconf_dev_config(dev);
2618 			break;
2619 		}
2620 
2621 		if (idev) {
2622 			if (run_pending)
2623 				addrconf_dad_run(idev);
2624 
2625 			/*
2626 			 * If the MTU changed during the interface down,
2627 			 * when the interface up, the changed MTU must be
2628 			 * reflected in the idev as well as routers.
2629 			 */
2630 			if (idev->cnf.mtu6 != dev->mtu &&
2631 			    dev->mtu >= IPV6_MIN_MTU) {
2632 				rt6_mtu_change(dev, dev->mtu);
2633 				idev->cnf.mtu6 = dev->mtu;
2634 			}
2635 			idev->tstamp = jiffies;
2636 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2637 
2638 			/*
2639 			 * If the changed mtu during down is lower than
2640 			 * IPV6_MIN_MTU stop IPv6 on this interface.
2641 			 */
2642 			if (dev->mtu < IPV6_MIN_MTU)
2643 				addrconf_ifdown(dev, 1);
2644 		}
2645 		break;
2646 
2647 	case NETDEV_CHANGEMTU:
2648 		if (idev && dev->mtu >= IPV6_MIN_MTU) {
2649 			rt6_mtu_change(dev, dev->mtu);
2650 			idev->cnf.mtu6 = dev->mtu;
2651 			break;
2652 		}
2653 
2654 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2655 			idev = ipv6_add_dev(dev);
2656 			if (idev)
2657 				break;
2658 		}
2659 
2660 		/*
2661 		 * MTU falled under IPV6_MIN_MTU.
2662 		 * Stop IPv6 on this interface.
2663 		 */
2664 
2665 	case NETDEV_DOWN:
2666 	case NETDEV_UNREGISTER:
2667 		/*
2668 		 *	Remove all addresses from this interface.
2669 		 */
2670 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2671 		break;
2672 
2673 	case NETDEV_CHANGENAME:
2674 		if (idev) {
2675 			snmp6_unregister_dev(idev);
2676 			addrconf_sysctl_unregister(idev);
2677 			addrconf_sysctl_register(idev);
2678 			err = snmp6_register_dev(idev);
2679 			if (err)
2680 				return notifier_from_errno(err);
2681 		}
2682 		break;
2683 
2684 	case NETDEV_PRE_TYPE_CHANGE:
2685 	case NETDEV_POST_TYPE_CHANGE:
2686 		addrconf_type_change(dev, event);
2687 		break;
2688 	}
2689 
2690 	return NOTIFY_OK;
2691 }
2692 
2693 /*
2694  *	addrconf module should be notified of a device going up
2695  */
2696 static struct notifier_block ipv6_dev_notf = {
2697 	.notifier_call = addrconf_notify,
2698 };
2699 
2700 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2701 {
2702 	struct inet6_dev *idev;
2703 	ASSERT_RTNL();
2704 
2705 	idev = __in6_dev_get(dev);
2706 
2707 	if (event == NETDEV_POST_TYPE_CHANGE)
2708 		ipv6_mc_remap(idev);
2709 	else if (event == NETDEV_PRE_TYPE_CHANGE)
2710 		ipv6_mc_unmap(idev);
2711 }
2712 
2713 static int addrconf_ifdown(struct net_device *dev, int how)
2714 {
2715 	struct net *net = dev_net(dev);
2716 	struct inet6_dev *idev;
2717 	struct inet6_ifaddr *ifa;
2718 	int state, i;
2719 
2720 	ASSERT_RTNL();
2721 
2722 	rt6_ifdown(net, dev);
2723 	neigh_ifdown(&nd_tbl, dev);
2724 
2725 	idev = __in6_dev_get(dev);
2726 	if (idev == NULL)
2727 		return -ENODEV;
2728 
2729 	/*
2730 	 * Step 1: remove reference to ipv6 device from parent device.
2731 	 *	   Do not dev_put!
2732 	 */
2733 	if (how) {
2734 		idev->dead = 1;
2735 
2736 		/* protected by rtnl_lock */
2737 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2738 
2739 		/* Step 1.5: remove snmp6 entry */
2740 		snmp6_unregister_dev(idev);
2741 
2742 	}
2743 
2744 	/* Step 2: clear hash table */
2745 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2746 		struct hlist_head *h = &inet6_addr_lst[i];
2747 		struct hlist_node *n;
2748 
2749 		spin_lock_bh(&addrconf_hash_lock);
2750 	restart:
2751 		hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2752 			if (ifa->idev == idev) {
2753 				hlist_del_init_rcu(&ifa->addr_lst);
2754 				addrconf_del_timer(ifa);
2755 				goto restart;
2756 			}
2757 		}
2758 		spin_unlock_bh(&addrconf_hash_lock);
2759 	}
2760 
2761 	write_lock_bh(&idev->lock);
2762 
2763 	/* Step 2: clear flags for stateless addrconf */
2764 	if (!how)
2765 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2766 
2767 #ifdef CONFIG_IPV6_PRIVACY
2768 	if (how && del_timer(&idev->regen_timer))
2769 		in6_dev_put(idev);
2770 
2771 	/* Step 3: clear tempaddr list */
2772 	while (!list_empty(&idev->tempaddr_list)) {
2773 		ifa = list_first_entry(&idev->tempaddr_list,
2774 				       struct inet6_ifaddr, tmp_list);
2775 		list_del(&ifa->tmp_list);
2776 		write_unlock_bh(&idev->lock);
2777 		spin_lock_bh(&ifa->lock);
2778 
2779 		if (ifa->ifpub) {
2780 			in6_ifa_put(ifa->ifpub);
2781 			ifa->ifpub = NULL;
2782 		}
2783 		spin_unlock_bh(&ifa->lock);
2784 		in6_ifa_put(ifa);
2785 		write_lock_bh(&idev->lock);
2786 	}
2787 #endif
2788 
2789 	while (!list_empty(&idev->addr_list)) {
2790 		ifa = list_first_entry(&idev->addr_list,
2791 				       struct inet6_ifaddr, if_list);
2792 		addrconf_del_timer(ifa);
2793 
2794 		list_del(&ifa->if_list);
2795 
2796 		write_unlock_bh(&idev->lock);
2797 
2798 		spin_lock_bh(&ifa->state_lock);
2799 		state = ifa->state;
2800 		ifa->state = INET6_IFADDR_STATE_DEAD;
2801 		spin_unlock_bh(&ifa->state_lock);
2802 
2803 		if (state != INET6_IFADDR_STATE_DEAD) {
2804 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
2805 			atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2806 		}
2807 		in6_ifa_put(ifa);
2808 
2809 		write_lock_bh(&idev->lock);
2810 	}
2811 
2812 	write_unlock_bh(&idev->lock);
2813 
2814 	/* Step 5: Discard multicast list */
2815 	if (how)
2816 		ipv6_mc_destroy_dev(idev);
2817 	else
2818 		ipv6_mc_down(idev);
2819 
2820 	idev->tstamp = jiffies;
2821 
2822 	/* Last: Shot the device (if unregistered) */
2823 	if (how) {
2824 		addrconf_sysctl_unregister(idev);
2825 		neigh_parms_release(&nd_tbl, idev->nd_parms);
2826 		neigh_ifdown(&nd_tbl, dev);
2827 		in6_dev_put(idev);
2828 	}
2829 	return 0;
2830 }
2831 
2832 static void addrconf_rs_timer(unsigned long data)
2833 {
2834 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2835 	struct inet6_dev *idev = ifp->idev;
2836 
2837 	read_lock(&idev->lock);
2838 	if (idev->dead || !(idev->if_flags & IF_READY))
2839 		goto out;
2840 
2841 	if (idev->cnf.forwarding)
2842 		goto out;
2843 
2844 	/* Announcement received after solicitation was sent */
2845 	if (idev->if_flags & IF_RA_RCVD)
2846 		goto out;
2847 
2848 	spin_lock(&ifp->lock);
2849 	if (ifp->probes++ < idev->cnf.rtr_solicits) {
2850 		/* The wait after the last probe can be shorter */
2851 		addrconf_mod_timer(ifp, AC_RS,
2852 				   (ifp->probes == idev->cnf.rtr_solicits) ?
2853 				   idev->cnf.rtr_solicit_delay :
2854 				   idev->cnf.rtr_solicit_interval);
2855 		spin_unlock(&ifp->lock);
2856 
2857 		ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2858 	} else {
2859 		spin_unlock(&ifp->lock);
2860 		/*
2861 		 * Note: we do not support deprecated "all on-link"
2862 		 * assumption any longer.
2863 		 */
2864 		printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2865 		       idev->dev->name);
2866 	}
2867 
2868 out:
2869 	read_unlock(&idev->lock);
2870 	in6_ifa_put(ifp);
2871 }
2872 
2873 /*
2874  *	Duplicate Address Detection
2875  */
2876 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2877 {
2878 	unsigned long rand_num;
2879 	struct inet6_dev *idev = ifp->idev;
2880 
2881 	if (ifp->flags & IFA_F_OPTIMISTIC)
2882 		rand_num = 0;
2883 	else
2884 		rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2885 
2886 	ifp->probes = idev->cnf.dad_transmits;
2887 	addrconf_mod_timer(ifp, AC_DAD, rand_num);
2888 }
2889 
2890 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2891 {
2892 	struct inet6_dev *idev = ifp->idev;
2893 	struct net_device *dev = idev->dev;
2894 
2895 	addrconf_join_solict(dev, &ifp->addr);
2896 
2897 	net_srandom(ifp->addr.s6_addr32[3]);
2898 
2899 	read_lock_bh(&idev->lock);
2900 	spin_lock(&ifp->lock);
2901 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
2902 		goto out;
2903 
2904 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2905 	    idev->cnf.accept_dad < 1 ||
2906 	    !(ifp->flags&IFA_F_TENTATIVE) ||
2907 	    ifp->flags & IFA_F_NODAD) {
2908 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2909 		spin_unlock(&ifp->lock);
2910 		read_unlock_bh(&idev->lock);
2911 
2912 		addrconf_dad_completed(ifp);
2913 		return;
2914 	}
2915 
2916 	if (!(idev->if_flags & IF_READY)) {
2917 		spin_unlock(&ifp->lock);
2918 		read_unlock_bh(&idev->lock);
2919 		/*
2920 		 * If the device is not ready:
2921 		 * - keep it tentative if it is a permanent address.
2922 		 * - otherwise, kill it.
2923 		 */
2924 		in6_ifa_hold(ifp);
2925 		addrconf_dad_stop(ifp, 0);
2926 		return;
2927 	}
2928 
2929 	/*
2930 	 * Optimistic nodes can start receiving
2931 	 * Frames right away
2932 	 */
2933 	if (ifp->flags & IFA_F_OPTIMISTIC)
2934 		ip6_ins_rt(ifp->rt);
2935 
2936 	addrconf_dad_kick(ifp);
2937 out:
2938 	spin_unlock(&ifp->lock);
2939 	read_unlock_bh(&idev->lock);
2940 }
2941 
2942 static void addrconf_dad_timer(unsigned long data)
2943 {
2944 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2945 	struct inet6_dev *idev = ifp->idev;
2946 	struct in6_addr mcaddr;
2947 
2948 	if (!ifp->probes && addrconf_dad_end(ifp))
2949 		goto out;
2950 
2951 	read_lock(&idev->lock);
2952 	if (idev->dead || !(idev->if_flags & IF_READY)) {
2953 		read_unlock(&idev->lock);
2954 		goto out;
2955 	}
2956 
2957 	spin_lock(&ifp->lock);
2958 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2959 		spin_unlock(&ifp->lock);
2960 		read_unlock(&idev->lock);
2961 		goto out;
2962 	}
2963 
2964 	if (ifp->probes == 0) {
2965 		/*
2966 		 * DAD was successful
2967 		 */
2968 
2969 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2970 		spin_unlock(&ifp->lock);
2971 		read_unlock(&idev->lock);
2972 
2973 		addrconf_dad_completed(ifp);
2974 
2975 		goto out;
2976 	}
2977 
2978 	ifp->probes--;
2979 	addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2980 	spin_unlock(&ifp->lock);
2981 	read_unlock(&idev->lock);
2982 
2983 	/* send a neighbour solicitation for our addr */
2984 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2985 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2986 out:
2987 	in6_ifa_put(ifp);
2988 }
2989 
2990 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2991 {
2992 	struct net_device *dev = ifp->idev->dev;
2993 
2994 	/*
2995 	 *	Configure the address for reception. Now it is valid.
2996 	 */
2997 
2998 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
2999 
3000 	/* If added prefix is link local and we are prepared to process
3001 	   router advertisements, start sending router solicitations.
3002 	 */
3003 
3004 	if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3005 	     ifp->idev->cnf.accept_ra == 2) &&
3006 	    ifp->idev->cnf.rtr_solicits > 0 &&
3007 	    (dev->flags&IFF_LOOPBACK) == 0 &&
3008 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3009 		/*
3010 		 *	If a host as already performed a random delay
3011 		 *	[...] as part of DAD [...] there is no need
3012 		 *	to delay again before sending the first RS
3013 		 */
3014 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3015 
3016 		spin_lock_bh(&ifp->lock);
3017 		ifp->probes = 1;
3018 		ifp->idev->if_flags |= IF_RS_SENT;
3019 		addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3020 		spin_unlock_bh(&ifp->lock);
3021 	}
3022 }
3023 
3024 static void addrconf_dad_run(struct inet6_dev *idev)
3025 {
3026 	struct inet6_ifaddr *ifp;
3027 
3028 	read_lock_bh(&idev->lock);
3029 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3030 		spin_lock(&ifp->lock);
3031 		if (ifp->flags & IFA_F_TENTATIVE &&
3032 		    ifp->state == INET6_IFADDR_STATE_DAD)
3033 			addrconf_dad_kick(ifp);
3034 		spin_unlock(&ifp->lock);
3035 	}
3036 	read_unlock_bh(&idev->lock);
3037 }
3038 
3039 #ifdef CONFIG_PROC_FS
3040 struct if6_iter_state {
3041 	struct seq_net_private p;
3042 	int bucket;
3043 };
3044 
3045 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3046 {
3047 	struct inet6_ifaddr *ifa = NULL;
3048 	struct if6_iter_state *state = seq->private;
3049 	struct net *net = seq_file_net(seq);
3050 
3051 	for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3052 		struct hlist_node *n;
3053 		hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3054 					 addr_lst)
3055 			if (net_eq(dev_net(ifa->idev->dev), net))
3056 				return ifa;
3057 	}
3058 	return NULL;
3059 }
3060 
3061 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3062 					 struct inet6_ifaddr *ifa)
3063 {
3064 	struct if6_iter_state *state = seq->private;
3065 	struct net *net = seq_file_net(seq);
3066 	struct hlist_node *n = &ifa->addr_lst;
3067 
3068 	hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3069 		if (net_eq(dev_net(ifa->idev->dev), net))
3070 			return ifa;
3071 
3072 	while (++state->bucket < IN6_ADDR_HSIZE) {
3073 		hlist_for_each_entry_rcu_bh(ifa, n,
3074 				     &inet6_addr_lst[state->bucket], addr_lst) {
3075 			if (net_eq(dev_net(ifa->idev->dev), net))
3076 				return ifa;
3077 		}
3078 	}
3079 
3080 	return NULL;
3081 }
3082 
3083 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3084 {
3085 	struct inet6_ifaddr *ifa = if6_get_first(seq);
3086 
3087 	if (ifa)
3088 		while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3089 			--pos;
3090 	return pos ? NULL : ifa;
3091 }
3092 
3093 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3094 	__acquires(rcu_bh)
3095 {
3096 	rcu_read_lock_bh();
3097 	return if6_get_idx(seq, *pos);
3098 }
3099 
3100 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3101 {
3102 	struct inet6_ifaddr *ifa;
3103 
3104 	ifa = if6_get_next(seq, v);
3105 	++*pos;
3106 	return ifa;
3107 }
3108 
3109 static void if6_seq_stop(struct seq_file *seq, void *v)
3110 	__releases(rcu_bh)
3111 {
3112 	rcu_read_unlock_bh();
3113 }
3114 
3115 static int if6_seq_show(struct seq_file *seq, void *v)
3116 {
3117 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3118 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3119 		   &ifp->addr,
3120 		   ifp->idev->dev->ifindex,
3121 		   ifp->prefix_len,
3122 		   ifp->scope,
3123 		   ifp->flags,
3124 		   ifp->idev->dev->name);
3125 	return 0;
3126 }
3127 
3128 static const struct seq_operations if6_seq_ops = {
3129 	.start	= if6_seq_start,
3130 	.next	= if6_seq_next,
3131 	.show	= if6_seq_show,
3132 	.stop	= if6_seq_stop,
3133 };
3134 
3135 static int if6_seq_open(struct inode *inode, struct file *file)
3136 {
3137 	return seq_open_net(inode, file, &if6_seq_ops,
3138 			    sizeof(struct if6_iter_state));
3139 }
3140 
3141 static const struct file_operations if6_fops = {
3142 	.owner		= THIS_MODULE,
3143 	.open		= if6_seq_open,
3144 	.read		= seq_read,
3145 	.llseek		= seq_lseek,
3146 	.release	= seq_release_net,
3147 };
3148 
3149 static int __net_init if6_proc_net_init(struct net *net)
3150 {
3151 	if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3152 		return -ENOMEM;
3153 	return 0;
3154 }
3155 
3156 static void __net_exit if6_proc_net_exit(struct net *net)
3157 {
3158        proc_net_remove(net, "if_inet6");
3159 }
3160 
3161 static struct pernet_operations if6_proc_net_ops = {
3162        .init = if6_proc_net_init,
3163        .exit = if6_proc_net_exit,
3164 };
3165 
3166 int __init if6_proc_init(void)
3167 {
3168 	return register_pernet_subsys(&if6_proc_net_ops);
3169 }
3170 
3171 void if6_proc_exit(void)
3172 {
3173 	unregister_pernet_subsys(&if6_proc_net_ops);
3174 }
3175 #endif	/* CONFIG_PROC_FS */
3176 
3177 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3178 /* Check if address is a home address configured on any interface. */
3179 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3180 {
3181 	int ret = 0;
3182 	struct inet6_ifaddr *ifp = NULL;
3183 	struct hlist_node *n;
3184 	unsigned int hash = ipv6_addr_hash(addr);
3185 
3186 	rcu_read_lock_bh();
3187 	hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3188 		if (!net_eq(dev_net(ifp->idev->dev), net))
3189 			continue;
3190 		if (ipv6_addr_equal(&ifp->addr, addr) &&
3191 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3192 			ret = 1;
3193 			break;
3194 		}
3195 	}
3196 	rcu_read_unlock_bh();
3197 	return ret;
3198 }
3199 #endif
3200 
3201 /*
3202  *	Periodic address status verification
3203  */
3204 
3205 static void addrconf_verify(unsigned long foo)
3206 {
3207 	unsigned long now, next, next_sec, next_sched;
3208 	struct inet6_ifaddr *ifp;
3209 	struct hlist_node *node;
3210 	int i;
3211 
3212 	rcu_read_lock_bh();
3213 	spin_lock(&addrconf_verify_lock);
3214 	now = jiffies;
3215 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3216 
3217 	del_timer(&addr_chk_timer);
3218 
3219 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3220 restart:
3221 		hlist_for_each_entry_rcu_bh(ifp, node,
3222 					 &inet6_addr_lst[i], addr_lst) {
3223 			unsigned long age;
3224 
3225 			if (ifp->flags & IFA_F_PERMANENT)
3226 				continue;
3227 
3228 			spin_lock(&ifp->lock);
3229 			/* We try to batch several events at once. */
3230 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3231 
3232 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3233 			    age >= ifp->valid_lft) {
3234 				spin_unlock(&ifp->lock);
3235 				in6_ifa_hold(ifp);
3236 				ipv6_del_addr(ifp);
3237 				goto restart;
3238 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3239 				spin_unlock(&ifp->lock);
3240 				continue;
3241 			} else if (age >= ifp->prefered_lft) {
3242 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3243 				int deprecate = 0;
3244 
3245 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3246 					deprecate = 1;
3247 					ifp->flags |= IFA_F_DEPRECATED;
3248 				}
3249 
3250 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3251 					next = ifp->tstamp + ifp->valid_lft * HZ;
3252 
3253 				spin_unlock(&ifp->lock);
3254 
3255 				if (deprecate) {
3256 					in6_ifa_hold(ifp);
3257 
3258 					ipv6_ifa_notify(0, ifp);
3259 					in6_ifa_put(ifp);
3260 					goto restart;
3261 				}
3262 #ifdef CONFIG_IPV6_PRIVACY
3263 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3264 				   !(ifp->flags&IFA_F_TENTATIVE)) {
3265 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3266 					ifp->idev->cnf.dad_transmits *
3267 					ifp->idev->nd_parms->retrans_time / HZ;
3268 
3269 				if (age >= ifp->prefered_lft - regen_advance) {
3270 					struct inet6_ifaddr *ifpub = ifp->ifpub;
3271 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3272 						next = ifp->tstamp + ifp->prefered_lft * HZ;
3273 					if (!ifp->regen_count && ifpub) {
3274 						ifp->regen_count++;
3275 						in6_ifa_hold(ifp);
3276 						in6_ifa_hold(ifpub);
3277 						spin_unlock(&ifp->lock);
3278 
3279 						spin_lock(&ifpub->lock);
3280 						ifpub->regen_count = 0;
3281 						spin_unlock(&ifpub->lock);
3282 						ipv6_create_tempaddr(ifpub, ifp);
3283 						in6_ifa_put(ifpub);
3284 						in6_ifa_put(ifp);
3285 						goto restart;
3286 					}
3287 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3288 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3289 				spin_unlock(&ifp->lock);
3290 #endif
3291 			} else {
3292 				/* ifp->prefered_lft <= ifp->valid_lft */
3293 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3294 					next = ifp->tstamp + ifp->prefered_lft * HZ;
3295 				spin_unlock(&ifp->lock);
3296 			}
3297 		}
3298 	}
3299 
3300 	next_sec = round_jiffies_up(next);
3301 	next_sched = next;
3302 
3303 	/* If rounded timeout is accurate enough, accept it. */
3304 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3305 		next_sched = next_sec;
3306 
3307 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3308 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3309 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3310 
3311 	ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3312 	      now, next, next_sec, next_sched));
3313 
3314 	addr_chk_timer.expires = next_sched;
3315 	add_timer(&addr_chk_timer);
3316 	spin_unlock(&addrconf_verify_lock);
3317 	rcu_read_unlock_bh();
3318 }
3319 
3320 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3321 {
3322 	struct in6_addr *pfx = NULL;
3323 
3324 	if (addr)
3325 		pfx = nla_data(addr);
3326 
3327 	if (local) {
3328 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3329 			pfx = NULL;
3330 		else
3331 			pfx = nla_data(local);
3332 	}
3333 
3334 	return pfx;
3335 }
3336 
3337 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3338 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
3339 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
3340 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
3341 };
3342 
3343 static int
3344 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3345 {
3346 	struct net *net = sock_net(skb->sk);
3347 	struct ifaddrmsg *ifm;
3348 	struct nlattr *tb[IFA_MAX+1];
3349 	struct in6_addr *pfx;
3350 	int err;
3351 
3352 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3353 	if (err < 0)
3354 		return err;
3355 
3356 	ifm = nlmsg_data(nlh);
3357 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3358 	if (pfx == NULL)
3359 		return -EINVAL;
3360 
3361 	return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3362 }
3363 
3364 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3365 			     u32 prefered_lft, u32 valid_lft)
3366 {
3367 	u32 flags;
3368 	clock_t expires;
3369 	unsigned long timeout;
3370 
3371 	if (!valid_lft || (prefered_lft > valid_lft))
3372 		return -EINVAL;
3373 
3374 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
3375 	if (addrconf_finite_timeout(timeout)) {
3376 		expires = jiffies_to_clock_t(timeout * HZ);
3377 		valid_lft = timeout;
3378 		flags = RTF_EXPIRES;
3379 	} else {
3380 		expires = 0;
3381 		flags = 0;
3382 		ifa_flags |= IFA_F_PERMANENT;
3383 	}
3384 
3385 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3386 	if (addrconf_finite_timeout(timeout)) {
3387 		if (timeout == 0)
3388 			ifa_flags |= IFA_F_DEPRECATED;
3389 		prefered_lft = timeout;
3390 	}
3391 
3392 	spin_lock_bh(&ifp->lock);
3393 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3394 	ifp->tstamp = jiffies;
3395 	ifp->valid_lft = valid_lft;
3396 	ifp->prefered_lft = prefered_lft;
3397 
3398 	spin_unlock_bh(&ifp->lock);
3399 	if (!(ifp->flags&IFA_F_TENTATIVE))
3400 		ipv6_ifa_notify(0, ifp);
3401 
3402 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3403 			      expires, flags);
3404 	addrconf_verify(0);
3405 
3406 	return 0;
3407 }
3408 
3409 static int
3410 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3411 {
3412 	struct net *net = sock_net(skb->sk);
3413 	struct ifaddrmsg *ifm;
3414 	struct nlattr *tb[IFA_MAX+1];
3415 	struct in6_addr *pfx;
3416 	struct inet6_ifaddr *ifa;
3417 	struct net_device *dev;
3418 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3419 	u8 ifa_flags;
3420 	int err;
3421 
3422 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3423 	if (err < 0)
3424 		return err;
3425 
3426 	ifm = nlmsg_data(nlh);
3427 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3428 	if (pfx == NULL)
3429 		return -EINVAL;
3430 
3431 	if (tb[IFA_CACHEINFO]) {
3432 		struct ifa_cacheinfo *ci;
3433 
3434 		ci = nla_data(tb[IFA_CACHEINFO]);
3435 		valid_lft = ci->ifa_valid;
3436 		preferred_lft = ci->ifa_prefered;
3437 	} else {
3438 		preferred_lft = INFINITY_LIFE_TIME;
3439 		valid_lft = INFINITY_LIFE_TIME;
3440 	}
3441 
3442 	dev =  __dev_get_by_index(net, ifm->ifa_index);
3443 	if (dev == NULL)
3444 		return -ENODEV;
3445 
3446 	/* We ignore other flags so far. */
3447 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3448 
3449 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3450 	if (ifa == NULL) {
3451 		/*
3452 		 * It would be best to check for !NLM_F_CREATE here but
3453 		 * userspace alreay relies on not having to provide this.
3454 		 */
3455 		return inet6_addr_add(net, ifm->ifa_index, pfx,
3456 				      ifm->ifa_prefixlen, ifa_flags,
3457 				      preferred_lft, valid_lft);
3458 	}
3459 
3460 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3461 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3462 		err = -EEXIST;
3463 	else
3464 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3465 
3466 	in6_ifa_put(ifa);
3467 
3468 	return err;
3469 }
3470 
3471 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3472 			  u8 scope, int ifindex)
3473 {
3474 	struct ifaddrmsg *ifm;
3475 
3476 	ifm = nlmsg_data(nlh);
3477 	ifm->ifa_family = AF_INET6;
3478 	ifm->ifa_prefixlen = prefixlen;
3479 	ifm->ifa_flags = flags;
3480 	ifm->ifa_scope = scope;
3481 	ifm->ifa_index = ifindex;
3482 }
3483 
3484 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3485 			 unsigned long tstamp, u32 preferred, u32 valid)
3486 {
3487 	struct ifa_cacheinfo ci;
3488 
3489 	ci.cstamp = cstamp_delta(cstamp);
3490 	ci.tstamp = cstamp_delta(tstamp);
3491 	ci.ifa_prefered = preferred;
3492 	ci.ifa_valid = valid;
3493 
3494 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3495 }
3496 
3497 static inline int rt_scope(int ifa_scope)
3498 {
3499 	if (ifa_scope & IFA_HOST)
3500 		return RT_SCOPE_HOST;
3501 	else if (ifa_scope & IFA_LINK)
3502 		return RT_SCOPE_LINK;
3503 	else if (ifa_scope & IFA_SITE)
3504 		return RT_SCOPE_SITE;
3505 	else
3506 		return RT_SCOPE_UNIVERSE;
3507 }
3508 
3509 static inline int inet6_ifaddr_msgsize(void)
3510 {
3511 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3512 	       + nla_total_size(16) /* IFA_ADDRESS */
3513 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3514 }
3515 
3516 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3517 			     u32 pid, u32 seq, int event, unsigned int flags)
3518 {
3519 	struct nlmsghdr  *nlh;
3520 	u32 preferred, valid;
3521 
3522 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3523 	if (nlh == NULL)
3524 		return -EMSGSIZE;
3525 
3526 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3527 		      ifa->idev->dev->ifindex);
3528 
3529 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3530 		preferred = ifa->prefered_lft;
3531 		valid = ifa->valid_lft;
3532 		if (preferred != INFINITY_LIFE_TIME) {
3533 			long tval = (jiffies - ifa->tstamp)/HZ;
3534 			if (preferred > tval)
3535 				preferred -= tval;
3536 			else
3537 				preferred = 0;
3538 			if (valid != INFINITY_LIFE_TIME) {
3539 				if (valid > tval)
3540 					valid -= tval;
3541 				else
3542 					valid = 0;
3543 			}
3544 		}
3545 	} else {
3546 		preferred = INFINITY_LIFE_TIME;
3547 		valid = INFINITY_LIFE_TIME;
3548 	}
3549 
3550 	if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3551 	    put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3552 		nlmsg_cancel(skb, nlh);
3553 		return -EMSGSIZE;
3554 	}
3555 
3556 	return nlmsg_end(skb, nlh);
3557 }
3558 
3559 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3560 				u32 pid, u32 seq, int event, u16 flags)
3561 {
3562 	struct nlmsghdr  *nlh;
3563 	u8 scope = RT_SCOPE_UNIVERSE;
3564 	int ifindex = ifmca->idev->dev->ifindex;
3565 
3566 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3567 		scope = RT_SCOPE_SITE;
3568 
3569 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3570 	if (nlh == NULL)
3571 		return -EMSGSIZE;
3572 
3573 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3574 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3575 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3576 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3577 		nlmsg_cancel(skb, nlh);
3578 		return -EMSGSIZE;
3579 	}
3580 
3581 	return nlmsg_end(skb, nlh);
3582 }
3583 
3584 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3585 				u32 pid, u32 seq, int event, unsigned int flags)
3586 {
3587 	struct nlmsghdr  *nlh;
3588 	u8 scope = RT_SCOPE_UNIVERSE;
3589 	int ifindex = ifaca->aca_idev->dev->ifindex;
3590 
3591 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3592 		scope = RT_SCOPE_SITE;
3593 
3594 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3595 	if (nlh == NULL)
3596 		return -EMSGSIZE;
3597 
3598 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3599 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3600 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3601 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3602 		nlmsg_cancel(skb, nlh);
3603 		return -EMSGSIZE;
3604 	}
3605 
3606 	return nlmsg_end(skb, nlh);
3607 }
3608 
3609 enum addr_type_t {
3610 	UNICAST_ADDR,
3611 	MULTICAST_ADDR,
3612 	ANYCAST_ADDR,
3613 };
3614 
3615 /* called with rcu_read_lock() */
3616 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3617 			  struct netlink_callback *cb, enum addr_type_t type,
3618 			  int s_ip_idx, int *p_ip_idx)
3619 {
3620 	struct ifmcaddr6 *ifmca;
3621 	struct ifacaddr6 *ifaca;
3622 	int err = 1;
3623 	int ip_idx = *p_ip_idx;
3624 
3625 	read_lock_bh(&idev->lock);
3626 	switch (type) {
3627 	case UNICAST_ADDR: {
3628 		struct inet6_ifaddr *ifa;
3629 
3630 		/* unicast address incl. temp addr */
3631 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
3632 			if (++ip_idx < s_ip_idx)
3633 				continue;
3634 			err = inet6_fill_ifaddr(skb, ifa,
3635 						NETLINK_CB(cb->skb).pid,
3636 						cb->nlh->nlmsg_seq,
3637 						RTM_NEWADDR,
3638 						NLM_F_MULTI);
3639 			if (err <= 0)
3640 				break;
3641 		}
3642 		break;
3643 	}
3644 	case MULTICAST_ADDR:
3645 		/* multicast address */
3646 		for (ifmca = idev->mc_list; ifmca;
3647 		     ifmca = ifmca->next, ip_idx++) {
3648 			if (ip_idx < s_ip_idx)
3649 				continue;
3650 			err = inet6_fill_ifmcaddr(skb, ifmca,
3651 						  NETLINK_CB(cb->skb).pid,
3652 						  cb->nlh->nlmsg_seq,
3653 						  RTM_GETMULTICAST,
3654 						  NLM_F_MULTI);
3655 			if (err <= 0)
3656 				break;
3657 		}
3658 		break;
3659 	case ANYCAST_ADDR:
3660 		/* anycast address */
3661 		for (ifaca = idev->ac_list; ifaca;
3662 		     ifaca = ifaca->aca_next, ip_idx++) {
3663 			if (ip_idx < s_ip_idx)
3664 				continue;
3665 			err = inet6_fill_ifacaddr(skb, ifaca,
3666 						  NETLINK_CB(cb->skb).pid,
3667 						  cb->nlh->nlmsg_seq,
3668 						  RTM_GETANYCAST,
3669 						  NLM_F_MULTI);
3670 			if (err <= 0)
3671 				break;
3672 		}
3673 		break;
3674 	default:
3675 		break;
3676 	}
3677 	read_unlock_bh(&idev->lock);
3678 	*p_ip_idx = ip_idx;
3679 	return err;
3680 }
3681 
3682 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3683 			   enum addr_type_t type)
3684 {
3685 	struct net *net = sock_net(skb->sk);
3686 	int h, s_h;
3687 	int idx, ip_idx;
3688 	int s_idx, s_ip_idx;
3689 	struct net_device *dev;
3690 	struct inet6_dev *idev;
3691 	struct hlist_head *head;
3692 	struct hlist_node *node;
3693 
3694 	s_h = cb->args[0];
3695 	s_idx = idx = cb->args[1];
3696 	s_ip_idx = ip_idx = cb->args[2];
3697 
3698 	rcu_read_lock();
3699 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3700 		idx = 0;
3701 		head = &net->dev_index_head[h];
3702 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3703 			if (idx < s_idx)
3704 				goto cont;
3705 			if (h > s_h || idx > s_idx)
3706 				s_ip_idx = 0;
3707 			ip_idx = 0;
3708 			idev = __in6_dev_get(dev);
3709 			if (!idev)
3710 				goto cont;
3711 
3712 			if (in6_dump_addrs(idev, skb, cb, type,
3713 					   s_ip_idx, &ip_idx) <= 0)
3714 				goto done;
3715 cont:
3716 			idx++;
3717 		}
3718 	}
3719 done:
3720 	rcu_read_unlock();
3721 	cb->args[0] = h;
3722 	cb->args[1] = idx;
3723 	cb->args[2] = ip_idx;
3724 
3725 	return skb->len;
3726 }
3727 
3728 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3729 {
3730 	enum addr_type_t type = UNICAST_ADDR;
3731 
3732 	return inet6_dump_addr(skb, cb, type);
3733 }
3734 
3735 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3736 {
3737 	enum addr_type_t type = MULTICAST_ADDR;
3738 
3739 	return inet6_dump_addr(skb, cb, type);
3740 }
3741 
3742 
3743 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3744 {
3745 	enum addr_type_t type = ANYCAST_ADDR;
3746 
3747 	return inet6_dump_addr(skb, cb, type);
3748 }
3749 
3750 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3751 			     void *arg)
3752 {
3753 	struct net *net = sock_net(in_skb->sk);
3754 	struct ifaddrmsg *ifm;
3755 	struct nlattr *tb[IFA_MAX+1];
3756 	struct in6_addr *addr = NULL;
3757 	struct net_device *dev = NULL;
3758 	struct inet6_ifaddr *ifa;
3759 	struct sk_buff *skb;
3760 	int err;
3761 
3762 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3763 	if (err < 0)
3764 		goto errout;
3765 
3766 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3767 	if (addr == NULL) {
3768 		err = -EINVAL;
3769 		goto errout;
3770 	}
3771 
3772 	ifm = nlmsg_data(nlh);
3773 	if (ifm->ifa_index)
3774 		dev = __dev_get_by_index(net, ifm->ifa_index);
3775 
3776 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3777 	if (!ifa) {
3778 		err = -EADDRNOTAVAIL;
3779 		goto errout;
3780 	}
3781 
3782 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3783 	if (!skb) {
3784 		err = -ENOBUFS;
3785 		goto errout_ifa;
3786 	}
3787 
3788 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3789 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
3790 	if (err < 0) {
3791 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3792 		WARN_ON(err == -EMSGSIZE);
3793 		kfree_skb(skb);
3794 		goto errout_ifa;
3795 	}
3796 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3797 errout_ifa:
3798 	in6_ifa_put(ifa);
3799 errout:
3800 	return err;
3801 }
3802 
3803 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3804 {
3805 	struct sk_buff *skb;
3806 	struct net *net = dev_net(ifa->idev->dev);
3807 	int err = -ENOBUFS;
3808 
3809 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3810 	if (skb == NULL)
3811 		goto errout;
3812 
3813 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3814 	if (err < 0) {
3815 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3816 		WARN_ON(err == -EMSGSIZE);
3817 		kfree_skb(skb);
3818 		goto errout;
3819 	}
3820 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3821 	return;
3822 errout:
3823 	if (err < 0)
3824 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3825 }
3826 
3827 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3828 				__s32 *array, int bytes)
3829 {
3830 	BUG_ON(bytes < (DEVCONF_MAX * 4));
3831 
3832 	memset(array, 0, bytes);
3833 	array[DEVCONF_FORWARDING] = cnf->forwarding;
3834 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3835 	array[DEVCONF_MTU6] = cnf->mtu6;
3836 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3837 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3838 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
3839 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3840 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3841 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3842 		jiffies_to_msecs(cnf->rtr_solicit_interval);
3843 	array[DEVCONF_RTR_SOLICIT_DELAY] =
3844 		jiffies_to_msecs(cnf->rtr_solicit_delay);
3845 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3846 #ifdef CONFIG_IPV6_PRIVACY
3847 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3848 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3849 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3850 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3851 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3852 #endif
3853 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3854 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3855 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3856 #ifdef CONFIG_IPV6_ROUTER_PREF
3857 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3858 	array[DEVCONF_RTR_PROBE_INTERVAL] =
3859 		jiffies_to_msecs(cnf->rtr_probe_interval);
3860 #ifdef CONFIG_IPV6_ROUTE_INFO
3861 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3862 #endif
3863 #endif
3864 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3865 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3866 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3867 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3868 #endif
3869 #ifdef CONFIG_IPV6_MROUTE
3870 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3871 #endif
3872 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3873 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3874 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3875 }
3876 
3877 static inline size_t inet6_ifla6_size(void)
3878 {
3879 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
3880 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
3881 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3882 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3883 	     + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3884 }
3885 
3886 static inline size_t inet6_if_nlmsg_size(void)
3887 {
3888 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3889 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3890 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3891 	       + nla_total_size(4) /* IFLA_MTU */
3892 	       + nla_total_size(4) /* IFLA_LINK */
3893 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3894 }
3895 
3896 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3897 				      int items, int bytes)
3898 {
3899 	int i;
3900 	int pad = bytes - sizeof(u64) * items;
3901 	BUG_ON(pad < 0);
3902 
3903 	/* Use put_unaligned() because stats may not be aligned for u64. */
3904 	put_unaligned(items, &stats[0]);
3905 	for (i = 1; i < items; i++)
3906 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3907 
3908 	memset(&stats[items], 0, pad);
3909 }
3910 
3911 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3912 				      int items, int bytes, size_t syncpoff)
3913 {
3914 	int i;
3915 	int pad = bytes - sizeof(u64) * items;
3916 	BUG_ON(pad < 0);
3917 
3918 	/* Use put_unaligned() because stats may not be aligned for u64. */
3919 	put_unaligned(items, &stats[0]);
3920 	for (i = 1; i < items; i++)
3921 		put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3922 
3923 	memset(&stats[items], 0, pad);
3924 }
3925 
3926 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3927 			     int bytes)
3928 {
3929 	switch (attrtype) {
3930 	case IFLA_INET6_STATS:
3931 		__snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3932 				     IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3933 		break;
3934 	case IFLA_INET6_ICMP6STATS:
3935 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3936 		break;
3937 	}
3938 }
3939 
3940 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3941 {
3942 	struct nlattr *nla;
3943 	struct ifla_cacheinfo ci;
3944 
3945 	NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3946 
3947 	ci.max_reasm_len = IPV6_MAXPLEN;
3948 	ci.tstamp = cstamp_delta(idev->tstamp);
3949 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3950 	ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3951 	NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3952 
3953 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3954 	if (nla == NULL)
3955 		goto nla_put_failure;
3956 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3957 
3958 	/* XXX - MC not implemented */
3959 
3960 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3961 	if (nla == NULL)
3962 		goto nla_put_failure;
3963 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3964 
3965 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3966 	if (nla == NULL)
3967 		goto nla_put_failure;
3968 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3969 
3970 	return 0;
3971 
3972 nla_put_failure:
3973 	return -EMSGSIZE;
3974 }
3975 
3976 static size_t inet6_get_link_af_size(const struct net_device *dev)
3977 {
3978 	if (!__in6_dev_get(dev))
3979 		return 0;
3980 
3981 	return inet6_ifla6_size();
3982 }
3983 
3984 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3985 {
3986 	struct inet6_dev *idev = __in6_dev_get(dev);
3987 
3988 	if (!idev)
3989 		return -ENODATA;
3990 
3991 	if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3992 		return -EMSGSIZE;
3993 
3994 	return 0;
3995 }
3996 
3997 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3998 			     u32 pid, u32 seq, int event, unsigned int flags)
3999 {
4000 	struct net_device *dev = idev->dev;
4001 	struct ifinfomsg *hdr;
4002 	struct nlmsghdr *nlh;
4003 	void *protoinfo;
4004 
4005 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4006 	if (nlh == NULL)
4007 		return -EMSGSIZE;
4008 
4009 	hdr = nlmsg_data(nlh);
4010 	hdr->ifi_family = AF_INET6;
4011 	hdr->__ifi_pad = 0;
4012 	hdr->ifi_type = dev->type;
4013 	hdr->ifi_index = dev->ifindex;
4014 	hdr->ifi_flags = dev_get_flags(dev);
4015 	hdr->ifi_change = 0;
4016 
4017 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4018 
4019 	if (dev->addr_len)
4020 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4021 
4022 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4023 	if (dev->ifindex != dev->iflink)
4024 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4025 
4026 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4027 	if (protoinfo == NULL)
4028 		goto nla_put_failure;
4029 
4030 	if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4031 		goto nla_put_failure;
4032 
4033 	nla_nest_end(skb, protoinfo);
4034 	return nlmsg_end(skb, nlh);
4035 
4036 nla_put_failure:
4037 	nlmsg_cancel(skb, nlh);
4038 	return -EMSGSIZE;
4039 }
4040 
4041 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4042 {
4043 	struct net *net = sock_net(skb->sk);
4044 	int h, s_h;
4045 	int idx = 0, s_idx;
4046 	struct net_device *dev;
4047 	struct inet6_dev *idev;
4048 	struct hlist_head *head;
4049 	struct hlist_node *node;
4050 
4051 	s_h = cb->args[0];
4052 	s_idx = cb->args[1];
4053 
4054 	rcu_read_lock();
4055 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4056 		idx = 0;
4057 		head = &net->dev_index_head[h];
4058 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4059 			if (idx < s_idx)
4060 				goto cont;
4061 			idev = __in6_dev_get(dev);
4062 			if (!idev)
4063 				goto cont;
4064 			if (inet6_fill_ifinfo(skb, idev,
4065 					      NETLINK_CB(cb->skb).pid,
4066 					      cb->nlh->nlmsg_seq,
4067 					      RTM_NEWLINK, NLM_F_MULTI) <= 0)
4068 				goto out;
4069 cont:
4070 			idx++;
4071 		}
4072 	}
4073 out:
4074 	rcu_read_unlock();
4075 	cb->args[1] = idx;
4076 	cb->args[0] = h;
4077 
4078 	return skb->len;
4079 }
4080 
4081 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4082 {
4083 	struct sk_buff *skb;
4084 	struct net *net = dev_net(idev->dev);
4085 	int err = -ENOBUFS;
4086 
4087 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4088 	if (skb == NULL)
4089 		goto errout;
4090 
4091 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4092 	if (err < 0) {
4093 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4094 		WARN_ON(err == -EMSGSIZE);
4095 		kfree_skb(skb);
4096 		goto errout;
4097 	}
4098 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4099 	return;
4100 errout:
4101 	if (err < 0)
4102 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4103 }
4104 
4105 static inline size_t inet6_prefix_nlmsg_size(void)
4106 {
4107 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
4108 	       + nla_total_size(sizeof(struct in6_addr))
4109 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
4110 }
4111 
4112 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4113 			     struct prefix_info *pinfo, u32 pid, u32 seq,
4114 			     int event, unsigned int flags)
4115 {
4116 	struct prefixmsg *pmsg;
4117 	struct nlmsghdr *nlh;
4118 	struct prefix_cacheinfo	ci;
4119 
4120 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4121 	if (nlh == NULL)
4122 		return -EMSGSIZE;
4123 
4124 	pmsg = nlmsg_data(nlh);
4125 	pmsg->prefix_family = AF_INET6;
4126 	pmsg->prefix_pad1 = 0;
4127 	pmsg->prefix_pad2 = 0;
4128 	pmsg->prefix_ifindex = idev->dev->ifindex;
4129 	pmsg->prefix_len = pinfo->prefix_len;
4130 	pmsg->prefix_type = pinfo->type;
4131 	pmsg->prefix_pad3 = 0;
4132 	pmsg->prefix_flags = 0;
4133 	if (pinfo->onlink)
4134 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4135 	if (pinfo->autoconf)
4136 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4137 
4138 	NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4139 
4140 	ci.preferred_time = ntohl(pinfo->prefered);
4141 	ci.valid_time = ntohl(pinfo->valid);
4142 	NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4143 
4144 	return nlmsg_end(skb, nlh);
4145 
4146 nla_put_failure:
4147 	nlmsg_cancel(skb, nlh);
4148 	return -EMSGSIZE;
4149 }
4150 
4151 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4152 			 struct prefix_info *pinfo)
4153 {
4154 	struct sk_buff *skb;
4155 	struct net *net = dev_net(idev->dev);
4156 	int err = -ENOBUFS;
4157 
4158 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4159 	if (skb == NULL)
4160 		goto errout;
4161 
4162 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4163 	if (err < 0) {
4164 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4165 		WARN_ON(err == -EMSGSIZE);
4166 		kfree_skb(skb);
4167 		goto errout;
4168 	}
4169 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4170 	return;
4171 errout:
4172 	if (err < 0)
4173 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4174 }
4175 
4176 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4177 {
4178 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4179 
4180 	switch (event) {
4181 	case RTM_NEWADDR:
4182 		/*
4183 		 * If the address was optimistic
4184 		 * we inserted the route at the start of
4185 		 * our DAD process, so we don't need
4186 		 * to do it again
4187 		 */
4188 		if (!(ifp->rt->rt6i_node))
4189 			ip6_ins_rt(ifp->rt);
4190 		if (ifp->idev->cnf.forwarding)
4191 			addrconf_join_anycast(ifp);
4192 		break;
4193 	case RTM_DELADDR:
4194 		if (ifp->idev->cnf.forwarding)
4195 			addrconf_leave_anycast(ifp);
4196 		addrconf_leave_solict(ifp->idev, &ifp->addr);
4197 		dst_hold(&ifp->rt->dst);
4198 
4199 		if (ip6_del_rt(ifp->rt))
4200 			dst_free(&ifp->rt->dst);
4201 		break;
4202 	}
4203 }
4204 
4205 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4206 {
4207 	rcu_read_lock_bh();
4208 	if (likely(ifp->idev->dead == 0))
4209 		__ipv6_ifa_notify(event, ifp);
4210 	rcu_read_unlock_bh();
4211 }
4212 
4213 #ifdef CONFIG_SYSCTL
4214 
4215 static
4216 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4217 			   void __user *buffer, size_t *lenp, loff_t *ppos)
4218 {
4219 	int *valp = ctl->data;
4220 	int val = *valp;
4221 	loff_t pos = *ppos;
4222 	int ret;
4223 
4224 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4225 
4226 	if (write)
4227 		ret = addrconf_fixup_forwarding(ctl, valp, val);
4228 	if (ret)
4229 		*ppos = pos;
4230 	return ret;
4231 }
4232 
4233 static void dev_disable_change(struct inet6_dev *idev)
4234 {
4235 	if (!idev || !idev->dev)
4236 		return;
4237 
4238 	if (idev->cnf.disable_ipv6)
4239 		addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4240 	else
4241 		addrconf_notify(NULL, NETDEV_UP, idev->dev);
4242 }
4243 
4244 static void addrconf_disable_change(struct net *net, __s32 newf)
4245 {
4246 	struct net_device *dev;
4247 	struct inet6_dev *idev;
4248 
4249 	rcu_read_lock();
4250 	for_each_netdev_rcu(net, dev) {
4251 		idev = __in6_dev_get(dev);
4252 		if (idev) {
4253 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4254 			idev->cnf.disable_ipv6 = newf;
4255 			if (changed)
4256 				dev_disable_change(idev);
4257 		}
4258 	}
4259 	rcu_read_unlock();
4260 }
4261 
4262 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4263 {
4264 	struct net *net;
4265 
4266 	net = (struct net *)table->extra2;
4267 
4268 	if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4269 		return 0;
4270 
4271 	if (!rtnl_trylock()) {
4272 		/* Restore the original values before restarting */
4273 		*p = old;
4274 		return restart_syscall();
4275 	}
4276 
4277 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
4278 		__s32 newf = net->ipv6.devconf_all->disable_ipv6;
4279 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
4280 		addrconf_disable_change(net, newf);
4281 	} else if ((!*p) ^ (!old))
4282 		dev_disable_change((struct inet6_dev *)table->extra1);
4283 
4284 	rtnl_unlock();
4285 	return 0;
4286 }
4287 
4288 static
4289 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4290 			    void __user *buffer, size_t *lenp, loff_t *ppos)
4291 {
4292 	int *valp = ctl->data;
4293 	int val = *valp;
4294 	loff_t pos = *ppos;
4295 	int ret;
4296 
4297 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4298 
4299 	if (write)
4300 		ret = addrconf_disable_ipv6(ctl, valp, val);
4301 	if (ret)
4302 		*ppos = pos;
4303 	return ret;
4304 }
4305 
4306 static struct addrconf_sysctl_table
4307 {
4308 	struct ctl_table_header *sysctl_header;
4309 	ctl_table addrconf_vars[DEVCONF_MAX+1];
4310 	char *dev_name;
4311 } addrconf_sysctl __read_mostly = {
4312 	.sysctl_header = NULL,
4313 	.addrconf_vars = {
4314 		{
4315 			.procname	= "forwarding",
4316 			.data		= &ipv6_devconf.forwarding,
4317 			.maxlen		= sizeof(int),
4318 			.mode		= 0644,
4319 			.proc_handler	= addrconf_sysctl_forward,
4320 		},
4321 		{
4322 			.procname	= "hop_limit",
4323 			.data		= &ipv6_devconf.hop_limit,
4324 			.maxlen		= sizeof(int),
4325 			.mode		= 0644,
4326 			.proc_handler	= proc_dointvec,
4327 		},
4328 		{
4329 			.procname	= "mtu",
4330 			.data		= &ipv6_devconf.mtu6,
4331 			.maxlen		= sizeof(int),
4332 			.mode		= 0644,
4333 			.proc_handler	= proc_dointvec,
4334 		},
4335 		{
4336 			.procname	= "accept_ra",
4337 			.data		= &ipv6_devconf.accept_ra,
4338 			.maxlen		= sizeof(int),
4339 			.mode		= 0644,
4340 			.proc_handler	= proc_dointvec,
4341 		},
4342 		{
4343 			.procname	= "accept_redirects",
4344 			.data		= &ipv6_devconf.accept_redirects,
4345 			.maxlen		= sizeof(int),
4346 			.mode		= 0644,
4347 			.proc_handler	= proc_dointvec,
4348 		},
4349 		{
4350 			.procname	= "autoconf",
4351 			.data		= &ipv6_devconf.autoconf,
4352 			.maxlen		= sizeof(int),
4353 			.mode		= 0644,
4354 			.proc_handler	= proc_dointvec,
4355 		},
4356 		{
4357 			.procname	= "dad_transmits",
4358 			.data		= &ipv6_devconf.dad_transmits,
4359 			.maxlen		= sizeof(int),
4360 			.mode		= 0644,
4361 			.proc_handler	= proc_dointvec,
4362 		},
4363 		{
4364 			.procname	= "router_solicitations",
4365 			.data		= &ipv6_devconf.rtr_solicits,
4366 			.maxlen		= sizeof(int),
4367 			.mode		= 0644,
4368 			.proc_handler	= proc_dointvec,
4369 		},
4370 		{
4371 			.procname	= "router_solicitation_interval",
4372 			.data		= &ipv6_devconf.rtr_solicit_interval,
4373 			.maxlen		= sizeof(int),
4374 			.mode		= 0644,
4375 			.proc_handler	= proc_dointvec_jiffies,
4376 		},
4377 		{
4378 			.procname	= "router_solicitation_delay",
4379 			.data		= &ipv6_devconf.rtr_solicit_delay,
4380 			.maxlen		= sizeof(int),
4381 			.mode		= 0644,
4382 			.proc_handler	= proc_dointvec_jiffies,
4383 		},
4384 		{
4385 			.procname	= "force_mld_version",
4386 			.data		= &ipv6_devconf.force_mld_version,
4387 			.maxlen		= sizeof(int),
4388 			.mode		= 0644,
4389 			.proc_handler	= proc_dointvec,
4390 		},
4391 #ifdef CONFIG_IPV6_PRIVACY
4392 		{
4393 			.procname	= "use_tempaddr",
4394 			.data		= &ipv6_devconf.use_tempaddr,
4395 			.maxlen		= sizeof(int),
4396 			.mode		= 0644,
4397 			.proc_handler	= proc_dointvec,
4398 		},
4399 		{
4400 			.procname	= "temp_valid_lft",
4401 			.data		= &ipv6_devconf.temp_valid_lft,
4402 			.maxlen		= sizeof(int),
4403 			.mode		= 0644,
4404 			.proc_handler	= proc_dointvec,
4405 		},
4406 		{
4407 			.procname	= "temp_prefered_lft",
4408 			.data		= &ipv6_devconf.temp_prefered_lft,
4409 			.maxlen		= sizeof(int),
4410 			.mode		= 0644,
4411 			.proc_handler	= proc_dointvec,
4412 		},
4413 		{
4414 			.procname	= "regen_max_retry",
4415 			.data		= &ipv6_devconf.regen_max_retry,
4416 			.maxlen		= sizeof(int),
4417 			.mode		= 0644,
4418 			.proc_handler	= proc_dointvec,
4419 		},
4420 		{
4421 			.procname	= "max_desync_factor",
4422 			.data		= &ipv6_devconf.max_desync_factor,
4423 			.maxlen		= sizeof(int),
4424 			.mode		= 0644,
4425 			.proc_handler	= proc_dointvec,
4426 		},
4427 #endif
4428 		{
4429 			.procname	= "max_addresses",
4430 			.data		= &ipv6_devconf.max_addresses,
4431 			.maxlen		= sizeof(int),
4432 			.mode		= 0644,
4433 			.proc_handler	= proc_dointvec,
4434 		},
4435 		{
4436 			.procname	= "accept_ra_defrtr",
4437 			.data		= &ipv6_devconf.accept_ra_defrtr,
4438 			.maxlen		= sizeof(int),
4439 			.mode		= 0644,
4440 			.proc_handler	= proc_dointvec,
4441 		},
4442 		{
4443 			.procname	= "accept_ra_pinfo",
4444 			.data		= &ipv6_devconf.accept_ra_pinfo,
4445 			.maxlen		= sizeof(int),
4446 			.mode		= 0644,
4447 			.proc_handler	= proc_dointvec,
4448 		},
4449 #ifdef CONFIG_IPV6_ROUTER_PREF
4450 		{
4451 			.procname	= "accept_ra_rtr_pref",
4452 			.data		= &ipv6_devconf.accept_ra_rtr_pref,
4453 			.maxlen		= sizeof(int),
4454 			.mode		= 0644,
4455 			.proc_handler	= proc_dointvec,
4456 		},
4457 		{
4458 			.procname	= "router_probe_interval",
4459 			.data		= &ipv6_devconf.rtr_probe_interval,
4460 			.maxlen		= sizeof(int),
4461 			.mode		= 0644,
4462 			.proc_handler	= proc_dointvec_jiffies,
4463 		},
4464 #ifdef CONFIG_IPV6_ROUTE_INFO
4465 		{
4466 			.procname	= "accept_ra_rt_info_max_plen",
4467 			.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
4468 			.maxlen		= sizeof(int),
4469 			.mode		= 0644,
4470 			.proc_handler	= proc_dointvec,
4471 		},
4472 #endif
4473 #endif
4474 		{
4475 			.procname	= "proxy_ndp",
4476 			.data		= &ipv6_devconf.proxy_ndp,
4477 			.maxlen		= sizeof(int),
4478 			.mode		= 0644,
4479 			.proc_handler	= proc_dointvec,
4480 		},
4481 		{
4482 			.procname	= "accept_source_route",
4483 			.data		= &ipv6_devconf.accept_source_route,
4484 			.maxlen		= sizeof(int),
4485 			.mode		= 0644,
4486 			.proc_handler	= proc_dointvec,
4487 		},
4488 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4489 		{
4490 			.procname       = "optimistic_dad",
4491 			.data           = &ipv6_devconf.optimistic_dad,
4492 			.maxlen         = sizeof(int),
4493 			.mode           = 0644,
4494 			.proc_handler   = proc_dointvec,
4495 
4496 		},
4497 #endif
4498 #ifdef CONFIG_IPV6_MROUTE
4499 		{
4500 			.procname	= "mc_forwarding",
4501 			.data		= &ipv6_devconf.mc_forwarding,
4502 			.maxlen		= sizeof(int),
4503 			.mode		= 0444,
4504 			.proc_handler	= proc_dointvec,
4505 		},
4506 #endif
4507 		{
4508 			.procname	= "disable_ipv6",
4509 			.data		= &ipv6_devconf.disable_ipv6,
4510 			.maxlen		= sizeof(int),
4511 			.mode		= 0644,
4512 			.proc_handler	= addrconf_sysctl_disable,
4513 		},
4514 		{
4515 			.procname	= "accept_dad",
4516 			.data		= &ipv6_devconf.accept_dad,
4517 			.maxlen		= sizeof(int),
4518 			.mode		= 0644,
4519 			.proc_handler	= proc_dointvec,
4520 		},
4521 		{
4522 			.procname       = "force_tllao",
4523 			.data           = &ipv6_devconf.force_tllao,
4524 			.maxlen         = sizeof(int),
4525 			.mode           = 0644,
4526 			.proc_handler   = proc_dointvec
4527 		},
4528 		{
4529 			/* sentinel */
4530 		}
4531 	},
4532 };
4533 
4534 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4535 		struct inet6_dev *idev, struct ipv6_devconf *p)
4536 {
4537 	int i;
4538 	struct addrconf_sysctl_table *t;
4539 
4540 #define ADDRCONF_CTL_PATH_DEV	3
4541 
4542 	struct ctl_path addrconf_ctl_path[] = {
4543 		{ .procname = "net", },
4544 		{ .procname = "ipv6", },
4545 		{ .procname = "conf", },
4546 		{ /* to be set */ },
4547 		{ },
4548 	};
4549 
4550 
4551 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4552 	if (t == NULL)
4553 		goto out;
4554 
4555 	for (i = 0; t->addrconf_vars[i].data; i++) {
4556 		t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4557 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4558 		t->addrconf_vars[i].extra2 = net;
4559 	}
4560 
4561 	/*
4562 	 * Make a copy of dev_name, because '.procname' is regarded as const
4563 	 * by sysctl and we wouldn't want anyone to change it under our feet
4564 	 * (see SIOCSIFNAME).
4565 	 */
4566 	t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4567 	if (!t->dev_name)
4568 		goto free;
4569 
4570 	addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4571 
4572 	t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4573 			t->addrconf_vars);
4574 	if (t->sysctl_header == NULL)
4575 		goto free_procname;
4576 
4577 	p->sysctl = t;
4578 	return 0;
4579 
4580 free_procname:
4581 	kfree(t->dev_name);
4582 free:
4583 	kfree(t);
4584 out:
4585 	return -ENOBUFS;
4586 }
4587 
4588 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4589 {
4590 	struct addrconf_sysctl_table *t;
4591 
4592 	if (p->sysctl == NULL)
4593 		return;
4594 
4595 	t = p->sysctl;
4596 	p->sysctl = NULL;
4597 	unregister_net_sysctl_table(t->sysctl_header);
4598 	kfree(t->dev_name);
4599 	kfree(t);
4600 }
4601 
4602 static void addrconf_sysctl_register(struct inet6_dev *idev)
4603 {
4604 	neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4605 			      &ndisc_ifinfo_sysctl_change);
4606 	__addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4607 					idev, &idev->cnf);
4608 }
4609 
4610 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4611 {
4612 	__addrconf_sysctl_unregister(&idev->cnf);
4613 	neigh_sysctl_unregister(idev->nd_parms);
4614 }
4615 
4616 
4617 #endif
4618 
4619 static int __net_init addrconf_init_net(struct net *net)
4620 {
4621 	int err;
4622 	struct ipv6_devconf *all, *dflt;
4623 
4624 	err = -ENOMEM;
4625 	all = &ipv6_devconf;
4626 	dflt = &ipv6_devconf_dflt;
4627 
4628 	if (!net_eq(net, &init_net)) {
4629 		all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4630 		if (all == NULL)
4631 			goto err_alloc_all;
4632 
4633 		dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4634 		if (dflt == NULL)
4635 			goto err_alloc_dflt;
4636 	} else {
4637 		/* these will be inherited by all namespaces */
4638 		dflt->autoconf = ipv6_defaults.autoconf;
4639 		dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4640 	}
4641 
4642 	net->ipv6.devconf_all = all;
4643 	net->ipv6.devconf_dflt = dflt;
4644 
4645 #ifdef CONFIG_SYSCTL
4646 	err = __addrconf_sysctl_register(net, "all", NULL, all);
4647 	if (err < 0)
4648 		goto err_reg_all;
4649 
4650 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4651 	if (err < 0)
4652 		goto err_reg_dflt;
4653 #endif
4654 	return 0;
4655 
4656 #ifdef CONFIG_SYSCTL
4657 err_reg_dflt:
4658 	__addrconf_sysctl_unregister(all);
4659 err_reg_all:
4660 	kfree(dflt);
4661 #endif
4662 err_alloc_dflt:
4663 	kfree(all);
4664 err_alloc_all:
4665 	return err;
4666 }
4667 
4668 static void __net_exit addrconf_exit_net(struct net *net)
4669 {
4670 #ifdef CONFIG_SYSCTL
4671 	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4672 	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
4673 #endif
4674 	if (!net_eq(net, &init_net)) {
4675 		kfree(net->ipv6.devconf_dflt);
4676 		kfree(net->ipv6.devconf_all);
4677 	}
4678 }
4679 
4680 static struct pernet_operations addrconf_ops = {
4681 	.init = addrconf_init_net,
4682 	.exit = addrconf_exit_net,
4683 };
4684 
4685 /*
4686  *      Device notifier
4687  */
4688 
4689 int register_inet6addr_notifier(struct notifier_block *nb)
4690 {
4691 	return atomic_notifier_chain_register(&inet6addr_chain, nb);
4692 }
4693 EXPORT_SYMBOL(register_inet6addr_notifier);
4694 
4695 int unregister_inet6addr_notifier(struct notifier_block *nb)
4696 {
4697 	return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4698 }
4699 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4700 
4701 static struct rtnl_af_ops inet6_ops = {
4702 	.family		  = AF_INET6,
4703 	.fill_link_af	  = inet6_fill_link_af,
4704 	.get_link_af_size = inet6_get_link_af_size,
4705 };
4706 
4707 /*
4708  *	Init / cleanup code
4709  */
4710 
4711 int __init addrconf_init(void)
4712 {
4713 	int i, err;
4714 
4715 	err = ipv6_addr_label_init();
4716 	if (err < 0) {
4717 		printk(KERN_CRIT "IPv6 Addrconf:"
4718 		       " cannot initialize default policy table: %d.\n", err);
4719 		goto out;
4720 	}
4721 
4722 	err = register_pernet_subsys(&addrconf_ops);
4723 	if (err < 0)
4724 		goto out_addrlabel;
4725 
4726 	/* The addrconf netdev notifier requires that loopback_dev
4727 	 * has it's ipv6 private information allocated and setup
4728 	 * before it can bring up and give link-local addresses
4729 	 * to other devices which are up.
4730 	 *
4731 	 * Unfortunately, loopback_dev is not necessarily the first
4732 	 * entry in the global dev_base list of net devices.  In fact,
4733 	 * it is likely to be the very last entry on that list.
4734 	 * So this causes the notifier registry below to try and
4735 	 * give link-local addresses to all devices besides loopback_dev
4736 	 * first, then loopback_dev, which cases all the non-loopback_dev
4737 	 * devices to fail to get a link-local address.
4738 	 *
4739 	 * So, as a temporary fix, allocate the ipv6 structure for
4740 	 * loopback_dev first by hand.
4741 	 * Longer term, all of the dependencies ipv6 has upon the loopback
4742 	 * device and it being up should be removed.
4743 	 */
4744 	rtnl_lock();
4745 	if (!ipv6_add_dev(init_net.loopback_dev))
4746 		err = -ENOMEM;
4747 	rtnl_unlock();
4748 	if (err)
4749 		goto errlo;
4750 
4751 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4752 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4753 
4754 	register_netdevice_notifier(&ipv6_dev_notf);
4755 
4756 	addrconf_verify(0);
4757 
4758 	err = rtnl_af_register(&inet6_ops);
4759 	if (err < 0)
4760 		goto errout_af;
4761 
4762 	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4763 			      NULL);
4764 	if (err < 0)
4765 		goto errout;
4766 
4767 	/* Only the first call to __rtnl_register can fail */
4768 	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4769 	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4770 	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4771 			inet6_dump_ifaddr, NULL);
4772 	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4773 			inet6_dump_ifmcaddr, NULL);
4774 	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4775 			inet6_dump_ifacaddr, NULL);
4776 
4777 	ipv6_addr_label_rtnl_register();
4778 
4779 	return 0;
4780 errout:
4781 	rtnl_af_unregister(&inet6_ops);
4782 errout_af:
4783 	unregister_netdevice_notifier(&ipv6_dev_notf);
4784 errlo:
4785 	unregister_pernet_subsys(&addrconf_ops);
4786 out_addrlabel:
4787 	ipv6_addr_label_cleanup();
4788 out:
4789 	return err;
4790 }
4791 
4792 void addrconf_cleanup(void)
4793 {
4794 	struct net_device *dev;
4795 	int i;
4796 
4797 	unregister_netdevice_notifier(&ipv6_dev_notf);
4798 	unregister_pernet_subsys(&addrconf_ops);
4799 	ipv6_addr_label_cleanup();
4800 
4801 	rtnl_lock();
4802 
4803 	__rtnl_af_unregister(&inet6_ops);
4804 
4805 	/* clean dev list */
4806 	for_each_netdev(&init_net, dev) {
4807 		if (__in6_dev_get(dev) == NULL)
4808 			continue;
4809 		addrconf_ifdown(dev, 1);
4810 	}
4811 	addrconf_ifdown(init_net.loopback_dev, 2);
4812 
4813 	/*
4814 	 *	Check hash table.
4815 	 */
4816 	spin_lock_bh(&addrconf_hash_lock);
4817 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4818 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4819 	spin_unlock_bh(&addrconf_hash_lock);
4820 
4821 	del_timer(&addr_chk_timer);
4822 	rtnl_unlock();
4823 }
4824