xref: /openbmc/linux/net/ipv6/addrconf.c (revision f2344a131bccdbfc5338e17fa71a807dee7944fa)
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 
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93 
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99 
100 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
101 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
102 
103 #define ADDRCONF_TIMER_FUZZ_MINUS	(HZ > 50 ? HZ/50 : 1)
104 #define ADDRCONF_TIMER_FUZZ		(HZ / 4)
105 #define ADDRCONF_TIMER_FUZZ_MAX		(HZ)
106 
107 #ifdef CONFIG_SYSCTL
108 static void addrconf_sysctl_register(struct inet6_dev *idev);
109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
110 #else
111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
112 {
113 }
114 
115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
116 {
117 }
118 #endif
119 
120 #ifdef CONFIG_IPV6_PRIVACY
121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
123 static void ipv6_regen_rndid(unsigned long data);
124 
125 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
126 #endif
127 
128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
129 static int ipv6_count_addresses(struct inet6_dev *idev);
130 
131 /*
132  *	Configured unicast address hash table
133  */
134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
135 static DEFINE_SPINLOCK(addrconf_hash_lock);
136 
137 static void addrconf_verify(unsigned long);
138 
139 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
140 static DEFINE_SPINLOCK(addrconf_verify_lock);
141 
142 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
143 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
144 
145 static void addrconf_type_change(struct net_device *dev,
146 				 unsigned long event);
147 static int addrconf_ifdown(struct net_device *dev, int how);
148 
149 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
150 static void addrconf_dad_timer(unsigned long data);
151 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
152 static void addrconf_dad_run(struct inet6_dev *idev);
153 static void addrconf_rs_timer(unsigned long data);
154 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
155 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
156 
157 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
158 				struct prefix_info *pinfo);
159 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
160 			       struct net_device *dev);
161 
162 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
163 
164 static struct ipv6_devconf ipv6_devconf __read_mostly = {
165 	.forwarding		= 0,
166 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
167 	.mtu6			= IPV6_MIN_MTU,
168 	.accept_ra		= 1,
169 	.accept_redirects	= 1,
170 	.autoconf		= 1,
171 	.force_mld_version	= 0,
172 	.dad_transmits		= 1,
173 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
174 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
175 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
176 #ifdef CONFIG_IPV6_PRIVACY
177 	.use_tempaddr 		= 0,
178 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
179 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
180 	.regen_max_retry	= REGEN_MAX_RETRY,
181 	.max_desync_factor	= MAX_DESYNC_FACTOR,
182 #endif
183 	.max_addresses		= IPV6_MAX_ADDRESSES,
184 	.accept_ra_defrtr	= 1,
185 	.accept_ra_pinfo	= 1,
186 #ifdef CONFIG_IPV6_ROUTER_PREF
187 	.accept_ra_rtr_pref	= 1,
188 	.rtr_probe_interval	= 60 * HZ,
189 #ifdef CONFIG_IPV6_ROUTE_INFO
190 	.accept_ra_rt_info_max_plen = 0,
191 #endif
192 #endif
193 	.proxy_ndp		= 0,
194 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
195 	.disable_ipv6		= 0,
196 	.accept_dad		= 1,
197 };
198 
199 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
200 	.forwarding		= 0,
201 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
202 	.mtu6			= IPV6_MIN_MTU,
203 	.accept_ra		= 1,
204 	.accept_redirects	= 1,
205 	.autoconf		= 1,
206 	.dad_transmits		= 1,
207 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
208 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
209 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
210 #ifdef CONFIG_IPV6_PRIVACY
211 	.use_tempaddr		= 0,
212 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
213 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
214 	.regen_max_retry	= REGEN_MAX_RETRY,
215 	.max_desync_factor	= MAX_DESYNC_FACTOR,
216 #endif
217 	.max_addresses		= IPV6_MAX_ADDRESSES,
218 	.accept_ra_defrtr	= 1,
219 	.accept_ra_pinfo	= 1,
220 #ifdef CONFIG_IPV6_ROUTER_PREF
221 	.accept_ra_rtr_pref	= 1,
222 	.rtr_probe_interval	= 60 * HZ,
223 #ifdef CONFIG_IPV6_ROUTE_INFO
224 	.accept_ra_rt_info_max_plen = 0,
225 #endif
226 #endif
227 	.proxy_ndp		= 0,
228 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
229 	.disable_ipv6		= 0,
230 	.accept_dad		= 1,
231 };
232 
233 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
234 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
235 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
236 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
237 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
238 
239 /* Check if a valid qdisc is available */
240 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
241 {
242 	return !qdisc_tx_is_noop(dev);
243 }
244 
245 /* Check if a route is valid prefix route */
246 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
247 {
248 	return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
249 }
250 
251 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
252 {
253 	if (del_timer(&ifp->timer))
254 		__in6_ifa_put(ifp);
255 }
256 
257 enum addrconf_timer_t {
258 	AC_NONE,
259 	AC_DAD,
260 	AC_RS,
261 };
262 
263 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
264 			       enum addrconf_timer_t what,
265 			       unsigned long when)
266 {
267 	if (!del_timer(&ifp->timer))
268 		in6_ifa_hold(ifp);
269 
270 	switch (what) {
271 	case AC_DAD:
272 		ifp->timer.function = addrconf_dad_timer;
273 		break;
274 	case AC_RS:
275 		ifp->timer.function = addrconf_rs_timer;
276 		break;
277 	default:
278 		break;
279 	}
280 	ifp->timer.expires = jiffies + when;
281 	add_timer(&ifp->timer);
282 }
283 
284 static int snmp6_alloc_dev(struct inet6_dev *idev)
285 {
286 	if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
287 			  sizeof(struct ipstats_mib)) < 0)
288 		goto err_ip;
289 	if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
290 			  sizeof(struct icmpv6_mib)) < 0)
291 		goto err_icmp;
292 	if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
293 			  sizeof(struct icmpv6msg_mib)) < 0)
294 		goto err_icmpmsg;
295 
296 	return 0;
297 
298 err_icmpmsg:
299 	snmp_mib_free((void __percpu **)idev->stats.icmpv6);
300 err_icmp:
301 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
302 err_ip:
303 	return -ENOMEM;
304 }
305 
306 static void snmp6_free_dev(struct inet6_dev *idev)
307 {
308 	snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
309 	snmp_mib_free((void __percpu **)idev->stats.icmpv6);
310 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
311 }
312 
313 /* Nobody refers to this device, we may destroy it. */
314 
315 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
316 {
317 	struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
318 	kfree(idev);
319 }
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 	call_rcu(&idev->rcu, in6_dev_finish_destroy_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 		ndev->dead = 1;
379 		in6_dev_finish_destroy(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 		printk(KERN_INFO
419 		       "%s: Disabled Privacy Extensions\n",
420 		       dev->name);
421 		ndev->cnf.use_tempaddr = -1;
422 	} else {
423 		in6_dev_hold(ndev);
424 		ipv6_regen_rndid((unsigned long) ndev);
425 	}
426 #endif
427 
428 	if (netif_running(dev) && addrconf_qdisc_ok(dev))
429 		ndev->if_flags |= IF_READY;
430 
431 	ipv6_mc_init_dev(ndev);
432 	ndev->tstamp = jiffies;
433 	addrconf_sysctl_register(ndev);
434 	/* protected by rtnl_lock */
435 	rcu_assign_pointer(dev->ip6_ptr, ndev);
436 
437 	/* Join all-node multicast group */
438 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439 
440 	return ndev;
441 }
442 
443 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
444 {
445 	struct inet6_dev *idev;
446 
447 	ASSERT_RTNL();
448 
449 	idev = __in6_dev_get(dev);
450 	if (!idev) {
451 		idev = ipv6_add_dev(dev);
452 		if (!idev)
453 			return NULL;
454 	}
455 
456 	if (dev->flags&IFF_UP)
457 		ipv6_mc_up(idev);
458 	return idev;
459 }
460 
461 #ifdef CONFIG_SYSCTL
462 static void dev_forward_change(struct inet6_dev *idev)
463 {
464 	struct net_device *dev;
465 	struct inet6_ifaddr *ifa;
466 
467 	if (!idev)
468 		return;
469 	dev = idev->dev;
470 	if (idev->cnf.forwarding)
471 		dev_disable_lro(dev);
472 	if (dev && (dev->flags & IFF_MULTICAST)) {
473 		if (idev->cnf.forwarding)
474 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
475 		else
476 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
477 	}
478 
479 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
480 		if (ifa->flags&IFA_F_TENTATIVE)
481 			continue;
482 		if (idev->cnf.forwarding)
483 			addrconf_join_anycast(ifa);
484 		else
485 			addrconf_leave_anycast(ifa);
486 	}
487 }
488 
489 
490 static void addrconf_forward_change(struct net *net, __s32 newf)
491 {
492 	struct net_device *dev;
493 	struct inet6_dev *idev;
494 
495 	rcu_read_lock();
496 	for_each_netdev_rcu(net, dev) {
497 		idev = __in6_dev_get(dev);
498 		if (idev) {
499 			int changed = (!idev->cnf.forwarding) ^ (!newf);
500 			idev->cnf.forwarding = newf;
501 			if (changed)
502 				dev_forward_change(idev);
503 		}
504 	}
505 	rcu_read_unlock();
506 }
507 
508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
509 {
510 	struct net *net;
511 
512 	net = (struct net *)table->extra2;
513 	if (p == &net->ipv6.devconf_dflt->forwarding)
514 		return 0;
515 
516 	if (!rtnl_trylock()) {
517 		/* Restore the original values before restarting */
518 		*p = old;
519 		return restart_syscall();
520 	}
521 
522 	if (p == &net->ipv6.devconf_all->forwarding) {
523 		__s32 newf = net->ipv6.devconf_all->forwarding;
524 		net->ipv6.devconf_dflt->forwarding = newf;
525 		addrconf_forward_change(net, newf);
526 	} else if ((!*p) ^ (!old))
527 		dev_forward_change((struct inet6_dev *)table->extra1);
528 	rtnl_unlock();
529 
530 	if (*p)
531 		rt6_purge_dflt_routers(net);
532 	return 1;
533 }
534 #endif
535 
536 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
537 {
538 	struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
539 	kfree(ifp);
540 }
541 
542 /* Nobody refers to this ifaddr, destroy it */
543 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
544 {
545 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
546 
547 #ifdef NET_REFCNT_DEBUG
548 	printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
549 #endif
550 
551 	in6_dev_put(ifp->idev);
552 
553 	if (del_timer(&ifp->timer))
554 		pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
555 
556 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
557 		pr_warning("Freeing alive inet6 address %p\n", ifp);
558 		return;
559 	}
560 	dst_release(&ifp->rt->u.dst);
561 
562 	call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
563 }
564 
565 static void
566 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
567 {
568 	struct list_head *p;
569 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
570 
571 	/*
572 	 * Each device address list is sorted in order of scope -
573 	 * global before linklocal.
574 	 */
575 	list_for_each(p, &idev->addr_list) {
576 		struct inet6_ifaddr *ifa
577 			= list_entry(p, struct inet6_ifaddr, if_list);
578 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
579 			break;
580 	}
581 
582 	list_add_tail(&ifp->if_list, p);
583 }
584 
585 static u32 ipv6_addr_hash(const struct in6_addr *addr)
586 {
587 	/*
588 	 * We perform the hash function over the last 64 bits of the address
589 	 * This will include the IEEE address token on links that support it.
590 	 */
591 	return jhash_2words((__force u32)addr->s6_addr32[2],
592 			    (__force u32)addr->s6_addr32[3], 0)
593 		& (IN6_ADDR_HSIZE - 1);
594 }
595 
596 /* On success it returns ifp with increased reference count */
597 
598 static struct inet6_ifaddr *
599 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
600 	      int scope, u32 flags)
601 {
602 	struct inet6_ifaddr *ifa = NULL;
603 	struct rt6_info *rt;
604 	unsigned int hash;
605 	int err = 0;
606 	int addr_type = ipv6_addr_type(addr);
607 
608 	if (addr_type == IPV6_ADDR_ANY ||
609 	    addr_type & IPV6_ADDR_MULTICAST ||
610 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
611 	     addr_type & IPV6_ADDR_LOOPBACK))
612 		return ERR_PTR(-EADDRNOTAVAIL);
613 
614 	rcu_read_lock_bh();
615 	if (idev->dead) {
616 		err = -ENODEV;			/*XXX*/
617 		goto out2;
618 	}
619 
620 	if (idev->cnf.disable_ipv6) {
621 		err = -EACCES;
622 		goto out2;
623 	}
624 
625 	spin_lock(&addrconf_hash_lock);
626 
627 	/* Ignore adding duplicate addresses on an interface */
628 	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
629 		ADBG(("ipv6_add_addr: already assigned\n"));
630 		err = -EEXIST;
631 		goto out;
632 	}
633 
634 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
635 
636 	if (ifa == NULL) {
637 		ADBG(("ipv6_add_addr: malloc failed\n"));
638 		err = -ENOBUFS;
639 		goto out;
640 	}
641 
642 	rt = addrconf_dst_alloc(idev, addr, 0);
643 	if (IS_ERR(rt)) {
644 		err = PTR_ERR(rt);
645 		goto out;
646 	}
647 
648 	ipv6_addr_copy(&ifa->addr, addr);
649 
650 	spin_lock_init(&ifa->lock);
651 	spin_lock_init(&ifa->state_lock);
652 	init_timer(&ifa->timer);
653 	INIT_HLIST_NODE(&ifa->addr_lst);
654 	ifa->timer.data = (unsigned long) ifa;
655 	ifa->scope = scope;
656 	ifa->prefix_len = pfxlen;
657 	ifa->flags = flags | IFA_F_TENTATIVE;
658 	ifa->cstamp = ifa->tstamp = jiffies;
659 
660 	ifa->rt = rt;
661 
662 	/*
663 	 * part one of RFC 4429, section 3.3
664 	 * We should not configure an address as
665 	 * optimistic if we do not yet know the link
666 	 * layer address of our nexhop router
667 	 */
668 
669 	if (rt->rt6i_nexthop == NULL)
670 		ifa->flags &= ~IFA_F_OPTIMISTIC;
671 
672 	ifa->idev = idev;
673 	in6_dev_hold(idev);
674 	/* For caller */
675 	in6_ifa_hold(ifa);
676 
677 	/* Add to big hash table */
678 	hash = ipv6_addr_hash(addr);
679 
680 	hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
681 	spin_unlock(&addrconf_hash_lock);
682 
683 	write_lock(&idev->lock);
684 	/* Add to inet6_dev unicast addr list. */
685 	ipv6_link_dev_addr(idev, ifa);
686 
687 #ifdef CONFIG_IPV6_PRIVACY
688 	if (ifa->flags&IFA_F_TEMPORARY) {
689 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
690 		in6_ifa_hold(ifa);
691 	}
692 #endif
693 
694 	in6_ifa_hold(ifa);
695 	write_unlock(&idev->lock);
696 out2:
697 	rcu_read_unlock_bh();
698 
699 	if (likely(err == 0))
700 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
701 	else {
702 		kfree(ifa);
703 		ifa = ERR_PTR(err);
704 	}
705 
706 	return ifa;
707 out:
708 	spin_unlock(&addrconf_hash_lock);
709 	goto out2;
710 }
711 
712 /* This function wants to get referenced ifp and releases it before return */
713 
714 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
715 {
716 	struct inet6_ifaddr *ifa, *ifn;
717 	struct inet6_dev *idev = ifp->idev;
718 	int state;
719 	int hash;
720 	int deleted = 0, onlink = 0;
721 	unsigned long expires = jiffies;
722 
723 	hash = ipv6_addr_hash(&ifp->addr);
724 
725 	spin_lock_bh(&ifp->state_lock);
726 	state = ifp->state;
727 	ifp->state = INET6_IFADDR_STATE_DEAD;
728 	spin_unlock_bh(&ifp->state_lock);
729 
730 	if (state == INET6_IFADDR_STATE_DEAD)
731 		goto out;
732 
733 	spin_lock_bh(&addrconf_hash_lock);
734 	hlist_del_init_rcu(&ifp->addr_lst);
735 	spin_unlock_bh(&addrconf_hash_lock);
736 
737 	write_lock_bh(&idev->lock);
738 #ifdef CONFIG_IPV6_PRIVACY
739 	if (ifp->flags&IFA_F_TEMPORARY) {
740 		list_del(&ifp->tmp_list);
741 		if (ifp->ifpub) {
742 			in6_ifa_put(ifp->ifpub);
743 			ifp->ifpub = NULL;
744 		}
745 		__in6_ifa_put(ifp);
746 	}
747 #endif
748 
749 	list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
750 		if (ifa == ifp) {
751 			list_del_init(&ifp->if_list);
752 			__in6_ifa_put(ifp);
753 
754 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
755 				break;
756 			deleted = 1;
757 			continue;
758 		} else if (ifp->flags & IFA_F_PERMANENT) {
759 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
760 					      ifp->prefix_len)) {
761 				if (ifa->flags & IFA_F_PERMANENT) {
762 					onlink = 1;
763 					if (deleted)
764 						break;
765 				} else {
766 					unsigned long lifetime;
767 
768 					if (!onlink)
769 						onlink = -1;
770 
771 					spin_lock(&ifa->lock);
772 
773 					lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
774 					/*
775 					 * Note: Because this address is
776 					 * not permanent, lifetime <
777 					 * LONG_MAX / HZ here.
778 					 */
779 					if (time_before(expires,
780 							ifa->tstamp + lifetime * HZ))
781 						expires = ifa->tstamp + lifetime * HZ;
782 					spin_unlock(&ifa->lock);
783 				}
784 			}
785 		}
786 	}
787 	write_unlock_bh(&idev->lock);
788 
789 	addrconf_del_timer(ifp);
790 
791 	ipv6_ifa_notify(RTM_DELADDR, ifp);
792 
793 	atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
794 
795 	/*
796 	 * Purge or update corresponding prefix
797 	 *
798 	 * 1) we don't purge prefix here if address was not permanent.
799 	 *    prefix is managed by its own lifetime.
800 	 * 2) if there're no addresses, delete prefix.
801 	 * 3) if there're still other permanent address(es),
802 	 *    corresponding prefix is still permanent.
803 	 * 4) otherwise, update prefix lifetime to the
804 	 *    longest valid lifetime among the corresponding
805 	 *    addresses on the device.
806 	 *    Note: subsequent RA will update lifetime.
807 	 *
808 	 * --yoshfuji
809 	 */
810 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
811 		struct in6_addr prefix;
812 		struct rt6_info *rt;
813 		struct net *net = dev_net(ifp->idev->dev);
814 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
815 		rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
816 
817 		if (rt && addrconf_is_prefix_route(rt)) {
818 			if (onlink == 0) {
819 				ip6_del_rt(rt);
820 				rt = NULL;
821 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
822 				rt->rt6i_expires = expires;
823 				rt->rt6i_flags |= RTF_EXPIRES;
824 			}
825 		}
826 		dst_release(&rt->u.dst);
827 	}
828 
829 out:
830 	in6_ifa_put(ifp);
831 }
832 
833 #ifdef CONFIG_IPV6_PRIVACY
834 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
835 {
836 	struct inet6_dev *idev = ifp->idev;
837 	struct in6_addr addr, *tmpaddr;
838 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
839 	unsigned long regen_advance;
840 	int tmp_plen;
841 	int ret = 0;
842 	int max_addresses;
843 	u32 addr_flags;
844 
845 	write_lock(&idev->lock);
846 	if (ift) {
847 		spin_lock_bh(&ift->lock);
848 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
849 		spin_unlock_bh(&ift->lock);
850 		tmpaddr = &addr;
851 	} else {
852 		tmpaddr = NULL;
853 	}
854 retry:
855 	in6_dev_hold(idev);
856 	if (idev->cnf.use_tempaddr <= 0) {
857 		write_unlock(&idev->lock);
858 		printk(KERN_INFO
859 			"ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
860 		in6_dev_put(idev);
861 		ret = -1;
862 		goto out;
863 	}
864 	spin_lock_bh(&ifp->lock);
865 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
866 		idev->cnf.use_tempaddr = -1;	/*XXX*/
867 		spin_unlock_bh(&ifp->lock);
868 		write_unlock(&idev->lock);
869 		printk(KERN_WARNING
870 			"ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
871 		in6_dev_put(idev);
872 		ret = -1;
873 		goto out;
874 	}
875 	in6_ifa_hold(ifp);
876 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
877 	if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
878 		spin_unlock_bh(&ifp->lock);
879 		write_unlock(&idev->lock);
880 		printk(KERN_WARNING
881 			"ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
882 		in6_ifa_put(ifp);
883 		in6_dev_put(idev);
884 		ret = -1;
885 		goto out;
886 	}
887 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
888 	tmp_valid_lft = min_t(__u32,
889 			      ifp->valid_lft,
890 			      idev->cnf.temp_valid_lft);
891 	tmp_prefered_lft = min_t(__u32,
892 				 ifp->prefered_lft,
893 				 idev->cnf.temp_prefered_lft - desync_factor / HZ);
894 	tmp_plen = ifp->prefix_len;
895 	max_addresses = idev->cnf.max_addresses;
896 	tmp_cstamp = ifp->cstamp;
897 	tmp_tstamp = ifp->tstamp;
898 	spin_unlock_bh(&ifp->lock);
899 
900 	regen_advance = idev->cnf.regen_max_retry *
901 	                idev->cnf.dad_transmits *
902 	                idev->nd_parms->retrans_time / HZ;
903 	write_unlock(&idev->lock);
904 
905 	/* A temporary address is created only if this calculated Preferred
906 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
907 	 * an implementation must not create a temporary address with a zero
908 	 * Preferred Lifetime.
909 	 */
910 	if (tmp_prefered_lft <= regen_advance) {
911 		in6_ifa_put(ifp);
912 		in6_dev_put(idev);
913 		ret = -1;
914 		goto out;
915 	}
916 
917 	addr_flags = IFA_F_TEMPORARY;
918 	/* set in addrconf_prefix_rcv() */
919 	if (ifp->flags & IFA_F_OPTIMISTIC)
920 		addr_flags |= IFA_F_OPTIMISTIC;
921 
922 	ift = !max_addresses ||
923 	      ipv6_count_addresses(idev) < max_addresses ?
924 		ipv6_add_addr(idev, &addr, tmp_plen,
925 			      ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
926 			      addr_flags) : NULL;
927 	if (!ift || IS_ERR(ift)) {
928 		in6_ifa_put(ifp);
929 		in6_dev_put(idev);
930 		printk(KERN_INFO
931 			"ipv6_create_tempaddr(): retry temporary address regeneration.\n");
932 		tmpaddr = &addr;
933 		write_lock(&idev->lock);
934 		goto retry;
935 	}
936 
937 	spin_lock_bh(&ift->lock);
938 	ift->ifpub = ifp;
939 	ift->valid_lft = tmp_valid_lft;
940 	ift->prefered_lft = tmp_prefered_lft;
941 	ift->cstamp = tmp_cstamp;
942 	ift->tstamp = tmp_tstamp;
943 	spin_unlock_bh(&ift->lock);
944 
945 	addrconf_dad_start(ift, 0);
946 	in6_ifa_put(ift);
947 	in6_dev_put(idev);
948 out:
949 	return ret;
950 }
951 #endif
952 
953 /*
954  *	Choose an appropriate source address (RFC3484)
955  */
956 enum {
957 	IPV6_SADDR_RULE_INIT = 0,
958 	IPV6_SADDR_RULE_LOCAL,
959 	IPV6_SADDR_RULE_SCOPE,
960 	IPV6_SADDR_RULE_PREFERRED,
961 #ifdef CONFIG_IPV6_MIP6
962 	IPV6_SADDR_RULE_HOA,
963 #endif
964 	IPV6_SADDR_RULE_OIF,
965 	IPV6_SADDR_RULE_LABEL,
966 #ifdef CONFIG_IPV6_PRIVACY
967 	IPV6_SADDR_RULE_PRIVACY,
968 #endif
969 	IPV6_SADDR_RULE_ORCHID,
970 	IPV6_SADDR_RULE_PREFIX,
971 	IPV6_SADDR_RULE_MAX
972 };
973 
974 struct ipv6_saddr_score {
975 	int			rule;
976 	int			addr_type;
977 	struct inet6_ifaddr	*ifa;
978 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
979 	int			scopedist;
980 	int			matchlen;
981 };
982 
983 struct ipv6_saddr_dst {
984 	const struct in6_addr *addr;
985 	int ifindex;
986 	int scope;
987 	int label;
988 	unsigned int prefs;
989 };
990 
991 static inline int ipv6_saddr_preferred(int type)
992 {
993 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
994 		return 1;
995 	return 0;
996 }
997 
998 static int ipv6_get_saddr_eval(struct net *net,
999 			       struct ipv6_saddr_score *score,
1000 			       struct ipv6_saddr_dst *dst,
1001 			       int i)
1002 {
1003 	int ret;
1004 
1005 	if (i <= score->rule) {
1006 		switch (i) {
1007 		case IPV6_SADDR_RULE_SCOPE:
1008 			ret = score->scopedist;
1009 			break;
1010 		case IPV6_SADDR_RULE_PREFIX:
1011 			ret = score->matchlen;
1012 			break;
1013 		default:
1014 			ret = !!test_bit(i, score->scorebits);
1015 		}
1016 		goto out;
1017 	}
1018 
1019 	switch (i) {
1020 	case IPV6_SADDR_RULE_INIT:
1021 		/* Rule 0: remember if hiscore is not ready yet */
1022 		ret = !!score->ifa;
1023 		break;
1024 	case IPV6_SADDR_RULE_LOCAL:
1025 		/* Rule 1: Prefer same address */
1026 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1027 		break;
1028 	case IPV6_SADDR_RULE_SCOPE:
1029 		/* Rule 2: Prefer appropriate scope
1030 		 *
1031 		 *      ret
1032 		 *       ^
1033 		 *    -1 |  d 15
1034 		 *    ---+--+-+---> scope
1035 		 *       |
1036 		 *       |             d is scope of the destination.
1037 		 *  B-d  |  \
1038 		 *       |   \      <- smaller scope is better if
1039 		 *  B-15 |    \        if scope is enough for destinaion.
1040 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1041 		 * d-C-1 | /
1042 		 *       |/         <- greater is better
1043 		 *   -C  /             if scope is not enough for destination.
1044 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1045 		 *
1046 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1047 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1048 		 * Assume B = 0 and we get C > 29.
1049 		 */
1050 		ret = __ipv6_addr_src_scope(score->addr_type);
1051 		if (ret >= dst->scope)
1052 			ret = -ret;
1053 		else
1054 			ret -= 128;	/* 30 is enough */
1055 		score->scopedist = ret;
1056 		break;
1057 	case IPV6_SADDR_RULE_PREFERRED:
1058 		/* Rule 3: Avoid deprecated and optimistic addresses */
1059 		ret = ipv6_saddr_preferred(score->addr_type) ||
1060 		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1061 		break;
1062 #ifdef CONFIG_IPV6_MIP6
1063 	case IPV6_SADDR_RULE_HOA:
1064 	    {
1065 		/* Rule 4: Prefer home address */
1066 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1067 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1068 		break;
1069 	    }
1070 #endif
1071 	case IPV6_SADDR_RULE_OIF:
1072 		/* Rule 5: Prefer outgoing interface */
1073 		ret = (!dst->ifindex ||
1074 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1075 		break;
1076 	case IPV6_SADDR_RULE_LABEL:
1077 		/* Rule 6: Prefer matching label */
1078 		ret = ipv6_addr_label(net,
1079 				      &score->ifa->addr, score->addr_type,
1080 				      score->ifa->idev->dev->ifindex) == dst->label;
1081 		break;
1082 #ifdef CONFIG_IPV6_PRIVACY
1083 	case IPV6_SADDR_RULE_PRIVACY:
1084 	    {
1085 		/* Rule 7: Prefer public address
1086 		 * Note: prefer temprary address if use_tempaddr >= 2
1087 		 */
1088 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1089 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1090 				score->ifa->idev->cnf.use_tempaddr >= 2;
1091 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1092 		break;
1093 	    }
1094 #endif
1095 	case IPV6_SADDR_RULE_ORCHID:
1096 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1097 		 *	    non-ORCHID vs non-ORCHID
1098 		 */
1099 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1100 			ipv6_addr_orchid(dst->addr));
1101 		break;
1102 	case IPV6_SADDR_RULE_PREFIX:
1103 		/* Rule 8: Use longest matching prefix */
1104 		score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1105 						       dst->addr);
1106 		break;
1107 	default:
1108 		ret = 0;
1109 	}
1110 
1111 	if (ret)
1112 		__set_bit(i, score->scorebits);
1113 	score->rule = i;
1114 out:
1115 	return ret;
1116 }
1117 
1118 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1119 		       const struct in6_addr *daddr, unsigned int prefs,
1120 		       struct in6_addr *saddr)
1121 {
1122 	struct ipv6_saddr_score scores[2],
1123 				*score = &scores[0], *hiscore = &scores[1];
1124 	struct ipv6_saddr_dst dst;
1125 	struct net_device *dev;
1126 	int dst_type;
1127 
1128 	dst_type = __ipv6_addr_type(daddr);
1129 	dst.addr = daddr;
1130 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1131 	dst.scope = __ipv6_addr_src_scope(dst_type);
1132 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1133 	dst.prefs = prefs;
1134 
1135 	hiscore->rule = -1;
1136 	hiscore->ifa = NULL;
1137 
1138 	rcu_read_lock();
1139 
1140 	for_each_netdev_rcu(net, dev) {
1141 		struct inet6_dev *idev;
1142 
1143 		/* Candidate Source Address (section 4)
1144 		 *  - multicast and link-local destination address,
1145 		 *    the set of candidate source address MUST only
1146 		 *    include addresses assigned to interfaces
1147 		 *    belonging to the same link as the outgoing
1148 		 *    interface.
1149 		 * (- For site-local destination addresses, the
1150 		 *    set of candidate source addresses MUST only
1151 		 *    include addresses assigned to interfaces
1152 		 *    belonging to the same site as the outgoing
1153 		 *    interface.)
1154 		 */
1155 		if (((dst_type & IPV6_ADDR_MULTICAST) ||
1156 		     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1157 		    dst.ifindex && dev->ifindex != dst.ifindex)
1158 			continue;
1159 
1160 		idev = __in6_dev_get(dev);
1161 		if (!idev)
1162 			continue;
1163 
1164 		read_lock_bh(&idev->lock);
1165 		list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1166 			int i;
1167 
1168 			/*
1169 			 * - Tentative Address (RFC2462 section 5.4)
1170 			 *  - A tentative address is not considered
1171 			 *    "assigned to an interface" in the traditional
1172 			 *    sense, unless it is also flagged as optimistic.
1173 			 * - Candidate Source Address (section 4)
1174 			 *  - In any case, anycast addresses, multicast
1175 			 *    addresses, and the unspecified address MUST
1176 			 *    NOT be included in a candidate set.
1177 			 */
1178 			if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1179 			    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1180 				continue;
1181 
1182 			score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1183 
1184 			if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1185 				     score->addr_type & IPV6_ADDR_MULTICAST)) {
1186 				LIMIT_NETDEBUG(KERN_DEBUG
1187 					       "ADDRCONF: unspecified / multicast address "
1188 					       "assigned as unicast address on %s",
1189 					       dev->name);
1190 				continue;
1191 			}
1192 
1193 			score->rule = -1;
1194 			bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1195 
1196 			for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1197 				int minihiscore, miniscore;
1198 
1199 				minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1200 				miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1201 
1202 				if (minihiscore > miniscore) {
1203 					if (i == IPV6_SADDR_RULE_SCOPE &&
1204 					    score->scopedist > 0) {
1205 						/*
1206 						 * special case:
1207 						 * each remaining entry
1208 						 * has too small (not enough)
1209 						 * scope, because ifa entries
1210 						 * are sorted by their scope
1211 						 * values.
1212 						 */
1213 						goto try_nextdev;
1214 					}
1215 					break;
1216 				} else if (minihiscore < miniscore) {
1217 					if (hiscore->ifa)
1218 						in6_ifa_put(hiscore->ifa);
1219 
1220 					in6_ifa_hold(score->ifa);
1221 
1222 					swap(hiscore, score);
1223 
1224 					/* restore our iterator */
1225 					score->ifa = hiscore->ifa;
1226 
1227 					break;
1228 				}
1229 			}
1230 		}
1231 try_nextdev:
1232 		read_unlock_bh(&idev->lock);
1233 	}
1234 	rcu_read_unlock();
1235 
1236 	if (!hiscore->ifa)
1237 		return -EADDRNOTAVAIL;
1238 
1239 	ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1240 	in6_ifa_put(hiscore->ifa);
1241 	return 0;
1242 }
1243 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1244 
1245 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1246 		    unsigned char banned_flags)
1247 {
1248 	struct inet6_dev *idev;
1249 	int err = -EADDRNOTAVAIL;
1250 
1251 	rcu_read_lock();
1252 	idev = __in6_dev_get(dev);
1253 	if (idev) {
1254 		struct inet6_ifaddr *ifp;
1255 
1256 		read_lock_bh(&idev->lock);
1257 		list_for_each_entry(ifp, &idev->addr_list, if_list) {
1258 			if (ifp->scope == IFA_LINK &&
1259 			    !(ifp->flags & banned_flags)) {
1260 				ipv6_addr_copy(addr, &ifp->addr);
1261 				err = 0;
1262 				break;
1263 			}
1264 		}
1265 		read_unlock_bh(&idev->lock);
1266 	}
1267 	rcu_read_unlock();
1268 	return err;
1269 }
1270 
1271 static int ipv6_count_addresses(struct inet6_dev *idev)
1272 {
1273 	int cnt = 0;
1274 	struct inet6_ifaddr *ifp;
1275 
1276 	read_lock_bh(&idev->lock);
1277 	list_for_each_entry(ifp, &idev->addr_list, if_list)
1278 		cnt++;
1279 	read_unlock_bh(&idev->lock);
1280 	return cnt;
1281 }
1282 
1283 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1284 		  struct net_device *dev, int strict)
1285 {
1286 	struct inet6_ifaddr *ifp;
1287 	struct hlist_node *node;
1288 	unsigned int hash = ipv6_addr_hash(addr);
1289 
1290 	rcu_read_lock_bh();
1291 	hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1292 		if (!net_eq(dev_net(ifp->idev->dev), net))
1293 			continue;
1294 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1295 		    !(ifp->flags&IFA_F_TENTATIVE) &&
1296 		    (dev == NULL || ifp->idev->dev == dev ||
1297 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1298 			rcu_read_unlock_bh();
1299 			return 1;
1300 		}
1301 	}
1302 
1303 	rcu_read_unlock_bh();
1304 	return 0;
1305 }
1306 EXPORT_SYMBOL(ipv6_chk_addr);
1307 
1308 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309 			       struct net_device *dev)
1310 {
1311 	unsigned int hash = ipv6_addr_hash(addr);
1312 	struct inet6_ifaddr *ifp;
1313 	struct hlist_node *node;
1314 
1315 	hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1316 		if (!net_eq(dev_net(ifp->idev->dev), net))
1317 			continue;
1318 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1319 			if (dev == NULL || ifp->idev->dev == dev)
1320 				return true;
1321 		}
1322 	}
1323 	return false;
1324 }
1325 
1326 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1327 {
1328 	struct inet6_dev *idev;
1329 	struct inet6_ifaddr *ifa;
1330 	int	onlink;
1331 
1332 	onlink = 0;
1333 	rcu_read_lock();
1334 	idev = __in6_dev_get(dev);
1335 	if (idev) {
1336 		read_lock_bh(&idev->lock);
1337 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1338 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1339 						   ifa->prefix_len);
1340 			if (onlink)
1341 				break;
1342 		}
1343 		read_unlock_bh(&idev->lock);
1344 	}
1345 	rcu_read_unlock();
1346 	return onlink;
1347 }
1348 
1349 EXPORT_SYMBOL(ipv6_chk_prefix);
1350 
1351 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1352 				     struct net_device *dev, int strict)
1353 {
1354 	struct inet6_ifaddr *ifp, *result = NULL;
1355 	unsigned int hash = ipv6_addr_hash(addr);
1356 	struct hlist_node *node;
1357 
1358 	rcu_read_lock_bh();
1359 	hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1360 		if (!net_eq(dev_net(ifp->idev->dev), net))
1361 			continue;
1362 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1363 			if (dev == NULL || ifp->idev->dev == dev ||
1364 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1365 				result = ifp;
1366 				in6_ifa_hold(ifp);
1367 				break;
1368 			}
1369 		}
1370 	}
1371 	rcu_read_unlock_bh();
1372 
1373 	return result;
1374 }
1375 
1376 /* Gets referenced address, destroys ifaddr */
1377 
1378 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1379 {
1380 	if (ifp->flags&IFA_F_PERMANENT) {
1381 		spin_lock_bh(&ifp->lock);
1382 		addrconf_del_timer(ifp);
1383 		ifp->flags |= IFA_F_TENTATIVE;
1384 		if (dad_failed)
1385 			ifp->flags |= IFA_F_DADFAILED;
1386 		spin_unlock_bh(&ifp->lock);
1387 		if (dad_failed)
1388 			ipv6_ifa_notify(0, ifp);
1389 		in6_ifa_put(ifp);
1390 #ifdef CONFIG_IPV6_PRIVACY
1391 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1392 		struct inet6_ifaddr *ifpub;
1393 		spin_lock_bh(&ifp->lock);
1394 		ifpub = ifp->ifpub;
1395 		if (ifpub) {
1396 			in6_ifa_hold(ifpub);
1397 			spin_unlock_bh(&ifp->lock);
1398 			ipv6_create_tempaddr(ifpub, ifp);
1399 			in6_ifa_put(ifpub);
1400 		} else {
1401 			spin_unlock_bh(&ifp->lock);
1402 		}
1403 		ipv6_del_addr(ifp);
1404 #endif
1405 	} else
1406 		ipv6_del_addr(ifp);
1407 }
1408 
1409 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1410 {
1411 	int err = -ENOENT;
1412 
1413 	spin_lock(&ifp->state_lock);
1414 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
1415 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
1416 		err = 0;
1417 	}
1418 	spin_unlock(&ifp->state_lock);
1419 
1420 	return err;
1421 }
1422 
1423 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1424 {
1425 	struct inet6_dev *idev = ifp->idev;
1426 
1427 	if (addrconf_dad_end(ifp))
1428 		return;
1429 
1430 	if (net_ratelimit())
1431 		printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1432 			ifp->idev->dev->name, &ifp->addr);
1433 
1434 	if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1435 		struct in6_addr addr;
1436 
1437 		addr.s6_addr32[0] = htonl(0xfe800000);
1438 		addr.s6_addr32[1] = 0;
1439 
1440 		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1441 		    ipv6_addr_equal(&ifp->addr, &addr)) {
1442 			/* DAD failed for link-local based on MAC address */
1443 			idev->cnf.disable_ipv6 = 1;
1444 
1445 			printk(KERN_INFO "%s: IPv6 being disabled!\n",
1446 				ifp->idev->dev->name);
1447 		}
1448 	}
1449 
1450 	addrconf_dad_stop(ifp, 1);
1451 }
1452 
1453 /* Join to solicited addr multicast group. */
1454 
1455 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1456 {
1457 	struct in6_addr maddr;
1458 
1459 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1460 		return;
1461 
1462 	addrconf_addr_solict_mult(addr, &maddr);
1463 	ipv6_dev_mc_inc(dev, &maddr);
1464 }
1465 
1466 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1467 {
1468 	struct in6_addr maddr;
1469 
1470 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1471 		return;
1472 
1473 	addrconf_addr_solict_mult(addr, &maddr);
1474 	__ipv6_dev_mc_dec(idev, &maddr);
1475 }
1476 
1477 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1478 {
1479 	struct in6_addr addr;
1480 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1481 	if (ipv6_addr_any(&addr))
1482 		return;
1483 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1484 }
1485 
1486 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1487 {
1488 	struct in6_addr addr;
1489 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490 	if (ipv6_addr_any(&addr))
1491 		return;
1492 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1493 }
1494 
1495 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1496 {
1497 	if (dev->addr_len != ETH_ALEN)
1498 		return -1;
1499 	memcpy(eui, dev->dev_addr, 3);
1500 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1501 
1502 	/*
1503 	 * The zSeries OSA network cards can be shared among various
1504 	 * OS instances, but the OSA cards have only one MAC address.
1505 	 * This leads to duplicate address conflicts in conjunction
1506 	 * with IPv6 if more than one instance uses the same card.
1507 	 *
1508 	 * The driver for these cards can deliver a unique 16-bit
1509 	 * identifier for each instance sharing the same card.  It is
1510 	 * placed instead of 0xFFFE in the interface identifier.  The
1511 	 * "u" bit of the interface identifier is not inverted in this
1512 	 * case.  Hence the resulting interface identifier has local
1513 	 * scope according to RFC2373.
1514 	 */
1515 	if (dev->dev_id) {
1516 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1517 		eui[4] = dev->dev_id & 0xFF;
1518 	} else {
1519 		eui[3] = 0xFF;
1520 		eui[4] = 0xFE;
1521 		eui[0] ^= 2;
1522 	}
1523 	return 0;
1524 }
1525 
1526 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1527 {
1528 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1529 	if (dev->addr_len != ARCNET_ALEN)
1530 		return -1;
1531 	memset(eui, 0, 7);
1532 	eui[7] = *(u8*)dev->dev_addr;
1533 	return 0;
1534 }
1535 
1536 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1537 {
1538 	if (dev->addr_len != INFINIBAND_ALEN)
1539 		return -1;
1540 	memcpy(eui, dev->dev_addr + 12, 8);
1541 	eui[0] |= 2;
1542 	return 0;
1543 }
1544 
1545 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1546 {
1547 	if (addr == 0)
1548 		return -1;
1549 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1550 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1551 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1552 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1553 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1554 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1555 	eui[1] = 0;
1556 	eui[2] = 0x5E;
1557 	eui[3] = 0xFE;
1558 	memcpy(eui + 4, &addr, 4);
1559 	return 0;
1560 }
1561 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1562 
1563 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1564 {
1565 	if (dev->priv_flags & IFF_ISATAP)
1566 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1567 	return -1;
1568 }
1569 
1570 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1571 {
1572 	switch (dev->type) {
1573 	case ARPHRD_ETHER:
1574 	case ARPHRD_FDDI:
1575 	case ARPHRD_IEEE802_TR:
1576 		return addrconf_ifid_eui48(eui, dev);
1577 	case ARPHRD_ARCNET:
1578 		return addrconf_ifid_arcnet(eui, dev);
1579 	case ARPHRD_INFINIBAND:
1580 		return addrconf_ifid_infiniband(eui, dev);
1581 	case ARPHRD_SIT:
1582 		return addrconf_ifid_sit(eui, dev);
1583 	}
1584 	return -1;
1585 }
1586 
1587 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1588 {
1589 	int err = -1;
1590 	struct inet6_ifaddr *ifp;
1591 
1592 	read_lock_bh(&idev->lock);
1593 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
1594 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1595 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1596 			err = 0;
1597 			break;
1598 		}
1599 	}
1600 	read_unlock_bh(&idev->lock);
1601 	return err;
1602 }
1603 
1604 #ifdef CONFIG_IPV6_PRIVACY
1605 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1606 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1607 {
1608 regen:
1609 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1610 	idev->rndid[0] &= ~0x02;
1611 
1612 	/*
1613 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1614 	 * check if generated address is not inappropriate
1615 	 *
1616 	 *  - Reserved subnet anycast (RFC 2526)
1617 	 *	11111101 11....11 1xxxxxxx
1618 	 *  - ISATAP (RFC4214) 6.1
1619 	 *	00-00-5E-FE-xx-xx-xx-xx
1620 	 *  - value 0
1621 	 *  - XXX: already assigned to an address on the device
1622 	 */
1623 	if (idev->rndid[0] == 0xfd &&
1624 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1625 	    (idev->rndid[7]&0x80))
1626 		goto regen;
1627 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1628 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1629 			goto regen;
1630 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1631 			goto regen;
1632 	}
1633 
1634 	return 0;
1635 }
1636 
1637 static void ipv6_regen_rndid(unsigned long data)
1638 {
1639 	struct inet6_dev *idev = (struct inet6_dev *) data;
1640 	unsigned long expires;
1641 
1642 	rcu_read_lock_bh();
1643 	write_lock_bh(&idev->lock);
1644 
1645 	if (idev->dead)
1646 		goto out;
1647 
1648 	if (__ipv6_regen_rndid(idev) < 0)
1649 		goto out;
1650 
1651 	expires = jiffies +
1652 		idev->cnf.temp_prefered_lft * HZ -
1653 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1654 	if (time_before(expires, jiffies)) {
1655 		printk(KERN_WARNING
1656 			"ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1657 			idev->dev->name);
1658 		goto out;
1659 	}
1660 
1661 	if (!mod_timer(&idev->regen_timer, expires))
1662 		in6_dev_hold(idev);
1663 
1664 out:
1665 	write_unlock_bh(&idev->lock);
1666 	rcu_read_unlock_bh();
1667 	in6_dev_put(idev);
1668 }
1669 
1670 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1671 	int ret = 0;
1672 
1673 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1674 		ret = __ipv6_regen_rndid(idev);
1675 	return ret;
1676 }
1677 #endif
1678 
1679 /*
1680  *	Add prefix route.
1681  */
1682 
1683 static void
1684 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1685 		      unsigned long expires, u32 flags)
1686 {
1687 	struct fib6_config cfg = {
1688 		.fc_table = RT6_TABLE_PREFIX,
1689 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1690 		.fc_ifindex = dev->ifindex,
1691 		.fc_expires = expires,
1692 		.fc_dst_len = plen,
1693 		.fc_flags = RTF_UP | flags,
1694 		.fc_nlinfo.nl_net = dev_net(dev),
1695 		.fc_protocol = RTPROT_KERNEL,
1696 	};
1697 
1698 	ipv6_addr_copy(&cfg.fc_dst, pfx);
1699 
1700 	/* Prevent useless cloning on PtP SIT.
1701 	   This thing is done here expecting that the whole
1702 	   class of non-broadcast devices need not cloning.
1703 	 */
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1706 		cfg.fc_flags |= RTF_NONEXTHOP;
1707 #endif
1708 
1709 	ip6_route_add(&cfg);
1710 }
1711 
1712 /* Create "default" multicast route to the interface */
1713 
1714 static void addrconf_add_mroute(struct net_device *dev)
1715 {
1716 	struct fib6_config cfg = {
1717 		.fc_table = RT6_TABLE_LOCAL,
1718 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1719 		.fc_ifindex = dev->ifindex,
1720 		.fc_dst_len = 8,
1721 		.fc_flags = RTF_UP,
1722 		.fc_nlinfo.nl_net = dev_net(dev),
1723 	};
1724 
1725 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1726 
1727 	ip6_route_add(&cfg);
1728 }
1729 
1730 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1731 static void sit_route_add(struct net_device *dev)
1732 {
1733 	struct fib6_config cfg = {
1734 		.fc_table = RT6_TABLE_MAIN,
1735 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1736 		.fc_ifindex = dev->ifindex,
1737 		.fc_dst_len = 96,
1738 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
1739 		.fc_nlinfo.nl_net = dev_net(dev),
1740 	};
1741 
1742 	/* prefix length - 96 bits "::d.d.d.d" */
1743 	ip6_route_add(&cfg);
1744 }
1745 #endif
1746 
1747 static void addrconf_add_lroute(struct net_device *dev)
1748 {
1749 	struct in6_addr addr;
1750 
1751 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1752 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
1753 }
1754 
1755 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1756 {
1757 	struct inet6_dev *idev;
1758 
1759 	ASSERT_RTNL();
1760 
1761 	idev = ipv6_find_idev(dev);
1762 	if (!idev)
1763 		return NULL;
1764 
1765 	/* Add default multicast route */
1766 	addrconf_add_mroute(dev);
1767 
1768 	/* Add link local route */
1769 	addrconf_add_lroute(dev);
1770 	return idev;
1771 }
1772 
1773 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1774 {
1775 	struct prefix_info *pinfo;
1776 	__u32 valid_lft;
1777 	__u32 prefered_lft;
1778 	int addr_type;
1779 	struct inet6_dev *in6_dev;
1780 	struct net *net = dev_net(dev);
1781 
1782 	pinfo = (struct prefix_info *) opt;
1783 
1784 	if (len < sizeof(struct prefix_info)) {
1785 		ADBG(("addrconf: prefix option too short\n"));
1786 		return;
1787 	}
1788 
1789 	/*
1790 	 *	Validation checks ([ADDRCONF], page 19)
1791 	 */
1792 
1793 	addr_type = ipv6_addr_type(&pinfo->prefix);
1794 
1795 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1796 		return;
1797 
1798 	valid_lft = ntohl(pinfo->valid);
1799 	prefered_lft = ntohl(pinfo->prefered);
1800 
1801 	if (prefered_lft > valid_lft) {
1802 		if (net_ratelimit())
1803 			printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1804 		return;
1805 	}
1806 
1807 	in6_dev = in6_dev_get(dev);
1808 
1809 	if (in6_dev == NULL) {
1810 		if (net_ratelimit())
1811 			printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1812 		return;
1813 	}
1814 
1815 	/*
1816 	 *	Two things going on here:
1817 	 *	1) Add routes for on-link prefixes
1818 	 *	2) Configure prefixes with the auto flag set
1819 	 */
1820 
1821 	if (pinfo->onlink) {
1822 		struct rt6_info *rt;
1823 		unsigned long rt_expires;
1824 
1825 		/* Avoid arithmetic overflow. Really, we could
1826 		 * save rt_expires in seconds, likely valid_lft,
1827 		 * but it would require division in fib gc, that it
1828 		 * not good.
1829 		 */
1830 		if (HZ > USER_HZ)
1831 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1832 		else
1833 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1834 
1835 		if (addrconf_finite_timeout(rt_expires))
1836 			rt_expires *= HZ;
1837 
1838 		rt = rt6_lookup(net, &pinfo->prefix, NULL,
1839 				dev->ifindex, 1);
1840 
1841 		if (rt && addrconf_is_prefix_route(rt)) {
1842 			/* Autoconf prefix route */
1843 			if (valid_lft == 0) {
1844 				ip6_del_rt(rt);
1845 				rt = NULL;
1846 			} else if (addrconf_finite_timeout(rt_expires)) {
1847 				/* not infinity */
1848 				rt->rt6i_expires = jiffies + rt_expires;
1849 				rt->rt6i_flags |= RTF_EXPIRES;
1850 			} else {
1851 				rt->rt6i_flags &= ~RTF_EXPIRES;
1852 				rt->rt6i_expires = 0;
1853 			}
1854 		} else if (valid_lft) {
1855 			clock_t expires = 0;
1856 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1857 			if (addrconf_finite_timeout(rt_expires)) {
1858 				/* not infinity */
1859 				flags |= RTF_EXPIRES;
1860 				expires = jiffies_to_clock_t(rt_expires);
1861 			}
1862 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1863 					      dev, expires, flags);
1864 		}
1865 		if (rt)
1866 			dst_release(&rt->u.dst);
1867 	}
1868 
1869 	/* Try to figure out our local address for this prefix */
1870 
1871 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1872 		struct inet6_ifaddr * ifp;
1873 		struct in6_addr addr;
1874 		int create = 0, update_lft = 0;
1875 
1876 		if (pinfo->prefix_len == 64) {
1877 			memcpy(&addr, &pinfo->prefix, 8);
1878 			if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1879 			    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1880 				in6_dev_put(in6_dev);
1881 				return;
1882 			}
1883 			goto ok;
1884 		}
1885 		if (net_ratelimit())
1886 			printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1887 			       pinfo->prefix_len);
1888 		in6_dev_put(in6_dev);
1889 		return;
1890 
1891 ok:
1892 
1893 		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1894 
1895 		if (ifp == NULL && valid_lft) {
1896 			int max_addresses = in6_dev->cnf.max_addresses;
1897 			u32 addr_flags = 0;
1898 
1899 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1900 			if (in6_dev->cnf.optimistic_dad &&
1901 			    !net->ipv6.devconf_all->forwarding)
1902 				addr_flags = IFA_F_OPTIMISTIC;
1903 #endif
1904 
1905 			/* Do not allow to create too much of autoconfigured
1906 			 * addresses; this would be too easy way to crash kernel.
1907 			 */
1908 			if (!max_addresses ||
1909 			    ipv6_count_addresses(in6_dev) < max_addresses)
1910 				ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1911 						    addr_type&IPV6_ADDR_SCOPE_MASK,
1912 						    addr_flags);
1913 
1914 			if (!ifp || IS_ERR(ifp)) {
1915 				in6_dev_put(in6_dev);
1916 				return;
1917 			}
1918 
1919 			update_lft = create = 1;
1920 			ifp->cstamp = jiffies;
1921 			addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1922 		}
1923 
1924 		if (ifp) {
1925 			int flags;
1926 			unsigned long now;
1927 #ifdef CONFIG_IPV6_PRIVACY
1928 			struct inet6_ifaddr *ift;
1929 #endif
1930 			u32 stored_lft;
1931 
1932 			/* update lifetime (RFC2462 5.5.3 e) */
1933 			spin_lock(&ifp->lock);
1934 			now = jiffies;
1935 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1936 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1937 			else
1938 				stored_lft = 0;
1939 			if (!update_lft && stored_lft) {
1940 				if (valid_lft > MIN_VALID_LIFETIME ||
1941 				    valid_lft > stored_lft)
1942 					update_lft = 1;
1943 				else if (stored_lft <= MIN_VALID_LIFETIME) {
1944 					/* valid_lft <= stored_lft is always true */
1945 					/*
1946 					 * RFC 4862 Section 5.5.3e:
1947 					 * "Note that the preferred lifetime of
1948 					 *  the corresponding address is always
1949 					 *  reset to the Preferred Lifetime in
1950 					 *  the received Prefix Information
1951 					 *  option, regardless of whether the
1952 					 *  valid lifetime is also reset or
1953 					 *  ignored."
1954 					 *
1955 					 *  So if the preferred lifetime in
1956 					 *  this advertisement is different
1957 					 *  than what we have stored, but the
1958 					 *  valid lifetime is invalid, just
1959 					 *  reset prefered_lft.
1960 					 *
1961 					 *  We must set the valid lifetime
1962 					 *  to the stored lifetime since we'll
1963 					 *  be updating the timestamp below,
1964 					 *  else we'll set it back to the
1965 					 *  minumum.
1966 					 */
1967 					if (prefered_lft != ifp->prefered_lft) {
1968 						valid_lft = stored_lft;
1969 						update_lft = 1;
1970 					}
1971 				} else {
1972 					valid_lft = MIN_VALID_LIFETIME;
1973 					if (valid_lft < prefered_lft)
1974 						prefered_lft = valid_lft;
1975 					update_lft = 1;
1976 				}
1977 			}
1978 
1979 			if (update_lft) {
1980 				ifp->valid_lft = valid_lft;
1981 				ifp->prefered_lft = prefered_lft;
1982 				ifp->tstamp = now;
1983 				flags = ifp->flags;
1984 				ifp->flags &= ~IFA_F_DEPRECATED;
1985 				spin_unlock(&ifp->lock);
1986 
1987 				if (!(flags&IFA_F_TENTATIVE))
1988 					ipv6_ifa_notify(0, ifp);
1989 			} else
1990 				spin_unlock(&ifp->lock);
1991 
1992 #ifdef CONFIG_IPV6_PRIVACY
1993 			read_lock_bh(&in6_dev->lock);
1994 			/* update all temporary addresses in the list */
1995 			list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1996 				/*
1997 				 * When adjusting the lifetimes of an existing
1998 				 * temporary address, only lower the lifetimes.
1999 				 * Implementations must not increase the
2000 				 * lifetimes of an existing temporary address
2001 				 * when processing a Prefix Information Option.
2002 				 */
2003 				if (ifp != ift->ifpub)
2004 					continue;
2005 
2006 				spin_lock(&ift->lock);
2007 				flags = ift->flags;
2008 				if (ift->valid_lft > valid_lft &&
2009 				    ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2010 					ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2011 				if (ift->prefered_lft > prefered_lft &&
2012 				    ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2013 					ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2014 				spin_unlock(&ift->lock);
2015 				if (!(flags&IFA_F_TENTATIVE))
2016 					ipv6_ifa_notify(0, ift);
2017 			}
2018 
2019 			if (create && in6_dev->cnf.use_tempaddr > 0) {
2020 				/*
2021 				 * When a new public address is created as described in [ADDRCONF],
2022 				 * also create a new temporary address.
2023 				 */
2024 				read_unlock_bh(&in6_dev->lock);
2025 				ipv6_create_tempaddr(ifp, NULL);
2026 			} else {
2027 				read_unlock_bh(&in6_dev->lock);
2028 			}
2029 #endif
2030 			in6_ifa_put(ifp);
2031 			addrconf_verify(0);
2032 		}
2033 	}
2034 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2035 	in6_dev_put(in6_dev);
2036 }
2037 
2038 /*
2039  *	Set destination address.
2040  *	Special case for SIT interfaces where we create a new "virtual"
2041  *	device.
2042  */
2043 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2044 {
2045 	struct in6_ifreq ireq;
2046 	struct net_device *dev;
2047 	int err = -EINVAL;
2048 
2049 	rtnl_lock();
2050 
2051 	err = -EFAULT;
2052 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2053 		goto err_exit;
2054 
2055 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2056 
2057 	err = -ENODEV;
2058 	if (dev == NULL)
2059 		goto err_exit;
2060 
2061 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2062 	if (dev->type == ARPHRD_SIT) {
2063 		const struct net_device_ops *ops = dev->netdev_ops;
2064 		struct ifreq ifr;
2065 		struct ip_tunnel_parm p;
2066 
2067 		err = -EADDRNOTAVAIL;
2068 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2069 			goto err_exit;
2070 
2071 		memset(&p, 0, sizeof(p));
2072 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2073 		p.iph.saddr = 0;
2074 		p.iph.version = 4;
2075 		p.iph.ihl = 5;
2076 		p.iph.protocol = IPPROTO_IPV6;
2077 		p.iph.ttl = 64;
2078 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2079 
2080 		if (ops->ndo_do_ioctl) {
2081 			mm_segment_t oldfs = get_fs();
2082 
2083 			set_fs(KERNEL_DS);
2084 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2085 			set_fs(oldfs);
2086 		} else
2087 			err = -EOPNOTSUPP;
2088 
2089 		if (err == 0) {
2090 			err = -ENOBUFS;
2091 			dev = __dev_get_by_name(net, p.name);
2092 			if (!dev)
2093 				goto err_exit;
2094 			err = dev_open(dev);
2095 		}
2096 	}
2097 #endif
2098 
2099 err_exit:
2100 	rtnl_unlock();
2101 	return err;
2102 }
2103 
2104 /*
2105  *	Manual configuration of address on an interface
2106  */
2107 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2108 			  unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2109 			  __u32 valid_lft)
2110 {
2111 	struct inet6_ifaddr *ifp;
2112 	struct inet6_dev *idev;
2113 	struct net_device *dev;
2114 	int scope;
2115 	u32 flags;
2116 	clock_t expires;
2117 	unsigned long timeout;
2118 
2119 	ASSERT_RTNL();
2120 
2121 	if (plen > 128)
2122 		return -EINVAL;
2123 
2124 	/* check the lifetime */
2125 	if (!valid_lft || prefered_lft > valid_lft)
2126 		return -EINVAL;
2127 
2128 	dev = __dev_get_by_index(net, ifindex);
2129 	if (!dev)
2130 		return -ENODEV;
2131 
2132 	if ((idev = addrconf_add_dev(dev)) == NULL)
2133 		return -ENOBUFS;
2134 
2135 	scope = ipv6_addr_scope(pfx);
2136 
2137 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2138 	if (addrconf_finite_timeout(timeout)) {
2139 		expires = jiffies_to_clock_t(timeout * HZ);
2140 		valid_lft = timeout;
2141 		flags = RTF_EXPIRES;
2142 	} else {
2143 		expires = 0;
2144 		flags = 0;
2145 		ifa_flags |= IFA_F_PERMANENT;
2146 	}
2147 
2148 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2149 	if (addrconf_finite_timeout(timeout)) {
2150 		if (timeout == 0)
2151 			ifa_flags |= IFA_F_DEPRECATED;
2152 		prefered_lft = timeout;
2153 	}
2154 
2155 	ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2156 
2157 	if (!IS_ERR(ifp)) {
2158 		spin_lock_bh(&ifp->lock);
2159 		ifp->valid_lft = valid_lft;
2160 		ifp->prefered_lft = prefered_lft;
2161 		ifp->tstamp = jiffies;
2162 		spin_unlock_bh(&ifp->lock);
2163 
2164 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2165 				      expires, flags);
2166 		/*
2167 		 * Note that section 3.1 of RFC 4429 indicates
2168 		 * that the Optimistic flag should not be set for
2169 		 * manually configured addresses
2170 		 */
2171 		addrconf_dad_start(ifp, 0);
2172 		in6_ifa_put(ifp);
2173 		addrconf_verify(0);
2174 		return 0;
2175 	}
2176 
2177 	return PTR_ERR(ifp);
2178 }
2179 
2180 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2181 			  unsigned int plen)
2182 {
2183 	struct inet6_ifaddr *ifp;
2184 	struct inet6_dev *idev;
2185 	struct net_device *dev;
2186 
2187 	if (plen > 128)
2188 		return -EINVAL;
2189 
2190 	dev = __dev_get_by_index(net, ifindex);
2191 	if (!dev)
2192 		return -ENODEV;
2193 
2194 	if ((idev = __in6_dev_get(dev)) == NULL)
2195 		return -ENXIO;
2196 
2197 	read_lock_bh(&idev->lock);
2198 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2199 		if (ifp->prefix_len == plen &&
2200 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2201 			in6_ifa_hold(ifp);
2202 			read_unlock_bh(&idev->lock);
2203 
2204 			ipv6_del_addr(ifp);
2205 
2206 			/* If the last address is deleted administratively,
2207 			   disable IPv6 on this interface.
2208 			 */
2209 			if (list_empty(&idev->addr_list))
2210 				addrconf_ifdown(idev->dev, 1);
2211 			return 0;
2212 		}
2213 	}
2214 	read_unlock_bh(&idev->lock);
2215 	return -EADDRNOTAVAIL;
2216 }
2217 
2218 
2219 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2220 {
2221 	struct in6_ifreq ireq;
2222 	int err;
2223 
2224 	if (!capable(CAP_NET_ADMIN))
2225 		return -EPERM;
2226 
2227 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2228 		return -EFAULT;
2229 
2230 	rtnl_lock();
2231 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2232 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2233 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2234 	rtnl_unlock();
2235 	return err;
2236 }
2237 
2238 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2239 {
2240 	struct in6_ifreq ireq;
2241 	int err;
2242 
2243 	if (!capable(CAP_NET_ADMIN))
2244 		return -EPERM;
2245 
2246 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2247 		return -EFAULT;
2248 
2249 	rtnl_lock();
2250 	err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2251 			     ireq.ifr6_prefixlen);
2252 	rtnl_unlock();
2253 	return err;
2254 }
2255 
2256 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2257 		     int plen, int scope)
2258 {
2259 	struct inet6_ifaddr *ifp;
2260 
2261 	ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2262 	if (!IS_ERR(ifp)) {
2263 		spin_lock_bh(&ifp->lock);
2264 		ifp->flags &= ~IFA_F_TENTATIVE;
2265 		spin_unlock_bh(&ifp->lock);
2266 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2267 		in6_ifa_put(ifp);
2268 	}
2269 }
2270 
2271 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2272 static void sit_add_v4_addrs(struct inet6_dev *idev)
2273 {
2274 	struct in6_addr addr;
2275 	struct net_device *dev;
2276 	struct net *net = dev_net(idev->dev);
2277 	int scope;
2278 
2279 	ASSERT_RTNL();
2280 
2281 	memset(&addr, 0, sizeof(struct in6_addr));
2282 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2283 
2284 	if (idev->dev->flags&IFF_POINTOPOINT) {
2285 		addr.s6_addr32[0] = htonl(0xfe800000);
2286 		scope = IFA_LINK;
2287 	} else {
2288 		scope = IPV6_ADDR_COMPATv4;
2289 	}
2290 
2291 	if (addr.s6_addr32[3]) {
2292 		add_addr(idev, &addr, 128, scope);
2293 		return;
2294 	}
2295 
2296 	for_each_netdev(net, dev) {
2297 		struct in_device * in_dev = __in_dev_get_rtnl(dev);
2298 		if (in_dev && (dev->flags & IFF_UP)) {
2299 			struct in_ifaddr * ifa;
2300 
2301 			int flag = scope;
2302 
2303 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2304 				int plen;
2305 
2306 				addr.s6_addr32[3] = ifa->ifa_local;
2307 
2308 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2309 					continue;
2310 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2311 					if (idev->dev->flags&IFF_POINTOPOINT)
2312 						continue;
2313 					flag |= IFA_HOST;
2314 				}
2315 				if (idev->dev->flags&IFF_POINTOPOINT)
2316 					plen = 64;
2317 				else
2318 					plen = 96;
2319 
2320 				add_addr(idev, &addr, plen, flag);
2321 			}
2322 		}
2323 	}
2324 }
2325 #endif
2326 
2327 static void init_loopback(struct net_device *dev)
2328 {
2329 	struct inet6_dev  *idev;
2330 
2331 	/* ::1 */
2332 
2333 	ASSERT_RTNL();
2334 
2335 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2336 		printk(KERN_DEBUG "init loopback: add_dev failed\n");
2337 		return;
2338 	}
2339 
2340 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2341 }
2342 
2343 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2344 {
2345 	struct inet6_ifaddr * ifp;
2346 	u32 addr_flags = IFA_F_PERMANENT;
2347 
2348 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2349 	if (idev->cnf.optimistic_dad &&
2350 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2351 		addr_flags |= IFA_F_OPTIMISTIC;
2352 #endif
2353 
2354 
2355 	ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2356 	if (!IS_ERR(ifp)) {
2357 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2358 		addrconf_dad_start(ifp, 0);
2359 		in6_ifa_put(ifp);
2360 	}
2361 }
2362 
2363 static void addrconf_dev_config(struct net_device *dev)
2364 {
2365 	struct in6_addr addr;
2366 	struct inet6_dev    * idev;
2367 
2368 	ASSERT_RTNL();
2369 
2370 	if ((dev->type != ARPHRD_ETHER) &&
2371 	    (dev->type != ARPHRD_FDDI) &&
2372 	    (dev->type != ARPHRD_IEEE802_TR) &&
2373 	    (dev->type != ARPHRD_ARCNET) &&
2374 	    (dev->type != ARPHRD_INFINIBAND)) {
2375 		/* Alas, we support only Ethernet autoconfiguration. */
2376 		return;
2377 	}
2378 
2379 	idev = addrconf_add_dev(dev);
2380 	if (idev == NULL)
2381 		return;
2382 
2383 	memset(&addr, 0, sizeof(struct in6_addr));
2384 	addr.s6_addr32[0] = htonl(0xFE800000);
2385 
2386 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2387 		addrconf_add_linklocal(idev, &addr);
2388 }
2389 
2390 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2391 static void addrconf_sit_config(struct net_device *dev)
2392 {
2393 	struct inet6_dev *idev;
2394 
2395 	ASSERT_RTNL();
2396 
2397 	/*
2398 	 * Configure the tunnel with one of our IPv4
2399 	 * addresses... we should configure all of
2400 	 * our v4 addrs in the tunnel
2401 	 */
2402 
2403 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2404 		printk(KERN_DEBUG "init sit: add_dev failed\n");
2405 		return;
2406 	}
2407 
2408 	if (dev->priv_flags & IFF_ISATAP) {
2409 		struct in6_addr addr;
2410 
2411 		ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2412 		addrconf_prefix_route(&addr, 64, dev, 0, 0);
2413 		if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2414 			addrconf_add_linklocal(idev, &addr);
2415 		return;
2416 	}
2417 
2418 	sit_add_v4_addrs(idev);
2419 
2420 	if (dev->flags&IFF_POINTOPOINT) {
2421 		addrconf_add_mroute(dev);
2422 		addrconf_add_lroute(dev);
2423 	} else
2424 		sit_route_add(dev);
2425 }
2426 #endif
2427 
2428 static inline int
2429 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2430 {
2431 	struct in6_addr lladdr;
2432 
2433 	if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2434 		addrconf_add_linklocal(idev, &lladdr);
2435 		return 0;
2436 	}
2437 	return -1;
2438 }
2439 
2440 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2441 {
2442 	struct net_device *link_dev;
2443 	struct net *net = dev_net(idev->dev);
2444 
2445 	/* first try to inherit the link-local address from the link device */
2446 	if (idev->dev->iflink &&
2447 	    (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2448 		if (!ipv6_inherit_linklocal(idev, link_dev))
2449 			return;
2450 	}
2451 	/* then try to inherit it from any device */
2452 	for_each_netdev(net, link_dev) {
2453 		if (!ipv6_inherit_linklocal(idev, link_dev))
2454 			return;
2455 	}
2456 	printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2457 }
2458 
2459 /*
2460  * Autoconfigure tunnel with a link-local address so routing protocols,
2461  * DHCPv6, MLD etc. can be run over the virtual link
2462  */
2463 
2464 static void addrconf_ip6_tnl_config(struct net_device *dev)
2465 {
2466 	struct inet6_dev *idev;
2467 
2468 	ASSERT_RTNL();
2469 
2470 	idev = addrconf_add_dev(dev);
2471 	if (!idev) {
2472 		printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2473 		return;
2474 	}
2475 	ip6_tnl_add_linklocal(idev);
2476 }
2477 
2478 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2479 			   void * data)
2480 {
2481 	struct net_device *dev = (struct net_device *) data;
2482 	struct inet6_dev *idev = __in6_dev_get(dev);
2483 	int run_pending = 0;
2484 	int err;
2485 
2486 	switch (event) {
2487 	case NETDEV_REGISTER:
2488 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2489 			idev = ipv6_add_dev(dev);
2490 			if (!idev)
2491 				return notifier_from_errno(-ENOMEM);
2492 		}
2493 		break;
2494 
2495 	case NETDEV_UP:
2496 	case NETDEV_CHANGE:
2497 		if (dev->flags & IFF_SLAVE)
2498 			break;
2499 
2500 		if (event == NETDEV_UP) {
2501 			if (!addrconf_qdisc_ok(dev)) {
2502 				/* device is not ready yet. */
2503 				printk(KERN_INFO
2504 					"ADDRCONF(NETDEV_UP): %s: "
2505 					"link is not ready\n",
2506 					dev->name);
2507 				break;
2508 			}
2509 
2510 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
2511 				idev = ipv6_add_dev(dev);
2512 
2513 			if (idev) {
2514 				idev->if_flags |= IF_READY;
2515 				run_pending = 1;
2516 			}
2517 		} else {
2518 			if (!addrconf_qdisc_ok(dev)) {
2519 				/* device is still not ready. */
2520 				break;
2521 			}
2522 
2523 			if (idev) {
2524 				if (idev->if_flags & IF_READY)
2525 					/* device is already configured. */
2526 					break;
2527 				idev->if_flags |= IF_READY;
2528 			}
2529 
2530 			printk(KERN_INFO
2531 					"ADDRCONF(NETDEV_CHANGE): %s: "
2532 					"link becomes ready\n",
2533 					dev->name);
2534 
2535 			run_pending = 1;
2536 		}
2537 
2538 		switch (dev->type) {
2539 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2540 		case ARPHRD_SIT:
2541 			addrconf_sit_config(dev);
2542 			break;
2543 #endif
2544 		case ARPHRD_TUNNEL6:
2545 			addrconf_ip6_tnl_config(dev);
2546 			break;
2547 		case ARPHRD_LOOPBACK:
2548 			init_loopback(dev);
2549 			break;
2550 
2551 		default:
2552 			addrconf_dev_config(dev);
2553 			break;
2554 		}
2555 
2556 		if (idev) {
2557 			if (run_pending)
2558 				addrconf_dad_run(idev);
2559 
2560 			/*
2561 			 * If the MTU changed during the interface down,
2562 			 * when the interface up, the changed MTU must be
2563 			 * reflected in the idev as well as routers.
2564 			 */
2565 			if (idev->cnf.mtu6 != dev->mtu &&
2566 			    dev->mtu >= IPV6_MIN_MTU) {
2567 				rt6_mtu_change(dev, dev->mtu);
2568 				idev->cnf.mtu6 = dev->mtu;
2569 			}
2570 			idev->tstamp = jiffies;
2571 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2572 
2573 			/*
2574 			 * If the changed mtu during down is lower than
2575 			 * IPV6_MIN_MTU stop IPv6 on this interface.
2576 			 */
2577 			if (dev->mtu < IPV6_MIN_MTU)
2578 				addrconf_ifdown(dev, 1);
2579 		}
2580 		break;
2581 
2582 	case NETDEV_CHANGEMTU:
2583 		if (idev && dev->mtu >= IPV6_MIN_MTU) {
2584 			rt6_mtu_change(dev, dev->mtu);
2585 			idev->cnf.mtu6 = dev->mtu;
2586 			break;
2587 		}
2588 
2589 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2590 			idev = ipv6_add_dev(dev);
2591 			if (idev)
2592 				break;
2593 		}
2594 
2595 		/*
2596 		 * MTU falled under IPV6_MIN_MTU.
2597 		 * Stop IPv6 on this interface.
2598 		 */
2599 
2600 	case NETDEV_DOWN:
2601 	case NETDEV_UNREGISTER:
2602 		/*
2603 		 *	Remove all addresses from this interface.
2604 		 */
2605 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2606 		break;
2607 
2608 	case NETDEV_CHANGENAME:
2609 		if (idev) {
2610 			snmp6_unregister_dev(idev);
2611 			addrconf_sysctl_unregister(idev);
2612 			addrconf_sysctl_register(idev);
2613 			err = snmp6_register_dev(idev);
2614 			if (err)
2615 				return notifier_from_errno(err);
2616 		}
2617 		break;
2618 
2619 	case NETDEV_PRE_TYPE_CHANGE:
2620 	case NETDEV_POST_TYPE_CHANGE:
2621 		addrconf_type_change(dev, event);
2622 		break;
2623 	}
2624 
2625 	return NOTIFY_OK;
2626 }
2627 
2628 /*
2629  *	addrconf module should be notified of a device going up
2630  */
2631 static struct notifier_block ipv6_dev_notf = {
2632 	.notifier_call = addrconf_notify,
2633 };
2634 
2635 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2636 {
2637 	struct inet6_dev *idev;
2638 	ASSERT_RTNL();
2639 
2640 	idev = __in6_dev_get(dev);
2641 
2642 	if (event == NETDEV_POST_TYPE_CHANGE)
2643 		ipv6_mc_remap(idev);
2644 	else if (event == NETDEV_PRE_TYPE_CHANGE)
2645 		ipv6_mc_unmap(idev);
2646 }
2647 
2648 static int addrconf_ifdown(struct net_device *dev, int how)
2649 {
2650 	struct net *net = dev_net(dev);
2651 	struct inet6_dev *idev;
2652 	struct inet6_ifaddr *ifa;
2653 	LIST_HEAD(keep_list);
2654 	int state;
2655 
2656 	ASSERT_RTNL();
2657 
2658 	rt6_ifdown(net, dev);
2659 	neigh_ifdown(&nd_tbl, dev);
2660 
2661 	idev = __in6_dev_get(dev);
2662 	if (idev == NULL)
2663 		return -ENODEV;
2664 
2665 	/*
2666 	 * Step 1: remove reference to ipv6 device from parent device.
2667 	 *	   Do not dev_put!
2668 	 */
2669 	if (how) {
2670 		idev->dead = 1;
2671 
2672 		/* protected by rtnl_lock */
2673 		rcu_assign_pointer(dev->ip6_ptr, NULL);
2674 
2675 		/* Step 1.5: remove snmp6 entry */
2676 		snmp6_unregister_dev(idev);
2677 
2678 	}
2679 
2680 	write_lock_bh(&idev->lock);
2681 
2682 	/* Step 2: clear flags for stateless addrconf */
2683 	if (!how)
2684 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2685 
2686 #ifdef CONFIG_IPV6_PRIVACY
2687 	if (how && del_timer(&idev->regen_timer))
2688 		in6_dev_put(idev);
2689 
2690 	/* Step 3: clear tempaddr list */
2691 	while (!list_empty(&idev->tempaddr_list)) {
2692 		ifa = list_first_entry(&idev->tempaddr_list,
2693 				       struct inet6_ifaddr, tmp_list);
2694 		list_del(&ifa->tmp_list);
2695 		write_unlock_bh(&idev->lock);
2696 		spin_lock_bh(&ifa->lock);
2697 
2698 		if (ifa->ifpub) {
2699 			in6_ifa_put(ifa->ifpub);
2700 			ifa->ifpub = NULL;
2701 		}
2702 		spin_unlock_bh(&ifa->lock);
2703 		in6_ifa_put(ifa);
2704 		write_lock_bh(&idev->lock);
2705 	}
2706 #endif
2707 
2708 	while (!list_empty(&idev->addr_list)) {
2709 		ifa = list_first_entry(&idev->addr_list,
2710 				       struct inet6_ifaddr, if_list);
2711 		addrconf_del_timer(ifa);
2712 
2713 		/* If just doing link down, and address is permanent
2714 		   and not link-local, then retain it. */
2715 		if (!how &&
2716 		    (ifa->flags&IFA_F_PERMANENT) &&
2717 		    !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2718 			list_move_tail(&ifa->if_list, &keep_list);
2719 
2720 			/* If not doing DAD on this address, just keep it. */
2721 			if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2722 			    idev->cnf.accept_dad <= 0 ||
2723 			    (ifa->flags & IFA_F_NODAD))
2724 				continue;
2725 
2726 			/* If it was tentative already, no need to notify */
2727 			if (ifa->flags & IFA_F_TENTATIVE)
2728 				continue;
2729 
2730 			/* Flag it for later restoration when link comes up */
2731 			ifa->flags |= IFA_F_TENTATIVE;
2732 			ifa->state = INET6_IFADDR_STATE_DAD;
2733 
2734 			write_unlock_bh(&idev->lock);
2735 
2736 			in6_ifa_hold(ifa);
2737 		} else {
2738 			list_del(&ifa->if_list);
2739 
2740 			/* clear hash table */
2741 			spin_lock_bh(&addrconf_hash_lock);
2742 			hlist_del_init_rcu(&ifa->addr_lst);
2743 			spin_unlock_bh(&addrconf_hash_lock);
2744 
2745 			write_unlock_bh(&idev->lock);
2746 			spin_lock_bh(&ifa->state_lock);
2747 			state = ifa->state;
2748 			ifa->state = INET6_IFADDR_STATE_DEAD;
2749 			spin_unlock_bh(&ifa->state_lock);
2750 
2751 			if (state == INET6_IFADDR_STATE_DEAD)
2752 				goto put_ifa;
2753 		}
2754 
2755 		__ipv6_ifa_notify(RTM_DELADDR, ifa);
2756 		if (ifa->state == INET6_IFADDR_STATE_DEAD)
2757 			atomic_notifier_call_chain(&inet6addr_chain,
2758 						   NETDEV_DOWN, ifa);
2759 
2760 put_ifa:
2761 		in6_ifa_put(ifa);
2762 
2763 		write_lock_bh(&idev->lock);
2764 	}
2765 
2766 	list_splice(&keep_list, &idev->addr_list);
2767 
2768 	write_unlock_bh(&idev->lock);
2769 
2770 	/* Step 5: Discard multicast list */
2771 	if (how)
2772 		ipv6_mc_destroy_dev(idev);
2773 	else
2774 		ipv6_mc_down(idev);
2775 
2776 	idev->tstamp = jiffies;
2777 
2778 	/* Last: Shot the device (if unregistered) */
2779 	if (how) {
2780 		addrconf_sysctl_unregister(idev);
2781 		neigh_parms_release(&nd_tbl, idev->nd_parms);
2782 		neigh_ifdown(&nd_tbl, dev);
2783 		in6_dev_put(idev);
2784 	}
2785 	return 0;
2786 }
2787 
2788 static void addrconf_rs_timer(unsigned long data)
2789 {
2790 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2791 	struct inet6_dev *idev = ifp->idev;
2792 
2793 	read_lock(&idev->lock);
2794 	if (idev->dead || !(idev->if_flags & IF_READY))
2795 		goto out;
2796 
2797 	if (idev->cnf.forwarding)
2798 		goto out;
2799 
2800 	/* Announcement received after solicitation was sent */
2801 	if (idev->if_flags & IF_RA_RCVD)
2802 		goto out;
2803 
2804 	spin_lock(&ifp->lock);
2805 	if (ifp->probes++ < idev->cnf.rtr_solicits) {
2806 		/* The wait after the last probe can be shorter */
2807 		addrconf_mod_timer(ifp, AC_RS,
2808 				   (ifp->probes == idev->cnf.rtr_solicits) ?
2809 				   idev->cnf.rtr_solicit_delay :
2810 				   idev->cnf.rtr_solicit_interval);
2811 		spin_unlock(&ifp->lock);
2812 
2813 		ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2814 	} else {
2815 		spin_unlock(&ifp->lock);
2816 		/*
2817 		 * Note: we do not support deprecated "all on-link"
2818 		 * assumption any longer.
2819 		 */
2820 		printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2821 		       idev->dev->name);
2822 	}
2823 
2824 out:
2825 	read_unlock(&idev->lock);
2826 	in6_ifa_put(ifp);
2827 }
2828 
2829 /*
2830  *	Duplicate Address Detection
2831  */
2832 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2833 {
2834 	unsigned long rand_num;
2835 	struct inet6_dev *idev = ifp->idev;
2836 
2837 	if (ifp->flags & IFA_F_OPTIMISTIC)
2838 		rand_num = 0;
2839 	else
2840 		rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2841 
2842 	ifp->probes = idev->cnf.dad_transmits;
2843 	addrconf_mod_timer(ifp, AC_DAD, rand_num);
2844 }
2845 
2846 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2847 {
2848 	struct inet6_dev *idev = ifp->idev;
2849 	struct net_device *dev = idev->dev;
2850 
2851 	addrconf_join_solict(dev, &ifp->addr);
2852 
2853 	net_srandom(ifp->addr.s6_addr32[3]);
2854 
2855 	read_lock_bh(&idev->lock);
2856 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
2857 		goto out;
2858 
2859 	spin_lock(&ifp->lock);
2860 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2861 	    idev->cnf.accept_dad < 1 ||
2862 	    !(ifp->flags&IFA_F_TENTATIVE) ||
2863 	    ifp->flags & IFA_F_NODAD) {
2864 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2865 		spin_unlock(&ifp->lock);
2866 		read_unlock_bh(&idev->lock);
2867 
2868 		addrconf_dad_completed(ifp);
2869 		return;
2870 	}
2871 
2872 	if (!(idev->if_flags & IF_READY)) {
2873 		spin_unlock(&ifp->lock);
2874 		read_unlock_bh(&idev->lock);
2875 		/*
2876 		 * If the device is not ready:
2877 		 * - keep it tentative if it is a permanent address.
2878 		 * - otherwise, kill it.
2879 		 */
2880 		in6_ifa_hold(ifp);
2881 		addrconf_dad_stop(ifp, 0);
2882 		return;
2883 	}
2884 
2885 	/*
2886 	 * Optimistic nodes can start receiving
2887 	 * Frames right away
2888 	 */
2889 	if (ifp->flags & IFA_F_OPTIMISTIC)
2890 		ip6_ins_rt(ifp->rt);
2891 
2892 	addrconf_dad_kick(ifp);
2893 	spin_unlock(&ifp->lock);
2894 out:
2895 	read_unlock_bh(&idev->lock);
2896 }
2897 
2898 static void addrconf_dad_timer(unsigned long data)
2899 {
2900 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2901 	struct inet6_dev *idev = ifp->idev;
2902 	struct in6_addr mcaddr;
2903 
2904 	if (!ifp->probes && addrconf_dad_end(ifp))
2905 		goto out;
2906 
2907 	read_lock(&idev->lock);
2908 	if (idev->dead || !(idev->if_flags & IF_READY)) {
2909 		read_unlock(&idev->lock);
2910 		goto out;
2911 	}
2912 
2913 	spin_lock(&ifp->lock);
2914 	if (ifp->probes == 0) {
2915 		/*
2916 		 * DAD was successful
2917 		 */
2918 
2919 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2920 		spin_unlock(&ifp->lock);
2921 		read_unlock(&idev->lock);
2922 
2923 		addrconf_dad_completed(ifp);
2924 
2925 		goto out;
2926 	}
2927 
2928 	ifp->probes--;
2929 	addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2930 	spin_unlock(&ifp->lock);
2931 	read_unlock(&idev->lock);
2932 
2933 	/* send a neighbour solicitation for our addr */
2934 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2935 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2936 out:
2937 	in6_ifa_put(ifp);
2938 }
2939 
2940 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2941 {
2942 	struct net_device *dev = ifp->idev->dev;
2943 
2944 	/*
2945 	 *	Configure the address for reception. Now it is valid.
2946 	 */
2947 
2948 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
2949 
2950 	/* If added prefix is link local and forwarding is off,
2951 	   start sending router solicitations.
2952 	 */
2953 
2954 	if (ifp->idev->cnf.forwarding == 0 &&
2955 	    ifp->idev->cnf.rtr_solicits > 0 &&
2956 	    (dev->flags&IFF_LOOPBACK) == 0 &&
2957 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2958 		/*
2959 		 *	If a host as already performed a random delay
2960 		 *	[...] as part of DAD [...] there is no need
2961 		 *	to delay again before sending the first RS
2962 		 */
2963 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2964 
2965 		spin_lock_bh(&ifp->lock);
2966 		ifp->probes = 1;
2967 		ifp->idev->if_flags |= IF_RS_SENT;
2968 		addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2969 		spin_unlock_bh(&ifp->lock);
2970 	}
2971 }
2972 
2973 static void addrconf_dad_run(struct inet6_dev *idev)
2974 {
2975 	struct inet6_ifaddr *ifp;
2976 
2977 	read_lock_bh(&idev->lock);
2978 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2979 		spin_lock(&ifp->lock);
2980 		if (ifp->flags & IFA_F_TENTATIVE &&
2981 		    ifp->state == INET6_IFADDR_STATE_DAD)
2982 			addrconf_dad_kick(ifp);
2983 		spin_unlock(&ifp->lock);
2984 	}
2985 	read_unlock_bh(&idev->lock);
2986 }
2987 
2988 #ifdef CONFIG_PROC_FS
2989 struct if6_iter_state {
2990 	struct seq_net_private p;
2991 	int bucket;
2992 };
2993 
2994 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2995 {
2996 	struct inet6_ifaddr *ifa = NULL;
2997 	struct if6_iter_state *state = seq->private;
2998 	struct net *net = seq_file_net(seq);
2999 
3000 	for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3001 		struct hlist_node *n;
3002 		hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3003 					 addr_lst)
3004 			if (net_eq(dev_net(ifa->idev->dev), net))
3005 				return ifa;
3006 	}
3007 	return NULL;
3008 }
3009 
3010 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3011 					 struct inet6_ifaddr *ifa)
3012 {
3013 	struct if6_iter_state *state = seq->private;
3014 	struct net *net = seq_file_net(seq);
3015 	struct hlist_node *n = &ifa->addr_lst;
3016 
3017 	hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3018 		if (net_eq(dev_net(ifa->idev->dev), net))
3019 			return ifa;
3020 
3021 	while (++state->bucket < IN6_ADDR_HSIZE) {
3022 		hlist_for_each_entry_rcu_bh(ifa, n,
3023 				     &inet6_addr_lst[state->bucket], addr_lst) {
3024 			if (net_eq(dev_net(ifa->idev->dev), net))
3025 				return ifa;
3026 		}
3027 	}
3028 
3029 	return NULL;
3030 }
3031 
3032 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3033 {
3034 	struct inet6_ifaddr *ifa = if6_get_first(seq);
3035 
3036 	if (ifa)
3037 		while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3038 			--pos;
3039 	return pos ? NULL : ifa;
3040 }
3041 
3042 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3043 	__acquires(rcu_bh)
3044 {
3045 	rcu_read_lock_bh();
3046 	return if6_get_idx(seq, *pos);
3047 }
3048 
3049 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3050 {
3051 	struct inet6_ifaddr *ifa;
3052 
3053 	ifa = if6_get_next(seq, v);
3054 	++*pos;
3055 	return ifa;
3056 }
3057 
3058 static void if6_seq_stop(struct seq_file *seq, void *v)
3059 	__releases(rcu_bh)
3060 {
3061 	rcu_read_unlock_bh();
3062 }
3063 
3064 static int if6_seq_show(struct seq_file *seq, void *v)
3065 {
3066 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3067 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3068 		   &ifp->addr,
3069 		   ifp->idev->dev->ifindex,
3070 		   ifp->prefix_len,
3071 		   ifp->scope,
3072 		   ifp->flags,
3073 		   ifp->idev->dev->name);
3074 	return 0;
3075 }
3076 
3077 static const struct seq_operations if6_seq_ops = {
3078 	.start	= if6_seq_start,
3079 	.next	= if6_seq_next,
3080 	.show	= if6_seq_show,
3081 	.stop	= if6_seq_stop,
3082 };
3083 
3084 static int if6_seq_open(struct inode *inode, struct file *file)
3085 {
3086 	return seq_open_net(inode, file, &if6_seq_ops,
3087 			    sizeof(struct if6_iter_state));
3088 }
3089 
3090 static const struct file_operations if6_fops = {
3091 	.owner		= THIS_MODULE,
3092 	.open		= if6_seq_open,
3093 	.read		= seq_read,
3094 	.llseek		= seq_lseek,
3095 	.release	= seq_release_net,
3096 };
3097 
3098 static int __net_init if6_proc_net_init(struct net *net)
3099 {
3100 	if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3101 		return -ENOMEM;
3102 	return 0;
3103 }
3104 
3105 static void __net_exit if6_proc_net_exit(struct net *net)
3106 {
3107        proc_net_remove(net, "if_inet6");
3108 }
3109 
3110 static struct pernet_operations if6_proc_net_ops = {
3111        .init = if6_proc_net_init,
3112        .exit = if6_proc_net_exit,
3113 };
3114 
3115 int __init if6_proc_init(void)
3116 {
3117 	return register_pernet_subsys(&if6_proc_net_ops);
3118 }
3119 
3120 void if6_proc_exit(void)
3121 {
3122 	unregister_pernet_subsys(&if6_proc_net_ops);
3123 }
3124 #endif	/* CONFIG_PROC_FS */
3125 
3126 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3127 /* Check if address is a home address configured on any interface. */
3128 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3129 {
3130 	int ret = 0;
3131 	struct inet6_ifaddr *ifp = NULL;
3132 	struct hlist_node *n;
3133 	unsigned int hash = ipv6_addr_hash(addr);
3134 
3135 	rcu_read_lock_bh();
3136 	hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3137 		if (!net_eq(dev_net(ifp->idev->dev), net))
3138 			continue;
3139 		if (ipv6_addr_equal(&ifp->addr, addr) &&
3140 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3141 			ret = 1;
3142 			break;
3143 		}
3144 	}
3145 	rcu_read_unlock_bh();
3146 	return ret;
3147 }
3148 #endif
3149 
3150 /*
3151  *	Periodic address status verification
3152  */
3153 
3154 static void addrconf_verify(unsigned long foo)
3155 {
3156 	unsigned long now, next, next_sec, next_sched;
3157 	struct inet6_ifaddr *ifp;
3158 	struct hlist_node *node;
3159 	int i;
3160 
3161 	rcu_read_lock_bh();
3162 	spin_lock(&addrconf_verify_lock);
3163 	now = jiffies;
3164 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3165 
3166 	del_timer(&addr_chk_timer);
3167 
3168 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3169 restart:
3170 		hlist_for_each_entry_rcu_bh(ifp, node,
3171 					 &inet6_addr_lst[i], addr_lst) {
3172 			unsigned long age;
3173 
3174 			if (ifp->flags & IFA_F_PERMANENT)
3175 				continue;
3176 
3177 			spin_lock(&ifp->lock);
3178 			/* We try to batch several events at once. */
3179 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3180 
3181 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3182 			    age >= ifp->valid_lft) {
3183 				spin_unlock(&ifp->lock);
3184 				in6_ifa_hold(ifp);
3185 				ipv6_del_addr(ifp);
3186 				goto restart;
3187 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3188 				spin_unlock(&ifp->lock);
3189 				continue;
3190 			} else if (age >= ifp->prefered_lft) {
3191 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3192 				int deprecate = 0;
3193 
3194 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3195 					deprecate = 1;
3196 					ifp->flags |= IFA_F_DEPRECATED;
3197 				}
3198 
3199 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3200 					next = ifp->tstamp + ifp->valid_lft * HZ;
3201 
3202 				spin_unlock(&ifp->lock);
3203 
3204 				if (deprecate) {
3205 					in6_ifa_hold(ifp);
3206 
3207 					ipv6_ifa_notify(0, ifp);
3208 					in6_ifa_put(ifp);
3209 					goto restart;
3210 				}
3211 #ifdef CONFIG_IPV6_PRIVACY
3212 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3213 				   !(ifp->flags&IFA_F_TENTATIVE)) {
3214 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3215 					ifp->idev->cnf.dad_transmits *
3216 					ifp->idev->nd_parms->retrans_time / HZ;
3217 
3218 				if (age >= ifp->prefered_lft - regen_advance) {
3219 					struct inet6_ifaddr *ifpub = ifp->ifpub;
3220 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3221 						next = ifp->tstamp + ifp->prefered_lft * HZ;
3222 					if (!ifp->regen_count && ifpub) {
3223 						ifp->regen_count++;
3224 						in6_ifa_hold(ifp);
3225 						in6_ifa_hold(ifpub);
3226 						spin_unlock(&ifp->lock);
3227 
3228 						spin_lock(&ifpub->lock);
3229 						ifpub->regen_count = 0;
3230 						spin_unlock(&ifpub->lock);
3231 						ipv6_create_tempaddr(ifpub, ifp);
3232 						in6_ifa_put(ifpub);
3233 						in6_ifa_put(ifp);
3234 						goto restart;
3235 					}
3236 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3237 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3238 				spin_unlock(&ifp->lock);
3239 #endif
3240 			} else {
3241 				/* ifp->prefered_lft <= ifp->valid_lft */
3242 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3243 					next = ifp->tstamp + ifp->prefered_lft * HZ;
3244 				spin_unlock(&ifp->lock);
3245 			}
3246 		}
3247 	}
3248 
3249 	next_sec = round_jiffies_up(next);
3250 	next_sched = next;
3251 
3252 	/* If rounded timeout is accurate enough, accept it. */
3253 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3254 		next_sched = next_sec;
3255 
3256 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3257 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3258 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3259 
3260 	ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3261 	      now, next, next_sec, next_sched));
3262 
3263 	addr_chk_timer.expires = next_sched;
3264 	add_timer(&addr_chk_timer);
3265 	spin_unlock(&addrconf_verify_lock);
3266 	rcu_read_unlock_bh();
3267 }
3268 
3269 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3270 {
3271 	struct in6_addr *pfx = NULL;
3272 
3273 	if (addr)
3274 		pfx = nla_data(addr);
3275 
3276 	if (local) {
3277 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3278 			pfx = NULL;
3279 		else
3280 			pfx = nla_data(local);
3281 	}
3282 
3283 	return pfx;
3284 }
3285 
3286 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3287 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
3288 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
3289 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
3290 };
3291 
3292 static int
3293 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3294 {
3295 	struct net *net = sock_net(skb->sk);
3296 	struct ifaddrmsg *ifm;
3297 	struct nlattr *tb[IFA_MAX+1];
3298 	struct in6_addr *pfx;
3299 	int err;
3300 
3301 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3302 	if (err < 0)
3303 		return err;
3304 
3305 	ifm = nlmsg_data(nlh);
3306 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3307 	if (pfx == NULL)
3308 		return -EINVAL;
3309 
3310 	return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3311 }
3312 
3313 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3314 			     u32 prefered_lft, u32 valid_lft)
3315 {
3316 	u32 flags;
3317 	clock_t expires;
3318 	unsigned long timeout;
3319 
3320 	if (!valid_lft || (prefered_lft > valid_lft))
3321 		return -EINVAL;
3322 
3323 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
3324 	if (addrconf_finite_timeout(timeout)) {
3325 		expires = jiffies_to_clock_t(timeout * HZ);
3326 		valid_lft = timeout;
3327 		flags = RTF_EXPIRES;
3328 	} else {
3329 		expires = 0;
3330 		flags = 0;
3331 		ifa_flags |= IFA_F_PERMANENT;
3332 	}
3333 
3334 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3335 	if (addrconf_finite_timeout(timeout)) {
3336 		if (timeout == 0)
3337 			ifa_flags |= IFA_F_DEPRECATED;
3338 		prefered_lft = timeout;
3339 	}
3340 
3341 	spin_lock_bh(&ifp->lock);
3342 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3343 	ifp->tstamp = jiffies;
3344 	ifp->valid_lft = valid_lft;
3345 	ifp->prefered_lft = prefered_lft;
3346 
3347 	spin_unlock_bh(&ifp->lock);
3348 	if (!(ifp->flags&IFA_F_TENTATIVE))
3349 		ipv6_ifa_notify(0, ifp);
3350 
3351 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3352 			      expires, flags);
3353 	addrconf_verify(0);
3354 
3355 	return 0;
3356 }
3357 
3358 static int
3359 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3360 {
3361 	struct net *net = sock_net(skb->sk);
3362 	struct ifaddrmsg *ifm;
3363 	struct nlattr *tb[IFA_MAX+1];
3364 	struct in6_addr *pfx;
3365 	struct inet6_ifaddr *ifa;
3366 	struct net_device *dev;
3367 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3368 	u8 ifa_flags;
3369 	int err;
3370 
3371 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3372 	if (err < 0)
3373 		return err;
3374 
3375 	ifm = nlmsg_data(nlh);
3376 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3377 	if (pfx == NULL)
3378 		return -EINVAL;
3379 
3380 	if (tb[IFA_CACHEINFO]) {
3381 		struct ifa_cacheinfo *ci;
3382 
3383 		ci = nla_data(tb[IFA_CACHEINFO]);
3384 		valid_lft = ci->ifa_valid;
3385 		preferred_lft = ci->ifa_prefered;
3386 	} else {
3387 		preferred_lft = INFINITY_LIFE_TIME;
3388 		valid_lft = INFINITY_LIFE_TIME;
3389 	}
3390 
3391 	dev =  __dev_get_by_index(net, ifm->ifa_index);
3392 	if (dev == NULL)
3393 		return -ENODEV;
3394 
3395 	/* We ignore other flags so far. */
3396 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3397 
3398 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3399 	if (ifa == NULL) {
3400 		/*
3401 		 * It would be best to check for !NLM_F_CREATE here but
3402 		 * userspace alreay relies on not having to provide this.
3403 		 */
3404 		return inet6_addr_add(net, ifm->ifa_index, pfx,
3405 				      ifm->ifa_prefixlen, ifa_flags,
3406 				      preferred_lft, valid_lft);
3407 	}
3408 
3409 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3410 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3411 		err = -EEXIST;
3412 	else
3413 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3414 
3415 	in6_ifa_put(ifa);
3416 
3417 	return err;
3418 }
3419 
3420 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3421 			  u8 scope, int ifindex)
3422 {
3423 	struct ifaddrmsg *ifm;
3424 
3425 	ifm = nlmsg_data(nlh);
3426 	ifm->ifa_family = AF_INET6;
3427 	ifm->ifa_prefixlen = prefixlen;
3428 	ifm->ifa_flags = flags;
3429 	ifm->ifa_scope = scope;
3430 	ifm->ifa_index = ifindex;
3431 }
3432 
3433 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3434 			 unsigned long tstamp, u32 preferred, u32 valid)
3435 {
3436 	struct ifa_cacheinfo ci;
3437 
3438 	ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3439 			+ TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3440 	ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3441 			+ TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3442 	ci.ifa_prefered = preferred;
3443 	ci.ifa_valid = valid;
3444 
3445 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3446 }
3447 
3448 static inline int rt_scope(int ifa_scope)
3449 {
3450 	if (ifa_scope & IFA_HOST)
3451 		return RT_SCOPE_HOST;
3452 	else if (ifa_scope & IFA_LINK)
3453 		return RT_SCOPE_LINK;
3454 	else if (ifa_scope & IFA_SITE)
3455 		return RT_SCOPE_SITE;
3456 	else
3457 		return RT_SCOPE_UNIVERSE;
3458 }
3459 
3460 static inline int inet6_ifaddr_msgsize(void)
3461 {
3462 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3463 	       + nla_total_size(16) /* IFA_ADDRESS */
3464 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3465 }
3466 
3467 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3468 			     u32 pid, u32 seq, int event, unsigned int flags)
3469 {
3470 	struct nlmsghdr  *nlh;
3471 	u32 preferred, valid;
3472 
3473 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3474 	if (nlh == NULL)
3475 		return -EMSGSIZE;
3476 
3477 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3478 		      ifa->idev->dev->ifindex);
3479 
3480 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3481 		preferred = ifa->prefered_lft;
3482 		valid = ifa->valid_lft;
3483 		if (preferred != INFINITY_LIFE_TIME) {
3484 			long tval = (jiffies - ifa->tstamp)/HZ;
3485 			if (preferred > tval)
3486 				preferred -= tval;
3487 			else
3488 				preferred = 0;
3489 			if (valid != INFINITY_LIFE_TIME)
3490 				valid -= tval;
3491 		}
3492 	} else {
3493 		preferred = INFINITY_LIFE_TIME;
3494 		valid = INFINITY_LIFE_TIME;
3495 	}
3496 
3497 	if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3498 	    put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3499 		nlmsg_cancel(skb, nlh);
3500 		return -EMSGSIZE;
3501 	}
3502 
3503 	return nlmsg_end(skb, nlh);
3504 }
3505 
3506 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3507 				u32 pid, u32 seq, int event, u16 flags)
3508 {
3509 	struct nlmsghdr  *nlh;
3510 	u8 scope = RT_SCOPE_UNIVERSE;
3511 	int ifindex = ifmca->idev->dev->ifindex;
3512 
3513 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3514 		scope = RT_SCOPE_SITE;
3515 
3516 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3517 	if (nlh == NULL)
3518 		return -EMSGSIZE;
3519 
3520 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3521 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3522 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3523 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3524 		nlmsg_cancel(skb, nlh);
3525 		return -EMSGSIZE;
3526 	}
3527 
3528 	return nlmsg_end(skb, nlh);
3529 }
3530 
3531 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3532 				u32 pid, u32 seq, int event, unsigned int flags)
3533 {
3534 	struct nlmsghdr  *nlh;
3535 	u8 scope = RT_SCOPE_UNIVERSE;
3536 	int ifindex = ifaca->aca_idev->dev->ifindex;
3537 
3538 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3539 		scope = RT_SCOPE_SITE;
3540 
3541 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3542 	if (nlh == NULL)
3543 		return -EMSGSIZE;
3544 
3545 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3546 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3547 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3548 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3549 		nlmsg_cancel(skb, nlh);
3550 		return -EMSGSIZE;
3551 	}
3552 
3553 	return nlmsg_end(skb, nlh);
3554 }
3555 
3556 enum addr_type_t {
3557 	UNICAST_ADDR,
3558 	MULTICAST_ADDR,
3559 	ANYCAST_ADDR,
3560 };
3561 
3562 /* called with rcu_read_lock() */
3563 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3564 			  struct netlink_callback *cb, enum addr_type_t type,
3565 			  int s_ip_idx, int *p_ip_idx)
3566 {
3567 	struct ifmcaddr6 *ifmca;
3568 	struct ifacaddr6 *ifaca;
3569 	int err = 1;
3570 	int ip_idx = *p_ip_idx;
3571 
3572 	read_lock_bh(&idev->lock);
3573 	switch (type) {
3574 	case UNICAST_ADDR: {
3575 		struct inet6_ifaddr *ifa;
3576 
3577 		/* unicast address incl. temp addr */
3578 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
3579 			if (++ip_idx < s_ip_idx)
3580 				continue;
3581 			err = inet6_fill_ifaddr(skb, ifa,
3582 						NETLINK_CB(cb->skb).pid,
3583 						cb->nlh->nlmsg_seq,
3584 						RTM_NEWADDR,
3585 						NLM_F_MULTI);
3586 			if (err <= 0)
3587 				break;
3588 		}
3589 		break;
3590 	}
3591 	case MULTICAST_ADDR:
3592 		/* multicast address */
3593 		for (ifmca = idev->mc_list; ifmca;
3594 		     ifmca = ifmca->next, ip_idx++) {
3595 			if (ip_idx < s_ip_idx)
3596 				continue;
3597 			err = inet6_fill_ifmcaddr(skb, ifmca,
3598 						  NETLINK_CB(cb->skb).pid,
3599 						  cb->nlh->nlmsg_seq,
3600 						  RTM_GETMULTICAST,
3601 						  NLM_F_MULTI);
3602 			if (err <= 0)
3603 				break;
3604 		}
3605 		break;
3606 	case ANYCAST_ADDR:
3607 		/* anycast address */
3608 		for (ifaca = idev->ac_list; ifaca;
3609 		     ifaca = ifaca->aca_next, ip_idx++) {
3610 			if (ip_idx < s_ip_idx)
3611 				continue;
3612 			err = inet6_fill_ifacaddr(skb, ifaca,
3613 						  NETLINK_CB(cb->skb).pid,
3614 						  cb->nlh->nlmsg_seq,
3615 						  RTM_GETANYCAST,
3616 						  NLM_F_MULTI);
3617 			if (err <= 0)
3618 				break;
3619 		}
3620 		break;
3621 	default:
3622 		break;
3623 	}
3624 	read_unlock_bh(&idev->lock);
3625 	*p_ip_idx = ip_idx;
3626 	return err;
3627 }
3628 
3629 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3630 			   enum addr_type_t type)
3631 {
3632 	struct net *net = sock_net(skb->sk);
3633 	int h, s_h;
3634 	int idx, ip_idx;
3635 	int s_idx, s_ip_idx;
3636 	struct net_device *dev;
3637 	struct inet6_dev *idev;
3638 	struct hlist_head *head;
3639 	struct hlist_node *node;
3640 
3641 	s_h = cb->args[0];
3642 	s_idx = idx = cb->args[1];
3643 	s_ip_idx = ip_idx = cb->args[2];
3644 
3645 	rcu_read_lock();
3646 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3647 		idx = 0;
3648 		head = &net->dev_index_head[h];
3649 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3650 			if (idx < s_idx)
3651 				goto cont;
3652 			if (h > s_h || idx > s_idx)
3653 				s_ip_idx = 0;
3654 			ip_idx = 0;
3655 			idev = __in6_dev_get(dev);
3656 			if (!idev)
3657 				goto cont;
3658 
3659 			if (in6_dump_addrs(idev, skb, cb, type,
3660 					   s_ip_idx, &ip_idx) <= 0)
3661 				goto done;
3662 cont:
3663 			idx++;
3664 		}
3665 	}
3666 done:
3667 	rcu_read_unlock();
3668 	cb->args[0] = h;
3669 	cb->args[1] = idx;
3670 	cb->args[2] = ip_idx;
3671 
3672 	return skb->len;
3673 }
3674 
3675 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3676 {
3677 	enum addr_type_t type = UNICAST_ADDR;
3678 
3679 	return inet6_dump_addr(skb, cb, type);
3680 }
3681 
3682 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3683 {
3684 	enum addr_type_t type = MULTICAST_ADDR;
3685 
3686 	return inet6_dump_addr(skb, cb, type);
3687 }
3688 
3689 
3690 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3691 {
3692 	enum addr_type_t type = ANYCAST_ADDR;
3693 
3694 	return inet6_dump_addr(skb, cb, type);
3695 }
3696 
3697 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3698 			     void *arg)
3699 {
3700 	struct net *net = sock_net(in_skb->sk);
3701 	struct ifaddrmsg *ifm;
3702 	struct nlattr *tb[IFA_MAX+1];
3703 	struct in6_addr *addr = NULL;
3704 	struct net_device *dev = NULL;
3705 	struct inet6_ifaddr *ifa;
3706 	struct sk_buff *skb;
3707 	int err;
3708 
3709 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3710 	if (err < 0)
3711 		goto errout;
3712 
3713 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3714 	if (addr == NULL) {
3715 		err = -EINVAL;
3716 		goto errout;
3717 	}
3718 
3719 	ifm = nlmsg_data(nlh);
3720 	if (ifm->ifa_index)
3721 		dev = __dev_get_by_index(net, ifm->ifa_index);
3722 
3723 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3724 	if (!ifa) {
3725 		err = -EADDRNOTAVAIL;
3726 		goto errout;
3727 	}
3728 
3729 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3730 	if (!skb) {
3731 		err = -ENOBUFS;
3732 		goto errout_ifa;
3733 	}
3734 
3735 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3736 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
3737 	if (err < 0) {
3738 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3739 		WARN_ON(err == -EMSGSIZE);
3740 		kfree_skb(skb);
3741 		goto errout_ifa;
3742 	}
3743 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3744 errout_ifa:
3745 	in6_ifa_put(ifa);
3746 errout:
3747 	return err;
3748 }
3749 
3750 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3751 {
3752 	struct sk_buff *skb;
3753 	struct net *net = dev_net(ifa->idev->dev);
3754 	int err = -ENOBUFS;
3755 
3756 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3757 	if (skb == NULL)
3758 		goto errout;
3759 
3760 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3761 	if (err < 0) {
3762 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3763 		WARN_ON(err == -EMSGSIZE);
3764 		kfree_skb(skb);
3765 		goto errout;
3766 	}
3767 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3768 	return;
3769 errout:
3770 	if (err < 0)
3771 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3772 }
3773 
3774 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3775 				__s32 *array, int bytes)
3776 {
3777 	BUG_ON(bytes < (DEVCONF_MAX * 4));
3778 
3779 	memset(array, 0, bytes);
3780 	array[DEVCONF_FORWARDING] = cnf->forwarding;
3781 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3782 	array[DEVCONF_MTU6] = cnf->mtu6;
3783 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3784 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3785 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
3786 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3787 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3788 	array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3789 	array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3790 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3791 #ifdef CONFIG_IPV6_PRIVACY
3792 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3793 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3794 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3795 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3796 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3797 #endif
3798 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3799 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3800 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3801 #ifdef CONFIG_IPV6_ROUTER_PREF
3802 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3803 	array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3804 #ifdef CONFIG_IPV6_ROUTE_INFO
3805 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3806 #endif
3807 #endif
3808 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3809 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3810 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3811 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3812 #endif
3813 #ifdef CONFIG_IPV6_MROUTE
3814 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3815 #endif
3816 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3817 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3818 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3819 }
3820 
3821 static inline size_t inet6_if_nlmsg_size(void)
3822 {
3823 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3824 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3825 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3826 	       + nla_total_size(4) /* IFLA_MTU */
3827 	       + nla_total_size(4) /* IFLA_LINK */
3828 	       + nla_total_size( /* IFLA_PROTINFO */
3829 			nla_total_size(4) /* IFLA_INET6_FLAGS */
3830 			+ nla_total_size(sizeof(struct ifla_cacheinfo))
3831 			+ nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3832 			+ nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3833 			+ nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3834 		 );
3835 }
3836 
3837 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3838 				      int items, int bytes)
3839 {
3840 	int i;
3841 	int pad = bytes - sizeof(u64) * items;
3842 	BUG_ON(pad < 0);
3843 
3844 	/* Use put_unaligned() because stats may not be aligned for u64. */
3845 	put_unaligned(items, &stats[0]);
3846 	for (i = 1; i < items; i++)
3847 		put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3848 
3849 	memset(&stats[items], 0, pad);
3850 }
3851 
3852 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3853 			     int bytes)
3854 {
3855 	switch (attrtype) {
3856 	case IFLA_INET6_STATS:
3857 		__snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3858 		break;
3859 	case IFLA_INET6_ICMP6STATS:
3860 		__snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3861 		break;
3862 	}
3863 }
3864 
3865 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3866 			     u32 pid, u32 seq, int event, unsigned int flags)
3867 {
3868 	struct net_device *dev = idev->dev;
3869 	struct nlattr *nla;
3870 	struct ifinfomsg *hdr;
3871 	struct nlmsghdr *nlh;
3872 	void *protoinfo;
3873 	struct ifla_cacheinfo ci;
3874 
3875 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3876 	if (nlh == NULL)
3877 		return -EMSGSIZE;
3878 
3879 	hdr = nlmsg_data(nlh);
3880 	hdr->ifi_family = AF_INET6;
3881 	hdr->__ifi_pad = 0;
3882 	hdr->ifi_type = dev->type;
3883 	hdr->ifi_index = dev->ifindex;
3884 	hdr->ifi_flags = dev_get_flags(dev);
3885 	hdr->ifi_change = 0;
3886 
3887 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3888 
3889 	if (dev->addr_len)
3890 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3891 
3892 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3893 	if (dev->ifindex != dev->iflink)
3894 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3895 
3896 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3897 	if (protoinfo == NULL)
3898 		goto nla_put_failure;
3899 
3900 	NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3901 
3902 	ci.max_reasm_len = IPV6_MAXPLEN;
3903 	ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3904 		    + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3905 	ci.reachable_time = idev->nd_parms->reachable_time;
3906 	ci.retrans_time = idev->nd_parms->retrans_time;
3907 	NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3908 
3909 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3910 	if (nla == NULL)
3911 		goto nla_put_failure;
3912 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3913 
3914 	/* XXX - MC not implemented */
3915 
3916 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3917 	if (nla == NULL)
3918 		goto nla_put_failure;
3919 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3920 
3921 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3922 	if (nla == NULL)
3923 		goto nla_put_failure;
3924 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3925 
3926 	nla_nest_end(skb, protoinfo);
3927 	return nlmsg_end(skb, nlh);
3928 
3929 nla_put_failure:
3930 	nlmsg_cancel(skb, nlh);
3931 	return -EMSGSIZE;
3932 }
3933 
3934 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3935 {
3936 	struct net *net = sock_net(skb->sk);
3937 	int h, s_h;
3938 	int idx = 0, s_idx;
3939 	struct net_device *dev;
3940 	struct inet6_dev *idev;
3941 	struct hlist_head *head;
3942 	struct hlist_node *node;
3943 
3944 	s_h = cb->args[0];
3945 	s_idx = cb->args[1];
3946 
3947 	rcu_read_lock();
3948 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3949 		idx = 0;
3950 		head = &net->dev_index_head[h];
3951 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3952 			if (idx < s_idx)
3953 				goto cont;
3954 			idev = __in6_dev_get(dev);
3955 			if (!idev)
3956 				goto cont;
3957 			if (inet6_fill_ifinfo(skb, idev,
3958 					      NETLINK_CB(cb->skb).pid,
3959 					      cb->nlh->nlmsg_seq,
3960 					      RTM_NEWLINK, NLM_F_MULTI) <= 0)
3961 				goto out;
3962 cont:
3963 			idx++;
3964 		}
3965 	}
3966 out:
3967 	rcu_read_unlock();
3968 	cb->args[1] = idx;
3969 	cb->args[0] = h;
3970 
3971 	return skb->len;
3972 }
3973 
3974 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3975 {
3976 	struct sk_buff *skb;
3977 	struct net *net = dev_net(idev->dev);
3978 	int err = -ENOBUFS;
3979 
3980 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3981 	if (skb == NULL)
3982 		goto errout;
3983 
3984 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3985 	if (err < 0) {
3986 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3987 		WARN_ON(err == -EMSGSIZE);
3988 		kfree_skb(skb);
3989 		goto errout;
3990 	}
3991 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3992 	return;
3993 errout:
3994 	if (err < 0)
3995 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3996 }
3997 
3998 static inline size_t inet6_prefix_nlmsg_size(void)
3999 {
4000 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
4001 	       + nla_total_size(sizeof(struct in6_addr))
4002 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
4003 }
4004 
4005 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4006 			     struct prefix_info *pinfo, u32 pid, u32 seq,
4007 			     int event, unsigned int flags)
4008 {
4009 	struct prefixmsg *pmsg;
4010 	struct nlmsghdr *nlh;
4011 	struct prefix_cacheinfo	ci;
4012 
4013 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4014 	if (nlh == NULL)
4015 		return -EMSGSIZE;
4016 
4017 	pmsg = nlmsg_data(nlh);
4018 	pmsg->prefix_family = AF_INET6;
4019 	pmsg->prefix_pad1 = 0;
4020 	pmsg->prefix_pad2 = 0;
4021 	pmsg->prefix_ifindex = idev->dev->ifindex;
4022 	pmsg->prefix_len = pinfo->prefix_len;
4023 	pmsg->prefix_type = pinfo->type;
4024 	pmsg->prefix_pad3 = 0;
4025 	pmsg->prefix_flags = 0;
4026 	if (pinfo->onlink)
4027 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4028 	if (pinfo->autoconf)
4029 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4030 
4031 	NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4032 
4033 	ci.preferred_time = ntohl(pinfo->prefered);
4034 	ci.valid_time = ntohl(pinfo->valid);
4035 	NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4036 
4037 	return nlmsg_end(skb, nlh);
4038 
4039 nla_put_failure:
4040 	nlmsg_cancel(skb, nlh);
4041 	return -EMSGSIZE;
4042 }
4043 
4044 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4045 			 struct prefix_info *pinfo)
4046 {
4047 	struct sk_buff *skb;
4048 	struct net *net = dev_net(idev->dev);
4049 	int err = -ENOBUFS;
4050 
4051 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4052 	if (skb == NULL)
4053 		goto errout;
4054 
4055 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4056 	if (err < 0) {
4057 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4058 		WARN_ON(err == -EMSGSIZE);
4059 		kfree_skb(skb);
4060 		goto errout;
4061 	}
4062 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4063 	return;
4064 errout:
4065 	if (err < 0)
4066 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4067 }
4068 
4069 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4070 {
4071 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4072 
4073 	switch (event) {
4074 	case RTM_NEWADDR:
4075 		/*
4076 		 * If the address was optimistic
4077 		 * we inserted the route at the start of
4078 		 * our DAD process, so we don't need
4079 		 * to do it again
4080 		 */
4081 		if (!(ifp->rt->rt6i_node))
4082 			ip6_ins_rt(ifp->rt);
4083 		if (ifp->idev->cnf.forwarding)
4084 			addrconf_join_anycast(ifp);
4085 		break;
4086 	case RTM_DELADDR:
4087 		if (ifp->idev->cnf.forwarding)
4088 			addrconf_leave_anycast(ifp);
4089 		addrconf_leave_solict(ifp->idev, &ifp->addr);
4090 		dst_hold(&ifp->rt->u.dst);
4091 
4092 		if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4093 		    ip6_del_rt(ifp->rt))
4094 			dst_free(&ifp->rt->u.dst);
4095 		break;
4096 	}
4097 }
4098 
4099 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4100 {
4101 	rcu_read_lock_bh();
4102 	if (likely(ifp->idev->dead == 0))
4103 		__ipv6_ifa_notify(event, ifp);
4104 	rcu_read_unlock_bh();
4105 }
4106 
4107 #ifdef CONFIG_SYSCTL
4108 
4109 static
4110 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4111 			   void __user *buffer, size_t *lenp, loff_t *ppos)
4112 {
4113 	int *valp = ctl->data;
4114 	int val = *valp;
4115 	loff_t pos = *ppos;
4116 	int ret;
4117 
4118 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4119 
4120 	if (write)
4121 		ret = addrconf_fixup_forwarding(ctl, valp, val);
4122 	if (ret)
4123 		*ppos = pos;
4124 	return ret;
4125 }
4126 
4127 static void dev_disable_change(struct inet6_dev *idev)
4128 {
4129 	if (!idev || !idev->dev)
4130 		return;
4131 
4132 	if (idev->cnf.disable_ipv6)
4133 		addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4134 	else
4135 		addrconf_notify(NULL, NETDEV_UP, idev->dev);
4136 }
4137 
4138 static void addrconf_disable_change(struct net *net, __s32 newf)
4139 {
4140 	struct net_device *dev;
4141 	struct inet6_dev *idev;
4142 
4143 	rcu_read_lock();
4144 	for_each_netdev_rcu(net, dev) {
4145 		idev = __in6_dev_get(dev);
4146 		if (idev) {
4147 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4148 			idev->cnf.disable_ipv6 = newf;
4149 			if (changed)
4150 				dev_disable_change(idev);
4151 		}
4152 	}
4153 	rcu_read_unlock();
4154 }
4155 
4156 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4157 {
4158 	struct net *net;
4159 
4160 	net = (struct net *)table->extra2;
4161 
4162 	if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4163 		return 0;
4164 
4165 	if (!rtnl_trylock()) {
4166 		/* Restore the original values before restarting */
4167 		*p = old;
4168 		return restart_syscall();
4169 	}
4170 
4171 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
4172 		__s32 newf = net->ipv6.devconf_all->disable_ipv6;
4173 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
4174 		addrconf_disable_change(net, newf);
4175 	} else if ((!*p) ^ (!old))
4176 		dev_disable_change((struct inet6_dev *)table->extra1);
4177 
4178 	rtnl_unlock();
4179 	return 0;
4180 }
4181 
4182 static
4183 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4184 			    void __user *buffer, size_t *lenp, loff_t *ppos)
4185 {
4186 	int *valp = ctl->data;
4187 	int val = *valp;
4188 	loff_t pos = *ppos;
4189 	int ret;
4190 
4191 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4192 
4193 	if (write)
4194 		ret = addrconf_disable_ipv6(ctl, valp, val);
4195 	if (ret)
4196 		*ppos = pos;
4197 	return ret;
4198 }
4199 
4200 static struct addrconf_sysctl_table
4201 {
4202 	struct ctl_table_header *sysctl_header;
4203 	ctl_table addrconf_vars[DEVCONF_MAX+1];
4204 	char *dev_name;
4205 } addrconf_sysctl __read_mostly = {
4206 	.sysctl_header = NULL,
4207 	.addrconf_vars = {
4208 		{
4209 			.procname	= "forwarding",
4210 			.data		= &ipv6_devconf.forwarding,
4211 			.maxlen		= sizeof(int),
4212 			.mode		= 0644,
4213 			.proc_handler	= addrconf_sysctl_forward,
4214 		},
4215 		{
4216 			.procname	= "hop_limit",
4217 			.data		= &ipv6_devconf.hop_limit,
4218 			.maxlen		= sizeof(int),
4219 			.mode		= 0644,
4220 			.proc_handler	= proc_dointvec,
4221 		},
4222 		{
4223 			.procname	= "mtu",
4224 			.data		= &ipv6_devconf.mtu6,
4225 			.maxlen		= sizeof(int),
4226 			.mode		= 0644,
4227 			.proc_handler	= proc_dointvec,
4228 		},
4229 		{
4230 			.procname	= "accept_ra",
4231 			.data		= &ipv6_devconf.accept_ra,
4232 			.maxlen		= sizeof(int),
4233 			.mode		= 0644,
4234 			.proc_handler	= proc_dointvec,
4235 		},
4236 		{
4237 			.procname	= "accept_redirects",
4238 			.data		= &ipv6_devconf.accept_redirects,
4239 			.maxlen		= sizeof(int),
4240 			.mode		= 0644,
4241 			.proc_handler	= proc_dointvec,
4242 		},
4243 		{
4244 			.procname	= "autoconf",
4245 			.data		= &ipv6_devconf.autoconf,
4246 			.maxlen		= sizeof(int),
4247 			.mode		= 0644,
4248 			.proc_handler	= proc_dointvec,
4249 		},
4250 		{
4251 			.procname	= "dad_transmits",
4252 			.data		= &ipv6_devconf.dad_transmits,
4253 			.maxlen		= sizeof(int),
4254 			.mode		= 0644,
4255 			.proc_handler	= proc_dointvec,
4256 		},
4257 		{
4258 			.procname	= "router_solicitations",
4259 			.data		= &ipv6_devconf.rtr_solicits,
4260 			.maxlen		= sizeof(int),
4261 			.mode		= 0644,
4262 			.proc_handler	= proc_dointvec,
4263 		},
4264 		{
4265 			.procname	= "router_solicitation_interval",
4266 			.data		= &ipv6_devconf.rtr_solicit_interval,
4267 			.maxlen		= sizeof(int),
4268 			.mode		= 0644,
4269 			.proc_handler	= proc_dointvec_jiffies,
4270 		},
4271 		{
4272 			.procname	= "router_solicitation_delay",
4273 			.data		= &ipv6_devconf.rtr_solicit_delay,
4274 			.maxlen		= sizeof(int),
4275 			.mode		= 0644,
4276 			.proc_handler	= proc_dointvec_jiffies,
4277 		},
4278 		{
4279 			.procname	= "force_mld_version",
4280 			.data		= &ipv6_devconf.force_mld_version,
4281 			.maxlen		= sizeof(int),
4282 			.mode		= 0644,
4283 			.proc_handler	= proc_dointvec,
4284 		},
4285 #ifdef CONFIG_IPV6_PRIVACY
4286 		{
4287 			.procname	= "use_tempaddr",
4288 			.data		= &ipv6_devconf.use_tempaddr,
4289 			.maxlen		= sizeof(int),
4290 			.mode		= 0644,
4291 			.proc_handler	= proc_dointvec,
4292 		},
4293 		{
4294 			.procname	= "temp_valid_lft",
4295 			.data		= &ipv6_devconf.temp_valid_lft,
4296 			.maxlen		= sizeof(int),
4297 			.mode		= 0644,
4298 			.proc_handler	= proc_dointvec,
4299 		},
4300 		{
4301 			.procname	= "temp_prefered_lft",
4302 			.data		= &ipv6_devconf.temp_prefered_lft,
4303 			.maxlen		= sizeof(int),
4304 			.mode		= 0644,
4305 			.proc_handler	= proc_dointvec,
4306 		},
4307 		{
4308 			.procname	= "regen_max_retry",
4309 			.data		= &ipv6_devconf.regen_max_retry,
4310 			.maxlen		= sizeof(int),
4311 			.mode		= 0644,
4312 			.proc_handler	= proc_dointvec,
4313 		},
4314 		{
4315 			.procname	= "max_desync_factor",
4316 			.data		= &ipv6_devconf.max_desync_factor,
4317 			.maxlen		= sizeof(int),
4318 			.mode		= 0644,
4319 			.proc_handler	= proc_dointvec,
4320 		},
4321 #endif
4322 		{
4323 			.procname	= "max_addresses",
4324 			.data		= &ipv6_devconf.max_addresses,
4325 			.maxlen		= sizeof(int),
4326 			.mode		= 0644,
4327 			.proc_handler	= proc_dointvec,
4328 		},
4329 		{
4330 			.procname	= "accept_ra_defrtr",
4331 			.data		= &ipv6_devconf.accept_ra_defrtr,
4332 			.maxlen		= sizeof(int),
4333 			.mode		= 0644,
4334 			.proc_handler	= proc_dointvec,
4335 		},
4336 		{
4337 			.procname	= "accept_ra_pinfo",
4338 			.data		= &ipv6_devconf.accept_ra_pinfo,
4339 			.maxlen		= sizeof(int),
4340 			.mode		= 0644,
4341 			.proc_handler	= proc_dointvec,
4342 		},
4343 #ifdef CONFIG_IPV6_ROUTER_PREF
4344 		{
4345 			.procname	= "accept_ra_rtr_pref",
4346 			.data		= &ipv6_devconf.accept_ra_rtr_pref,
4347 			.maxlen		= sizeof(int),
4348 			.mode		= 0644,
4349 			.proc_handler	= proc_dointvec,
4350 		},
4351 		{
4352 			.procname	= "router_probe_interval",
4353 			.data		= &ipv6_devconf.rtr_probe_interval,
4354 			.maxlen		= sizeof(int),
4355 			.mode		= 0644,
4356 			.proc_handler	= proc_dointvec_jiffies,
4357 		},
4358 #ifdef CONFIG_IPV6_ROUTE_INFO
4359 		{
4360 			.procname	= "accept_ra_rt_info_max_plen",
4361 			.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
4362 			.maxlen		= sizeof(int),
4363 			.mode		= 0644,
4364 			.proc_handler	= proc_dointvec,
4365 		},
4366 #endif
4367 #endif
4368 		{
4369 			.procname	= "proxy_ndp",
4370 			.data		= &ipv6_devconf.proxy_ndp,
4371 			.maxlen		= sizeof(int),
4372 			.mode		= 0644,
4373 			.proc_handler	= proc_dointvec,
4374 		},
4375 		{
4376 			.procname	= "accept_source_route",
4377 			.data		= &ipv6_devconf.accept_source_route,
4378 			.maxlen		= sizeof(int),
4379 			.mode		= 0644,
4380 			.proc_handler	= proc_dointvec,
4381 		},
4382 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4383 		{
4384 			.procname       = "optimistic_dad",
4385 			.data           = &ipv6_devconf.optimistic_dad,
4386 			.maxlen         = sizeof(int),
4387 			.mode           = 0644,
4388 			.proc_handler   = proc_dointvec,
4389 
4390 		},
4391 #endif
4392 #ifdef CONFIG_IPV6_MROUTE
4393 		{
4394 			.procname	= "mc_forwarding",
4395 			.data		= &ipv6_devconf.mc_forwarding,
4396 			.maxlen		= sizeof(int),
4397 			.mode		= 0444,
4398 			.proc_handler	= proc_dointvec,
4399 		},
4400 #endif
4401 		{
4402 			.procname	= "disable_ipv6",
4403 			.data		= &ipv6_devconf.disable_ipv6,
4404 			.maxlen		= sizeof(int),
4405 			.mode		= 0644,
4406 			.proc_handler	= addrconf_sysctl_disable,
4407 		},
4408 		{
4409 			.procname	= "accept_dad",
4410 			.data		= &ipv6_devconf.accept_dad,
4411 			.maxlen		= sizeof(int),
4412 			.mode		= 0644,
4413 			.proc_handler	= proc_dointvec,
4414 		},
4415 		{
4416 			.procname       = "force_tllao",
4417 			.data           = &ipv6_devconf.force_tllao,
4418 			.maxlen         = sizeof(int),
4419 			.mode           = 0644,
4420 			.proc_handler   = proc_dointvec
4421 		},
4422 		{
4423 			/* sentinel */
4424 		}
4425 	},
4426 };
4427 
4428 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4429 		struct inet6_dev *idev, struct ipv6_devconf *p)
4430 {
4431 	int i;
4432 	struct addrconf_sysctl_table *t;
4433 
4434 #define ADDRCONF_CTL_PATH_DEV	3
4435 
4436 	struct ctl_path addrconf_ctl_path[] = {
4437 		{ .procname = "net", },
4438 		{ .procname = "ipv6", },
4439 		{ .procname = "conf", },
4440 		{ /* to be set */ },
4441 		{ },
4442 	};
4443 
4444 
4445 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4446 	if (t == NULL)
4447 		goto out;
4448 
4449 	for (i = 0; t->addrconf_vars[i].data; i++) {
4450 		t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4451 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4452 		t->addrconf_vars[i].extra2 = net;
4453 	}
4454 
4455 	/*
4456 	 * Make a copy of dev_name, because '.procname' is regarded as const
4457 	 * by sysctl and we wouldn't want anyone to change it under our feet
4458 	 * (see SIOCSIFNAME).
4459 	 */
4460 	t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4461 	if (!t->dev_name)
4462 		goto free;
4463 
4464 	addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4465 
4466 	t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4467 			t->addrconf_vars);
4468 	if (t->sysctl_header == NULL)
4469 		goto free_procname;
4470 
4471 	p->sysctl = t;
4472 	return 0;
4473 
4474 free_procname:
4475 	kfree(t->dev_name);
4476 free:
4477 	kfree(t);
4478 out:
4479 	return -ENOBUFS;
4480 }
4481 
4482 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4483 {
4484 	struct addrconf_sysctl_table *t;
4485 
4486 	if (p->sysctl == NULL)
4487 		return;
4488 
4489 	t = p->sysctl;
4490 	p->sysctl = NULL;
4491 	unregister_sysctl_table(t->sysctl_header);
4492 	kfree(t->dev_name);
4493 	kfree(t);
4494 }
4495 
4496 static void addrconf_sysctl_register(struct inet6_dev *idev)
4497 {
4498 	neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4499 			      &ndisc_ifinfo_sysctl_change);
4500 	__addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4501 					idev, &idev->cnf);
4502 }
4503 
4504 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4505 {
4506 	__addrconf_sysctl_unregister(&idev->cnf);
4507 	neigh_sysctl_unregister(idev->nd_parms);
4508 }
4509 
4510 
4511 #endif
4512 
4513 static int __net_init addrconf_init_net(struct net *net)
4514 {
4515 	int err;
4516 	struct ipv6_devconf *all, *dflt;
4517 
4518 	err = -ENOMEM;
4519 	all = &ipv6_devconf;
4520 	dflt = &ipv6_devconf_dflt;
4521 
4522 	if (!net_eq(net, &init_net)) {
4523 		all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4524 		if (all == NULL)
4525 			goto err_alloc_all;
4526 
4527 		dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4528 		if (dflt == NULL)
4529 			goto err_alloc_dflt;
4530 	} else {
4531 		/* these will be inherited by all namespaces */
4532 		dflt->autoconf = ipv6_defaults.autoconf;
4533 		dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4534 	}
4535 
4536 	net->ipv6.devconf_all = all;
4537 	net->ipv6.devconf_dflt = dflt;
4538 
4539 #ifdef CONFIG_SYSCTL
4540 	err = __addrconf_sysctl_register(net, "all", NULL, all);
4541 	if (err < 0)
4542 		goto err_reg_all;
4543 
4544 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4545 	if (err < 0)
4546 		goto err_reg_dflt;
4547 #endif
4548 	return 0;
4549 
4550 #ifdef CONFIG_SYSCTL
4551 err_reg_dflt:
4552 	__addrconf_sysctl_unregister(all);
4553 err_reg_all:
4554 	kfree(dflt);
4555 #endif
4556 err_alloc_dflt:
4557 	kfree(all);
4558 err_alloc_all:
4559 	return err;
4560 }
4561 
4562 static void __net_exit addrconf_exit_net(struct net *net)
4563 {
4564 #ifdef CONFIG_SYSCTL
4565 	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4566 	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
4567 #endif
4568 	if (!net_eq(net, &init_net)) {
4569 		kfree(net->ipv6.devconf_dflt);
4570 		kfree(net->ipv6.devconf_all);
4571 	}
4572 }
4573 
4574 static struct pernet_operations addrconf_ops = {
4575 	.init = addrconf_init_net,
4576 	.exit = addrconf_exit_net,
4577 };
4578 
4579 /*
4580  *      Device notifier
4581  */
4582 
4583 int register_inet6addr_notifier(struct notifier_block *nb)
4584 {
4585 	return atomic_notifier_chain_register(&inet6addr_chain, nb);
4586 }
4587 EXPORT_SYMBOL(register_inet6addr_notifier);
4588 
4589 int unregister_inet6addr_notifier(struct notifier_block *nb)
4590 {
4591 	return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4592 }
4593 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4594 
4595 /*
4596  *	Init / cleanup code
4597  */
4598 
4599 int __init addrconf_init(void)
4600 {
4601 	int i, err;
4602 
4603 	err = ipv6_addr_label_init();
4604 	if (err < 0) {
4605 		printk(KERN_CRIT "IPv6 Addrconf:"
4606 		       " cannot initialize default policy table: %d.\n", err);
4607 		return err;
4608 	}
4609 
4610 	register_pernet_subsys(&addrconf_ops);
4611 
4612 	/* The addrconf netdev notifier requires that loopback_dev
4613 	 * has it's ipv6 private information allocated and setup
4614 	 * before it can bring up and give link-local addresses
4615 	 * to other devices which are up.
4616 	 *
4617 	 * Unfortunately, loopback_dev is not necessarily the first
4618 	 * entry in the global dev_base list of net devices.  In fact,
4619 	 * it is likely to be the very last entry on that list.
4620 	 * So this causes the notifier registry below to try and
4621 	 * give link-local addresses to all devices besides loopback_dev
4622 	 * first, then loopback_dev, which cases all the non-loopback_dev
4623 	 * devices to fail to get a link-local address.
4624 	 *
4625 	 * So, as a temporary fix, allocate the ipv6 structure for
4626 	 * loopback_dev first by hand.
4627 	 * Longer term, all of the dependencies ipv6 has upon the loopback
4628 	 * device and it being up should be removed.
4629 	 */
4630 	rtnl_lock();
4631 	if (!ipv6_add_dev(init_net.loopback_dev))
4632 		err = -ENOMEM;
4633 	rtnl_unlock();
4634 	if (err)
4635 		goto errlo;
4636 
4637 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4638 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4639 
4640 	register_netdevice_notifier(&ipv6_dev_notf);
4641 
4642 	addrconf_verify(0);
4643 
4644 	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4645 	if (err < 0)
4646 		goto errout;
4647 
4648 	/* Only the first call to __rtnl_register can fail */
4649 	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4650 	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4651 	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4652 	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4653 	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4654 
4655 	ipv6_addr_label_rtnl_register();
4656 
4657 	return 0;
4658 errout:
4659 	unregister_netdevice_notifier(&ipv6_dev_notf);
4660 errlo:
4661 	unregister_pernet_subsys(&addrconf_ops);
4662 
4663 	return err;
4664 }
4665 
4666 void addrconf_cleanup(void)
4667 {
4668 	struct net_device *dev;
4669 	int i;
4670 
4671 	unregister_netdevice_notifier(&ipv6_dev_notf);
4672 	unregister_pernet_subsys(&addrconf_ops);
4673 
4674 	rtnl_lock();
4675 
4676 	/* clean dev list */
4677 	for_each_netdev(&init_net, dev) {
4678 		if (__in6_dev_get(dev) == NULL)
4679 			continue;
4680 		addrconf_ifdown(dev, 1);
4681 	}
4682 	addrconf_ifdown(init_net.loopback_dev, 2);
4683 
4684 	/*
4685 	 *	Check hash table.
4686 	 */
4687 	spin_lock_bh(&addrconf_hash_lock);
4688 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4689 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4690 	spin_unlock_bh(&addrconf_hash_lock);
4691 
4692 	del_timer(&addr_chk_timer);
4693 	rtnl_unlock();
4694 }
4695