xref: /openbmc/linux/net/ipv6/addrconf.c (revision 853dc2e03ddd0c49885ed55c48755d2b1073122a)
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 #endif
125 
126 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
127 static int ipv6_count_addresses(struct inet6_dev *idev);
128 
129 /*
130  *	Configured unicast address hash table
131  */
132 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
133 static DEFINE_SPINLOCK(addrconf_hash_lock);
134 
135 static void addrconf_verify(unsigned long);
136 
137 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
138 static DEFINE_SPINLOCK(addrconf_verify_lock);
139 
140 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
141 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
142 
143 static void addrconf_type_change(struct net_device *dev,
144 				 unsigned long event);
145 static int addrconf_ifdown(struct net_device *dev, int how);
146 
147 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
148 static void addrconf_dad_timer(unsigned long data);
149 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
150 static void addrconf_dad_run(struct inet6_dev *idev);
151 static void addrconf_rs_timer(unsigned long data);
152 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
153 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
154 
155 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
156 				struct prefix_info *pinfo);
157 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
158 			       struct net_device *dev);
159 
160 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
161 
162 static struct ipv6_devconf ipv6_devconf __read_mostly = {
163 	.forwarding		= 0,
164 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
165 	.mtu6			= IPV6_MIN_MTU,
166 	.accept_ra		= 1,
167 	.accept_redirects	= 1,
168 	.autoconf		= 1,
169 	.force_mld_version	= 0,
170 	.dad_transmits		= 1,
171 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
172 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
173 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
174 #ifdef CONFIG_IPV6_PRIVACY
175 	.use_tempaddr 		= 0,
176 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
177 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
178 	.regen_max_retry	= REGEN_MAX_RETRY,
179 	.max_desync_factor	= MAX_DESYNC_FACTOR,
180 #endif
181 	.max_addresses		= IPV6_MAX_ADDRESSES,
182 	.accept_ra_defrtr	= 1,
183 	.accept_ra_pinfo	= 1,
184 #ifdef CONFIG_IPV6_ROUTER_PREF
185 	.accept_ra_rtr_pref	= 1,
186 	.rtr_probe_interval	= 60 * HZ,
187 #ifdef CONFIG_IPV6_ROUTE_INFO
188 	.accept_ra_rt_info_max_plen = 0,
189 #endif
190 #endif
191 	.proxy_ndp		= 0,
192 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
193 	.disable_ipv6		= 0,
194 	.accept_dad		= 1,
195 };
196 
197 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
198 	.forwarding		= 0,
199 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
200 	.mtu6			= IPV6_MIN_MTU,
201 	.accept_ra		= 1,
202 	.accept_redirects	= 1,
203 	.autoconf		= 1,
204 	.dad_transmits		= 1,
205 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
206 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
207 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
208 #ifdef CONFIG_IPV6_PRIVACY
209 	.use_tempaddr		= 0,
210 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
211 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
212 	.regen_max_retry	= REGEN_MAX_RETRY,
213 	.max_desync_factor	= MAX_DESYNC_FACTOR,
214 #endif
215 	.max_addresses		= IPV6_MAX_ADDRESSES,
216 	.accept_ra_defrtr	= 1,
217 	.accept_ra_pinfo	= 1,
218 #ifdef CONFIG_IPV6_ROUTER_PREF
219 	.accept_ra_rtr_pref	= 1,
220 	.rtr_probe_interval	= 60 * HZ,
221 #ifdef CONFIG_IPV6_ROUTE_INFO
222 	.accept_ra_rt_info_max_plen = 0,
223 #endif
224 #endif
225 	.proxy_ndp		= 0,
226 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
227 	.disable_ipv6		= 0,
228 	.accept_dad		= 1,
229 };
230 
231 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
232 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
233 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
234 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
235 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
236 
237 /* Check if a valid qdisc is available */
238 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
239 {
240 	return !qdisc_tx_is_noop(dev);
241 }
242 
243 /* Check if a route is valid prefix route */
244 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
245 {
246 	return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
247 }
248 
249 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
250 {
251 	if (del_timer(&ifp->timer))
252 		__in6_ifa_put(ifp);
253 }
254 
255 enum addrconf_timer_t {
256 	AC_NONE,
257 	AC_DAD,
258 	AC_RS,
259 };
260 
261 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
262 			       enum addrconf_timer_t what,
263 			       unsigned long when)
264 {
265 	if (!del_timer(&ifp->timer))
266 		in6_ifa_hold(ifp);
267 
268 	switch (what) {
269 	case AC_DAD:
270 		ifp->timer.function = addrconf_dad_timer;
271 		break;
272 	case AC_RS:
273 		ifp->timer.function = addrconf_rs_timer;
274 		break;
275 	default:
276 		break;
277 	}
278 	ifp->timer.expires = jiffies + when;
279 	add_timer(&ifp->timer);
280 }
281 
282 static int snmp6_alloc_dev(struct inet6_dev *idev)
283 {
284 	if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
285 			  sizeof(struct ipstats_mib),
286 			  __alignof__(struct ipstats_mib)) < 0)
287 		goto err_ip;
288 	if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
289 			  sizeof(struct icmpv6_mib),
290 			  __alignof__(struct icmpv6_mib)) < 0)
291 		goto err_icmp;
292 	if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
293 			  sizeof(struct icmpv6msg_mib),
294 			  __alignof__(struct icmpv6msg_mib)) < 0)
295 		goto err_icmpmsg;
296 
297 	return 0;
298 
299 err_icmpmsg:
300 	snmp_mib_free((void __percpu **)idev->stats.icmpv6);
301 err_icmp:
302 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
303 err_ip:
304 	return -ENOMEM;
305 }
306 
307 static void snmp6_free_dev(struct inet6_dev *idev)
308 {
309 	snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
310 	snmp_mib_free((void __percpu **)idev->stats.icmpv6);
311 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
312 }
313 
314 /* Nobody refers to this device, we may destroy it. */
315 
316 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
317 {
318 	struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
319 	kfree(idev);
320 }
321 
322 void in6_dev_finish_destroy(struct inet6_dev *idev)
323 {
324 	struct net_device *dev = idev->dev;
325 
326 	WARN_ON(!list_empty(&idev->addr_list));
327 	WARN_ON(idev->mc_list != NULL);
328 
329 #ifdef NET_REFCNT_DEBUG
330 	printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
331 #endif
332 	dev_put(dev);
333 	if (!idev->dead) {
334 		pr_warning("Freeing alive inet6 device %p\n", idev);
335 		return;
336 	}
337 	snmp6_free_dev(idev);
338 	call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
339 }
340 
341 EXPORT_SYMBOL(in6_dev_finish_destroy);
342 
343 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 {
345 	struct inet6_dev *ndev;
346 
347 	ASSERT_RTNL();
348 
349 	if (dev->mtu < IPV6_MIN_MTU)
350 		return NULL;
351 
352 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
353 
354 	if (ndev == NULL)
355 		return NULL;
356 
357 	rwlock_init(&ndev->lock);
358 	ndev->dev = dev;
359 	INIT_LIST_HEAD(&ndev->addr_list);
360 
361 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
362 	ndev->cnf.mtu6 = dev->mtu;
363 	ndev->cnf.sysctl = NULL;
364 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
365 	if (ndev->nd_parms == NULL) {
366 		kfree(ndev);
367 		return NULL;
368 	}
369 	if (ndev->cnf.forwarding)
370 		dev_disable_lro(dev);
371 	/* We refer to the device */
372 	dev_hold(dev);
373 
374 	if (snmp6_alloc_dev(ndev) < 0) {
375 		ADBG((KERN_WARNING
376 			"%s(): cannot allocate memory for statistics; dev=%s.\n",
377 			__func__, dev->name));
378 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
379 		ndev->dead = 1;
380 		in6_dev_finish_destroy(ndev);
381 		return NULL;
382 	}
383 
384 	if (snmp6_register_dev(ndev) < 0) {
385 		ADBG((KERN_WARNING
386 			"%s(): cannot create /proc/net/dev_snmp6/%s\n",
387 			__func__, dev->name));
388 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
389 		ndev->dead = 1;
390 		in6_dev_finish_destroy(ndev);
391 		return NULL;
392 	}
393 
394 	/* One reference from device.  We must do this before
395 	 * we invoke __ipv6_regen_rndid().
396 	 */
397 	in6_dev_hold(ndev);
398 
399 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
400 		ndev->cnf.accept_dad = -1;
401 
402 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
403 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
404 		printk(KERN_INFO
405 		       "%s: Disabled Multicast RS\n",
406 		       dev->name);
407 		ndev->cnf.rtr_solicits = 0;
408 	}
409 #endif
410 
411 #ifdef CONFIG_IPV6_PRIVACY
412 	INIT_LIST_HEAD(&ndev->tempaddr_list);
413 	setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
414 	if ((dev->flags&IFF_LOOPBACK) ||
415 	    dev->type == ARPHRD_TUNNEL ||
416 	    dev->type == ARPHRD_TUNNEL6 ||
417 	    dev->type == ARPHRD_SIT ||
418 	    dev->type == ARPHRD_NONE) {
419 		printk(KERN_INFO
420 		       "%s: Disabled Privacy Extensions\n",
421 		       dev->name);
422 		ndev->cnf.use_tempaddr = -1;
423 	} else {
424 		in6_dev_hold(ndev);
425 		ipv6_regen_rndid((unsigned long) ndev);
426 	}
427 #endif
428 
429 	if (netif_running(dev) && addrconf_qdisc_ok(dev))
430 		ndev->if_flags |= IF_READY;
431 
432 	ipv6_mc_init_dev(ndev);
433 	ndev->tstamp = jiffies;
434 	addrconf_sysctl_register(ndev);
435 	/* protected by rtnl_lock */
436 	rcu_assign_pointer(dev->ip6_ptr, ndev);
437 
438 	/* Join all-node multicast group */
439 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
440 
441 	return ndev;
442 }
443 
444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
445 {
446 	struct inet6_dev *idev;
447 
448 	ASSERT_RTNL();
449 
450 	idev = __in6_dev_get(dev);
451 	if (!idev) {
452 		idev = ipv6_add_dev(dev);
453 		if (!idev)
454 			return NULL;
455 	}
456 
457 	if (dev->flags&IFF_UP)
458 		ipv6_mc_up(idev);
459 	return idev;
460 }
461 
462 #ifdef CONFIG_SYSCTL
463 static void dev_forward_change(struct inet6_dev *idev)
464 {
465 	struct net_device *dev;
466 	struct inet6_ifaddr *ifa;
467 
468 	if (!idev)
469 		return;
470 	dev = idev->dev;
471 	if (idev->cnf.forwarding)
472 		dev_disable_lro(dev);
473 	if (dev && (dev->flags & IFF_MULTICAST)) {
474 		if (idev->cnf.forwarding)
475 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
476 		else
477 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
478 	}
479 
480 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
481 		if (ifa->flags&IFA_F_TENTATIVE)
482 			continue;
483 		if (idev->cnf.forwarding)
484 			addrconf_join_anycast(ifa);
485 		else
486 			addrconf_leave_anycast(ifa);
487 	}
488 }
489 
490 
491 static void addrconf_forward_change(struct net *net, __s32 newf)
492 {
493 	struct net_device *dev;
494 	struct inet6_dev *idev;
495 
496 	rcu_read_lock();
497 	for_each_netdev_rcu(net, dev) {
498 		idev = __in6_dev_get(dev);
499 		if (idev) {
500 			int changed = (!idev->cnf.forwarding) ^ (!newf);
501 			idev->cnf.forwarding = newf;
502 			if (changed)
503 				dev_forward_change(idev);
504 		}
505 	}
506 	rcu_read_unlock();
507 }
508 
509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
510 {
511 	struct net *net;
512 
513 	net = (struct net *)table->extra2;
514 	if (p == &net->ipv6.devconf_dflt->forwarding)
515 		return 0;
516 
517 	if (!rtnl_trylock()) {
518 		/* Restore the original values before restarting */
519 		*p = old;
520 		return restart_syscall();
521 	}
522 
523 	if (p == &net->ipv6.devconf_all->forwarding) {
524 		__s32 newf = net->ipv6.devconf_all->forwarding;
525 		net->ipv6.devconf_dflt->forwarding = newf;
526 		addrconf_forward_change(net, newf);
527 	} else if ((!*p) ^ (!old))
528 		dev_forward_change((struct inet6_dev *)table->extra1);
529 	rtnl_unlock();
530 
531 	if (*p)
532 		rt6_purge_dflt_routers(net);
533 	return 1;
534 }
535 #endif
536 
537 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
538 {
539 	struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
540 	kfree(ifp);
541 }
542 
543 /* Nobody refers to this ifaddr, destroy it */
544 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
545 {
546 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
547 
548 #ifdef NET_REFCNT_DEBUG
549 	printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
550 #endif
551 
552 	in6_dev_put(ifp->idev);
553 
554 	if (del_timer(&ifp->timer))
555 		pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
556 
557 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
558 		pr_warning("Freeing alive inet6 address %p\n", ifp);
559 		return;
560 	}
561 	dst_release(&ifp->rt->dst);
562 
563 	call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
564 }
565 
566 static void
567 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
568 {
569 	struct list_head *p;
570 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
571 
572 	/*
573 	 * Each device address list is sorted in order of scope -
574 	 * global before linklocal.
575 	 */
576 	list_for_each(p, &idev->addr_list) {
577 		struct inet6_ifaddr *ifa
578 			= list_entry(p, struct inet6_ifaddr, if_list);
579 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
580 			break;
581 	}
582 
583 	list_add_tail(&ifp->if_list, p);
584 }
585 
586 static u32 ipv6_addr_hash(const struct in6_addr *addr)
587 {
588 	/*
589 	 * We perform the hash function over the last 64 bits of the address
590 	 * This will include the IEEE address token on links that support it.
591 	 */
592 	return jhash_2words((__force u32)addr->s6_addr32[2],
593 			    (__force u32)addr->s6_addr32[3], 0)
594 		& (IN6_ADDR_HSIZE - 1);
595 }
596 
597 /* On success it returns ifp with increased reference count */
598 
599 static struct inet6_ifaddr *
600 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
601 	      int scope, u32 flags)
602 {
603 	struct inet6_ifaddr *ifa = NULL;
604 	struct rt6_info *rt;
605 	unsigned int hash;
606 	int err = 0;
607 	int addr_type = ipv6_addr_type(addr);
608 
609 	if (addr_type == IPV6_ADDR_ANY ||
610 	    addr_type & IPV6_ADDR_MULTICAST ||
611 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
612 	     addr_type & IPV6_ADDR_LOOPBACK))
613 		return ERR_PTR(-EADDRNOTAVAIL);
614 
615 	rcu_read_lock_bh();
616 	if (idev->dead) {
617 		err = -ENODEV;			/*XXX*/
618 		goto out2;
619 	}
620 
621 	if (idev->cnf.disable_ipv6) {
622 		err = -EACCES;
623 		goto out2;
624 	}
625 
626 	spin_lock(&addrconf_hash_lock);
627 
628 	/* Ignore adding duplicate addresses on an interface */
629 	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
630 		ADBG(("ipv6_add_addr: already assigned\n"));
631 		err = -EEXIST;
632 		goto out;
633 	}
634 
635 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
636 
637 	if (ifa == NULL) {
638 		ADBG(("ipv6_add_addr: malloc failed\n"));
639 		err = -ENOBUFS;
640 		goto out;
641 	}
642 
643 	rt = addrconf_dst_alloc(idev, addr, 0);
644 	if (IS_ERR(rt)) {
645 		err = PTR_ERR(rt);
646 		goto out;
647 	}
648 
649 	ipv6_addr_copy(&ifa->addr, addr);
650 
651 	spin_lock_init(&ifa->lock);
652 	spin_lock_init(&ifa->state_lock);
653 	init_timer(&ifa->timer);
654 	INIT_HLIST_NODE(&ifa->addr_lst);
655 	ifa->timer.data = (unsigned long) ifa;
656 	ifa->scope = scope;
657 	ifa->prefix_len = pfxlen;
658 	ifa->flags = flags | IFA_F_TENTATIVE;
659 	ifa->cstamp = ifa->tstamp = jiffies;
660 
661 	ifa->rt = rt;
662 
663 	/*
664 	 * part one of RFC 4429, section 3.3
665 	 * We should not configure an address as
666 	 * optimistic if we do not yet know the link
667 	 * layer address of our nexhop router
668 	 */
669 
670 	if (rt->rt6i_nexthop == NULL)
671 		ifa->flags &= ~IFA_F_OPTIMISTIC;
672 
673 	ifa->idev = idev;
674 	in6_dev_hold(idev);
675 	/* For caller */
676 	in6_ifa_hold(ifa);
677 
678 	/* Add to big hash table */
679 	hash = ipv6_addr_hash(addr);
680 
681 	hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
682 	spin_unlock(&addrconf_hash_lock);
683 
684 	write_lock(&idev->lock);
685 	/* Add to inet6_dev unicast addr list. */
686 	ipv6_link_dev_addr(idev, ifa);
687 
688 #ifdef CONFIG_IPV6_PRIVACY
689 	if (ifa->flags&IFA_F_TEMPORARY) {
690 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
691 		in6_ifa_hold(ifa);
692 	}
693 #endif
694 
695 	in6_ifa_hold(ifa);
696 	write_unlock(&idev->lock);
697 out2:
698 	rcu_read_unlock_bh();
699 
700 	if (likely(err == 0))
701 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
702 	else {
703 		kfree(ifa);
704 		ifa = ERR_PTR(err);
705 	}
706 
707 	return ifa;
708 out:
709 	spin_unlock(&addrconf_hash_lock);
710 	goto out2;
711 }
712 
713 /* This function wants to get referenced ifp and releases it before return */
714 
715 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
716 {
717 	struct inet6_ifaddr *ifa, *ifn;
718 	struct inet6_dev *idev = ifp->idev;
719 	int state;
720 	int hash;
721 	int deleted = 0, onlink = 0;
722 	unsigned long expires = jiffies;
723 
724 	hash = ipv6_addr_hash(&ifp->addr);
725 
726 	spin_lock_bh(&ifp->state_lock);
727 	state = ifp->state;
728 	ifp->state = INET6_IFADDR_STATE_DEAD;
729 	spin_unlock_bh(&ifp->state_lock);
730 
731 	if (state == INET6_IFADDR_STATE_DEAD)
732 		goto out;
733 
734 	spin_lock_bh(&addrconf_hash_lock);
735 	hlist_del_init_rcu(&ifp->addr_lst);
736 	spin_unlock_bh(&addrconf_hash_lock);
737 
738 	write_lock_bh(&idev->lock);
739 #ifdef CONFIG_IPV6_PRIVACY
740 	if (ifp->flags&IFA_F_TEMPORARY) {
741 		list_del(&ifp->tmp_list);
742 		if (ifp->ifpub) {
743 			in6_ifa_put(ifp->ifpub);
744 			ifp->ifpub = NULL;
745 		}
746 		__in6_ifa_put(ifp);
747 	}
748 #endif
749 
750 	list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
751 		if (ifa == ifp) {
752 			list_del_init(&ifp->if_list);
753 			__in6_ifa_put(ifp);
754 
755 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
756 				break;
757 			deleted = 1;
758 			continue;
759 		} else if (ifp->flags & IFA_F_PERMANENT) {
760 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
761 					      ifp->prefix_len)) {
762 				if (ifa->flags & IFA_F_PERMANENT) {
763 					onlink = 1;
764 					if (deleted)
765 						break;
766 				} else {
767 					unsigned long lifetime;
768 
769 					if (!onlink)
770 						onlink = -1;
771 
772 					spin_lock(&ifa->lock);
773 
774 					lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
775 					/*
776 					 * Note: Because this address is
777 					 * not permanent, lifetime <
778 					 * LONG_MAX / HZ here.
779 					 */
780 					if (time_before(expires,
781 							ifa->tstamp + lifetime * HZ))
782 						expires = ifa->tstamp + lifetime * HZ;
783 					spin_unlock(&ifa->lock);
784 				}
785 			}
786 		}
787 	}
788 	write_unlock_bh(&idev->lock);
789 
790 	addrconf_del_timer(ifp);
791 
792 	ipv6_ifa_notify(RTM_DELADDR, ifp);
793 
794 	atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
795 
796 	/*
797 	 * Purge or update corresponding prefix
798 	 *
799 	 * 1) we don't purge prefix here if address was not permanent.
800 	 *    prefix is managed by its own lifetime.
801 	 * 2) if there're no addresses, delete prefix.
802 	 * 3) if there're still other permanent address(es),
803 	 *    corresponding prefix is still permanent.
804 	 * 4) otherwise, update prefix lifetime to the
805 	 *    longest valid lifetime among the corresponding
806 	 *    addresses on the device.
807 	 *    Note: subsequent RA will update lifetime.
808 	 *
809 	 * --yoshfuji
810 	 */
811 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
812 		struct in6_addr prefix;
813 		struct rt6_info *rt;
814 		struct net *net = dev_net(ifp->idev->dev);
815 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
816 		rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
817 
818 		if (rt && addrconf_is_prefix_route(rt)) {
819 			if (onlink == 0) {
820 				ip6_del_rt(rt);
821 				rt = NULL;
822 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
823 				rt->rt6i_expires = expires;
824 				rt->rt6i_flags |= RTF_EXPIRES;
825 			}
826 		}
827 		dst_release(&rt->dst);
828 	}
829 
830 out:
831 	in6_ifa_put(ifp);
832 }
833 
834 #ifdef CONFIG_IPV6_PRIVACY
835 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
836 {
837 	struct inet6_dev *idev = ifp->idev;
838 	struct in6_addr addr, *tmpaddr;
839 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
840 	unsigned long regen_advance;
841 	int tmp_plen;
842 	int ret = 0;
843 	int max_addresses;
844 	u32 addr_flags;
845 
846 	write_lock(&idev->lock);
847 	if (ift) {
848 		spin_lock_bh(&ift->lock);
849 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
850 		spin_unlock_bh(&ift->lock);
851 		tmpaddr = &addr;
852 	} else {
853 		tmpaddr = NULL;
854 	}
855 retry:
856 	in6_dev_hold(idev);
857 	if (idev->cnf.use_tempaddr <= 0) {
858 		write_unlock(&idev->lock);
859 		printk(KERN_INFO
860 			"ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
861 		in6_dev_put(idev);
862 		ret = -1;
863 		goto out;
864 	}
865 	spin_lock_bh(&ifp->lock);
866 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
867 		idev->cnf.use_tempaddr = -1;	/*XXX*/
868 		spin_unlock_bh(&ifp->lock);
869 		write_unlock(&idev->lock);
870 		printk(KERN_WARNING
871 			"ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
872 		in6_dev_put(idev);
873 		ret = -1;
874 		goto out;
875 	}
876 	in6_ifa_hold(ifp);
877 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
878 	if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
879 		spin_unlock_bh(&ifp->lock);
880 		write_unlock(&idev->lock);
881 		printk(KERN_WARNING
882 			"ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
883 		in6_ifa_put(ifp);
884 		in6_dev_put(idev);
885 		ret = -1;
886 		goto out;
887 	}
888 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
889 	age = (jiffies - ifp->tstamp) / HZ;
890 	tmp_valid_lft = min_t(__u32,
891 			      ifp->valid_lft,
892 			      idev->cnf.temp_valid_lft + age);
893 	tmp_prefered_lft = min_t(__u32,
894 				 ifp->prefered_lft,
895 				 idev->cnf.temp_prefered_lft + age -
896 				 idev->cnf.max_desync_factor);
897 	tmp_plen = ifp->prefix_len;
898 	max_addresses = idev->cnf.max_addresses;
899 	tmp_cstamp = ifp->cstamp;
900 	tmp_tstamp = ifp->tstamp;
901 	spin_unlock_bh(&ifp->lock);
902 
903 	regen_advance = idev->cnf.regen_max_retry *
904 	                idev->cnf.dad_transmits *
905 	                idev->nd_parms->retrans_time / HZ;
906 	write_unlock(&idev->lock);
907 
908 	/* A temporary address is created only if this calculated Preferred
909 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
910 	 * an implementation must not create a temporary address with a zero
911 	 * Preferred Lifetime.
912 	 */
913 	if (tmp_prefered_lft <= regen_advance) {
914 		in6_ifa_put(ifp);
915 		in6_dev_put(idev);
916 		ret = -1;
917 		goto out;
918 	}
919 
920 	addr_flags = IFA_F_TEMPORARY;
921 	/* set in addrconf_prefix_rcv() */
922 	if (ifp->flags & IFA_F_OPTIMISTIC)
923 		addr_flags |= IFA_F_OPTIMISTIC;
924 
925 	ift = !max_addresses ||
926 	      ipv6_count_addresses(idev) < max_addresses ?
927 		ipv6_add_addr(idev, &addr, tmp_plen,
928 			      ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
929 			      addr_flags) : NULL;
930 	if (!ift || IS_ERR(ift)) {
931 		in6_ifa_put(ifp);
932 		in6_dev_put(idev);
933 		printk(KERN_INFO
934 			"ipv6_create_tempaddr(): retry temporary address regeneration.\n");
935 		tmpaddr = &addr;
936 		write_lock(&idev->lock);
937 		goto retry;
938 	}
939 
940 	spin_lock_bh(&ift->lock);
941 	ift->ifpub = ifp;
942 	ift->valid_lft = tmp_valid_lft;
943 	ift->prefered_lft = tmp_prefered_lft;
944 	ift->cstamp = tmp_cstamp;
945 	ift->tstamp = tmp_tstamp;
946 	spin_unlock_bh(&ift->lock);
947 
948 	addrconf_dad_start(ift, 0);
949 	in6_ifa_put(ift);
950 	in6_dev_put(idev);
951 out:
952 	return ret;
953 }
954 #endif
955 
956 /*
957  *	Choose an appropriate source address (RFC3484)
958  */
959 enum {
960 	IPV6_SADDR_RULE_INIT = 0,
961 	IPV6_SADDR_RULE_LOCAL,
962 	IPV6_SADDR_RULE_SCOPE,
963 	IPV6_SADDR_RULE_PREFERRED,
964 #ifdef CONFIG_IPV6_MIP6
965 	IPV6_SADDR_RULE_HOA,
966 #endif
967 	IPV6_SADDR_RULE_OIF,
968 	IPV6_SADDR_RULE_LABEL,
969 #ifdef CONFIG_IPV6_PRIVACY
970 	IPV6_SADDR_RULE_PRIVACY,
971 #endif
972 	IPV6_SADDR_RULE_ORCHID,
973 	IPV6_SADDR_RULE_PREFIX,
974 	IPV6_SADDR_RULE_MAX
975 };
976 
977 struct ipv6_saddr_score {
978 	int			rule;
979 	int			addr_type;
980 	struct inet6_ifaddr	*ifa;
981 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
982 	int			scopedist;
983 	int			matchlen;
984 };
985 
986 struct ipv6_saddr_dst {
987 	const struct in6_addr *addr;
988 	int ifindex;
989 	int scope;
990 	int label;
991 	unsigned int prefs;
992 };
993 
994 static inline int ipv6_saddr_preferred(int type)
995 {
996 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
997 		return 1;
998 	return 0;
999 }
1000 
1001 static int ipv6_get_saddr_eval(struct net *net,
1002 			       struct ipv6_saddr_score *score,
1003 			       struct ipv6_saddr_dst *dst,
1004 			       int i)
1005 {
1006 	int ret;
1007 
1008 	if (i <= score->rule) {
1009 		switch (i) {
1010 		case IPV6_SADDR_RULE_SCOPE:
1011 			ret = score->scopedist;
1012 			break;
1013 		case IPV6_SADDR_RULE_PREFIX:
1014 			ret = score->matchlen;
1015 			break;
1016 		default:
1017 			ret = !!test_bit(i, score->scorebits);
1018 		}
1019 		goto out;
1020 	}
1021 
1022 	switch (i) {
1023 	case IPV6_SADDR_RULE_INIT:
1024 		/* Rule 0: remember if hiscore is not ready yet */
1025 		ret = !!score->ifa;
1026 		break;
1027 	case IPV6_SADDR_RULE_LOCAL:
1028 		/* Rule 1: Prefer same address */
1029 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030 		break;
1031 	case IPV6_SADDR_RULE_SCOPE:
1032 		/* Rule 2: Prefer appropriate scope
1033 		 *
1034 		 *      ret
1035 		 *       ^
1036 		 *    -1 |  d 15
1037 		 *    ---+--+-+---> scope
1038 		 *       |
1039 		 *       |             d is scope of the destination.
1040 		 *  B-d  |  \
1041 		 *       |   \      <- smaller scope is better if
1042 		 *  B-15 |    \        if scope is enough for destinaion.
1043 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1044 		 * d-C-1 | /
1045 		 *       |/         <- greater is better
1046 		 *   -C  /             if scope is not enough for destination.
1047 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1048 		 *
1049 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051 		 * Assume B = 0 and we get C > 29.
1052 		 */
1053 		ret = __ipv6_addr_src_scope(score->addr_type);
1054 		if (ret >= dst->scope)
1055 			ret = -ret;
1056 		else
1057 			ret -= 128;	/* 30 is enough */
1058 		score->scopedist = ret;
1059 		break;
1060 	case IPV6_SADDR_RULE_PREFERRED:
1061 		/* Rule 3: Avoid deprecated and optimistic addresses */
1062 		ret = ipv6_saddr_preferred(score->addr_type) ||
1063 		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064 		break;
1065 #ifdef CONFIG_IPV6_MIP6
1066 	case IPV6_SADDR_RULE_HOA:
1067 	    {
1068 		/* Rule 4: Prefer home address */
1069 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071 		break;
1072 	    }
1073 #endif
1074 	case IPV6_SADDR_RULE_OIF:
1075 		/* Rule 5: Prefer outgoing interface */
1076 		ret = (!dst->ifindex ||
1077 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1078 		break;
1079 	case IPV6_SADDR_RULE_LABEL:
1080 		/* Rule 6: Prefer matching label */
1081 		ret = ipv6_addr_label(net,
1082 				      &score->ifa->addr, score->addr_type,
1083 				      score->ifa->idev->dev->ifindex) == dst->label;
1084 		break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086 	case IPV6_SADDR_RULE_PRIVACY:
1087 	    {
1088 		/* Rule 7: Prefer public address
1089 		 * Note: prefer temprary address if use_tempaddr >= 2
1090 		 */
1091 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093 				score->ifa->idev->cnf.use_tempaddr >= 2;
1094 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095 		break;
1096 	    }
1097 #endif
1098 	case IPV6_SADDR_RULE_ORCHID:
1099 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1100 		 *	    non-ORCHID vs non-ORCHID
1101 		 */
1102 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103 			ipv6_addr_orchid(dst->addr));
1104 		break;
1105 	case IPV6_SADDR_RULE_PREFIX:
1106 		/* Rule 8: Use longest matching prefix */
1107 		score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108 						       dst->addr);
1109 		break;
1110 	default:
1111 		ret = 0;
1112 	}
1113 
1114 	if (ret)
1115 		__set_bit(i, score->scorebits);
1116 	score->rule = i;
1117 out:
1118 	return ret;
1119 }
1120 
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122 		       const struct in6_addr *daddr, unsigned int prefs,
1123 		       struct in6_addr *saddr)
1124 {
1125 	struct ipv6_saddr_score scores[2],
1126 				*score = &scores[0], *hiscore = &scores[1];
1127 	struct ipv6_saddr_dst dst;
1128 	struct net_device *dev;
1129 	int dst_type;
1130 
1131 	dst_type = __ipv6_addr_type(daddr);
1132 	dst.addr = daddr;
1133 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134 	dst.scope = __ipv6_addr_src_scope(dst_type);
1135 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136 	dst.prefs = prefs;
1137 
1138 	hiscore->rule = -1;
1139 	hiscore->ifa = NULL;
1140 
1141 	rcu_read_lock();
1142 
1143 	for_each_netdev_rcu(net, dev) {
1144 		struct inet6_dev *idev;
1145 
1146 		/* Candidate Source Address (section 4)
1147 		 *  - multicast and link-local destination address,
1148 		 *    the set of candidate source address MUST only
1149 		 *    include addresses assigned to interfaces
1150 		 *    belonging to the same link as the outgoing
1151 		 *    interface.
1152 		 * (- For site-local destination addresses, the
1153 		 *    set of candidate source addresses MUST only
1154 		 *    include addresses assigned to interfaces
1155 		 *    belonging to the same site as the outgoing
1156 		 *    interface.)
1157 		 */
1158 		if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159 		     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160 		    dst.ifindex && dev->ifindex != dst.ifindex)
1161 			continue;
1162 
1163 		idev = __in6_dev_get(dev);
1164 		if (!idev)
1165 			continue;
1166 
1167 		read_lock_bh(&idev->lock);
1168 		list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1169 			int i;
1170 
1171 			/*
1172 			 * - Tentative Address (RFC2462 section 5.4)
1173 			 *  - A tentative address is not considered
1174 			 *    "assigned to an interface" in the traditional
1175 			 *    sense, unless it is also flagged as optimistic.
1176 			 * - Candidate Source Address (section 4)
1177 			 *  - In any case, anycast addresses, multicast
1178 			 *    addresses, and the unspecified address MUST
1179 			 *    NOT be included in a candidate set.
1180 			 */
1181 			if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182 			    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183 				continue;
1184 
1185 			score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1186 
1187 			if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188 				     score->addr_type & IPV6_ADDR_MULTICAST)) {
1189 				LIMIT_NETDEBUG(KERN_DEBUG
1190 					       "ADDRCONF: unspecified / multicast address "
1191 					       "assigned as unicast address on %s",
1192 					       dev->name);
1193 				continue;
1194 			}
1195 
1196 			score->rule = -1;
1197 			bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1198 
1199 			for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200 				int minihiscore, miniscore;
1201 
1202 				minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203 				miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1204 
1205 				if (minihiscore > miniscore) {
1206 					if (i == IPV6_SADDR_RULE_SCOPE &&
1207 					    score->scopedist > 0) {
1208 						/*
1209 						 * special case:
1210 						 * each remaining entry
1211 						 * has too small (not enough)
1212 						 * scope, because ifa entries
1213 						 * are sorted by their scope
1214 						 * values.
1215 						 */
1216 						goto try_nextdev;
1217 					}
1218 					break;
1219 				} else if (minihiscore < miniscore) {
1220 					if (hiscore->ifa)
1221 						in6_ifa_put(hiscore->ifa);
1222 
1223 					in6_ifa_hold(score->ifa);
1224 
1225 					swap(hiscore, score);
1226 
1227 					/* restore our iterator */
1228 					score->ifa = hiscore->ifa;
1229 
1230 					break;
1231 				}
1232 			}
1233 		}
1234 try_nextdev:
1235 		read_unlock_bh(&idev->lock);
1236 	}
1237 	rcu_read_unlock();
1238 
1239 	if (!hiscore->ifa)
1240 		return -EADDRNOTAVAIL;
1241 
1242 	ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243 	in6_ifa_put(hiscore->ifa);
1244 	return 0;
1245 }
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1247 
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249 		    unsigned char banned_flags)
1250 {
1251 	struct inet6_dev *idev;
1252 	int err = -EADDRNOTAVAIL;
1253 
1254 	rcu_read_lock();
1255 	idev = __in6_dev_get(dev);
1256 	if (idev) {
1257 		struct inet6_ifaddr *ifp;
1258 
1259 		read_lock_bh(&idev->lock);
1260 		list_for_each_entry(ifp, &idev->addr_list, if_list) {
1261 			if (ifp->scope == IFA_LINK &&
1262 			    !(ifp->flags & banned_flags)) {
1263 				ipv6_addr_copy(addr, &ifp->addr);
1264 				err = 0;
1265 				break;
1266 			}
1267 		}
1268 		read_unlock_bh(&idev->lock);
1269 	}
1270 	rcu_read_unlock();
1271 	return err;
1272 }
1273 
1274 static int ipv6_count_addresses(struct inet6_dev *idev)
1275 {
1276 	int cnt = 0;
1277 	struct inet6_ifaddr *ifp;
1278 
1279 	read_lock_bh(&idev->lock);
1280 	list_for_each_entry(ifp, &idev->addr_list, if_list)
1281 		cnt++;
1282 	read_unlock_bh(&idev->lock);
1283 	return cnt;
1284 }
1285 
1286 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1287 		  struct net_device *dev, int strict)
1288 {
1289 	struct inet6_ifaddr *ifp;
1290 	struct hlist_node *node;
1291 	unsigned int hash = ipv6_addr_hash(addr);
1292 
1293 	rcu_read_lock_bh();
1294 	hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1295 		if (!net_eq(dev_net(ifp->idev->dev), net))
1296 			continue;
1297 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1298 		    !(ifp->flags&IFA_F_TENTATIVE) &&
1299 		    (dev == NULL || ifp->idev->dev == dev ||
1300 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1301 			rcu_read_unlock_bh();
1302 			return 1;
1303 		}
1304 	}
1305 
1306 	rcu_read_unlock_bh();
1307 	return 0;
1308 }
1309 EXPORT_SYMBOL(ipv6_chk_addr);
1310 
1311 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1312 			       struct net_device *dev)
1313 {
1314 	unsigned int hash = ipv6_addr_hash(addr);
1315 	struct inet6_ifaddr *ifp;
1316 	struct hlist_node *node;
1317 
1318 	hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1319 		if (!net_eq(dev_net(ifp->idev->dev), net))
1320 			continue;
1321 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1322 			if (dev == NULL || ifp->idev->dev == dev)
1323 				return true;
1324 		}
1325 	}
1326 	return false;
1327 }
1328 
1329 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1330 {
1331 	struct inet6_dev *idev;
1332 	struct inet6_ifaddr *ifa;
1333 	int	onlink;
1334 
1335 	onlink = 0;
1336 	rcu_read_lock();
1337 	idev = __in6_dev_get(dev);
1338 	if (idev) {
1339 		read_lock_bh(&idev->lock);
1340 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1341 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1342 						   ifa->prefix_len);
1343 			if (onlink)
1344 				break;
1345 		}
1346 		read_unlock_bh(&idev->lock);
1347 	}
1348 	rcu_read_unlock();
1349 	return onlink;
1350 }
1351 
1352 EXPORT_SYMBOL(ipv6_chk_prefix);
1353 
1354 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1355 				     struct net_device *dev, int strict)
1356 {
1357 	struct inet6_ifaddr *ifp, *result = NULL;
1358 	unsigned int hash = ipv6_addr_hash(addr);
1359 	struct hlist_node *node;
1360 
1361 	rcu_read_lock_bh();
1362 	hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1363 		if (!net_eq(dev_net(ifp->idev->dev), net))
1364 			continue;
1365 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1366 			if (dev == NULL || ifp->idev->dev == dev ||
1367 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1368 				result = ifp;
1369 				in6_ifa_hold(ifp);
1370 				break;
1371 			}
1372 		}
1373 	}
1374 	rcu_read_unlock_bh();
1375 
1376 	return result;
1377 }
1378 
1379 /* Gets referenced address, destroys ifaddr */
1380 
1381 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1382 {
1383 	if (ifp->flags&IFA_F_PERMANENT) {
1384 		spin_lock_bh(&ifp->lock);
1385 		addrconf_del_timer(ifp);
1386 		ifp->flags |= IFA_F_TENTATIVE;
1387 		if (dad_failed)
1388 			ifp->flags |= IFA_F_DADFAILED;
1389 		spin_unlock_bh(&ifp->lock);
1390 		if (dad_failed)
1391 			ipv6_ifa_notify(0, ifp);
1392 		in6_ifa_put(ifp);
1393 #ifdef CONFIG_IPV6_PRIVACY
1394 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1395 		struct inet6_ifaddr *ifpub;
1396 		spin_lock_bh(&ifp->lock);
1397 		ifpub = ifp->ifpub;
1398 		if (ifpub) {
1399 			in6_ifa_hold(ifpub);
1400 			spin_unlock_bh(&ifp->lock);
1401 			ipv6_create_tempaddr(ifpub, ifp);
1402 			in6_ifa_put(ifpub);
1403 		} else {
1404 			spin_unlock_bh(&ifp->lock);
1405 		}
1406 		ipv6_del_addr(ifp);
1407 #endif
1408 	} else
1409 		ipv6_del_addr(ifp);
1410 }
1411 
1412 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1413 {
1414 	int err = -ENOENT;
1415 
1416 	spin_lock(&ifp->state_lock);
1417 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
1418 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
1419 		err = 0;
1420 	}
1421 	spin_unlock(&ifp->state_lock);
1422 
1423 	return err;
1424 }
1425 
1426 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1427 {
1428 	struct inet6_dev *idev = ifp->idev;
1429 
1430 	if (addrconf_dad_end(ifp)) {
1431 		in6_ifa_put(ifp);
1432 		return;
1433 	}
1434 
1435 	if (net_ratelimit())
1436 		printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1437 			ifp->idev->dev->name, &ifp->addr);
1438 
1439 	if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1440 		struct in6_addr addr;
1441 
1442 		addr.s6_addr32[0] = htonl(0xfe800000);
1443 		addr.s6_addr32[1] = 0;
1444 
1445 		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1446 		    ipv6_addr_equal(&ifp->addr, &addr)) {
1447 			/* DAD failed for link-local based on MAC address */
1448 			idev->cnf.disable_ipv6 = 1;
1449 
1450 			printk(KERN_INFO "%s: IPv6 being disabled!\n",
1451 				ifp->idev->dev->name);
1452 		}
1453 	}
1454 
1455 	addrconf_dad_stop(ifp, 1);
1456 }
1457 
1458 /* Join to solicited addr multicast group. */
1459 
1460 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1461 {
1462 	struct in6_addr maddr;
1463 
1464 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1465 		return;
1466 
1467 	addrconf_addr_solict_mult(addr, &maddr);
1468 	ipv6_dev_mc_inc(dev, &maddr);
1469 }
1470 
1471 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1472 {
1473 	struct in6_addr maddr;
1474 
1475 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1476 		return;
1477 
1478 	addrconf_addr_solict_mult(addr, &maddr);
1479 	__ipv6_dev_mc_dec(idev, &maddr);
1480 }
1481 
1482 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1483 {
1484 	struct in6_addr addr;
1485 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1486 	if (ipv6_addr_any(&addr))
1487 		return;
1488 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1489 }
1490 
1491 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1492 {
1493 	struct in6_addr addr;
1494 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1495 	if (ipv6_addr_any(&addr))
1496 		return;
1497 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1498 }
1499 
1500 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1501 {
1502 	if (dev->addr_len != ETH_ALEN)
1503 		return -1;
1504 	memcpy(eui, dev->dev_addr, 3);
1505 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1506 
1507 	/*
1508 	 * The zSeries OSA network cards can be shared among various
1509 	 * OS instances, but the OSA cards have only one MAC address.
1510 	 * This leads to duplicate address conflicts in conjunction
1511 	 * with IPv6 if more than one instance uses the same card.
1512 	 *
1513 	 * The driver for these cards can deliver a unique 16-bit
1514 	 * identifier for each instance sharing the same card.  It is
1515 	 * placed instead of 0xFFFE in the interface identifier.  The
1516 	 * "u" bit of the interface identifier is not inverted in this
1517 	 * case.  Hence the resulting interface identifier has local
1518 	 * scope according to RFC2373.
1519 	 */
1520 	if (dev->dev_id) {
1521 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1522 		eui[4] = dev->dev_id & 0xFF;
1523 	} else {
1524 		eui[3] = 0xFF;
1525 		eui[4] = 0xFE;
1526 		eui[0] ^= 2;
1527 	}
1528 	return 0;
1529 }
1530 
1531 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1532 {
1533 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1534 	if (dev->addr_len != ARCNET_ALEN)
1535 		return -1;
1536 	memset(eui, 0, 7);
1537 	eui[7] = *(u8*)dev->dev_addr;
1538 	return 0;
1539 }
1540 
1541 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1542 {
1543 	if (dev->addr_len != INFINIBAND_ALEN)
1544 		return -1;
1545 	memcpy(eui, dev->dev_addr + 12, 8);
1546 	eui[0] |= 2;
1547 	return 0;
1548 }
1549 
1550 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1551 {
1552 	if (addr == 0)
1553 		return -1;
1554 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1555 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1556 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1557 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1558 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1559 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1560 	eui[1] = 0;
1561 	eui[2] = 0x5E;
1562 	eui[3] = 0xFE;
1563 	memcpy(eui + 4, &addr, 4);
1564 	return 0;
1565 }
1566 
1567 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1568 {
1569 	if (dev->priv_flags & IFF_ISATAP)
1570 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1571 	return -1;
1572 }
1573 
1574 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1575 {
1576 	switch (dev->type) {
1577 	case ARPHRD_ETHER:
1578 	case ARPHRD_FDDI:
1579 	case ARPHRD_IEEE802_TR:
1580 		return addrconf_ifid_eui48(eui, dev);
1581 	case ARPHRD_ARCNET:
1582 		return addrconf_ifid_arcnet(eui, dev);
1583 	case ARPHRD_INFINIBAND:
1584 		return addrconf_ifid_infiniband(eui, dev);
1585 	case ARPHRD_SIT:
1586 		return addrconf_ifid_sit(eui, dev);
1587 	}
1588 	return -1;
1589 }
1590 
1591 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1592 {
1593 	int err = -1;
1594 	struct inet6_ifaddr *ifp;
1595 
1596 	read_lock_bh(&idev->lock);
1597 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
1598 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1599 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1600 			err = 0;
1601 			break;
1602 		}
1603 	}
1604 	read_unlock_bh(&idev->lock);
1605 	return err;
1606 }
1607 
1608 #ifdef CONFIG_IPV6_PRIVACY
1609 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1610 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1611 {
1612 regen:
1613 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1614 	idev->rndid[0] &= ~0x02;
1615 
1616 	/*
1617 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1618 	 * check if generated address is not inappropriate
1619 	 *
1620 	 *  - Reserved subnet anycast (RFC 2526)
1621 	 *	11111101 11....11 1xxxxxxx
1622 	 *  - ISATAP (RFC4214) 6.1
1623 	 *	00-00-5E-FE-xx-xx-xx-xx
1624 	 *  - value 0
1625 	 *  - XXX: already assigned to an address on the device
1626 	 */
1627 	if (idev->rndid[0] == 0xfd &&
1628 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1629 	    (idev->rndid[7]&0x80))
1630 		goto regen;
1631 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1632 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1633 			goto regen;
1634 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1635 			goto regen;
1636 	}
1637 
1638 	return 0;
1639 }
1640 
1641 static void ipv6_regen_rndid(unsigned long data)
1642 {
1643 	struct inet6_dev *idev = (struct inet6_dev *) data;
1644 	unsigned long expires;
1645 
1646 	rcu_read_lock_bh();
1647 	write_lock_bh(&idev->lock);
1648 
1649 	if (idev->dead)
1650 		goto out;
1651 
1652 	if (__ipv6_regen_rndid(idev) < 0)
1653 		goto out;
1654 
1655 	expires = jiffies +
1656 		idev->cnf.temp_prefered_lft * HZ -
1657 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1658 		idev->cnf.max_desync_factor * HZ;
1659 	if (time_before(expires, jiffies)) {
1660 		printk(KERN_WARNING
1661 			"ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1662 			idev->dev->name);
1663 		goto out;
1664 	}
1665 
1666 	if (!mod_timer(&idev->regen_timer, expires))
1667 		in6_dev_hold(idev);
1668 
1669 out:
1670 	write_unlock_bh(&idev->lock);
1671 	rcu_read_unlock_bh();
1672 	in6_dev_put(idev);
1673 }
1674 
1675 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1676 	int ret = 0;
1677 
1678 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1679 		ret = __ipv6_regen_rndid(idev);
1680 	return ret;
1681 }
1682 #endif
1683 
1684 /*
1685  *	Add prefix route.
1686  */
1687 
1688 static void
1689 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1690 		      unsigned long expires, u32 flags)
1691 {
1692 	struct fib6_config cfg = {
1693 		.fc_table = RT6_TABLE_PREFIX,
1694 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1695 		.fc_ifindex = dev->ifindex,
1696 		.fc_expires = expires,
1697 		.fc_dst_len = plen,
1698 		.fc_flags = RTF_UP | flags,
1699 		.fc_nlinfo.nl_net = dev_net(dev),
1700 		.fc_protocol = RTPROT_KERNEL,
1701 	};
1702 
1703 	ipv6_addr_copy(&cfg.fc_dst, pfx);
1704 
1705 	/* Prevent useless cloning on PtP SIT.
1706 	   This thing is done here expecting that the whole
1707 	   class of non-broadcast devices need not cloning.
1708 	 */
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1711 		cfg.fc_flags |= RTF_NONEXTHOP;
1712 #endif
1713 
1714 	ip6_route_add(&cfg);
1715 }
1716 
1717 /* Create "default" multicast route to the interface */
1718 
1719 static void addrconf_add_mroute(struct net_device *dev)
1720 {
1721 	struct fib6_config cfg = {
1722 		.fc_table = RT6_TABLE_LOCAL,
1723 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1724 		.fc_ifindex = dev->ifindex,
1725 		.fc_dst_len = 8,
1726 		.fc_flags = RTF_UP,
1727 		.fc_nlinfo.nl_net = dev_net(dev),
1728 	};
1729 
1730 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1731 
1732 	ip6_route_add(&cfg);
1733 }
1734 
1735 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1736 static void sit_route_add(struct net_device *dev)
1737 {
1738 	struct fib6_config cfg = {
1739 		.fc_table = RT6_TABLE_MAIN,
1740 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1741 		.fc_ifindex = dev->ifindex,
1742 		.fc_dst_len = 96,
1743 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
1744 		.fc_nlinfo.nl_net = dev_net(dev),
1745 	};
1746 
1747 	/* prefix length - 96 bits "::d.d.d.d" */
1748 	ip6_route_add(&cfg);
1749 }
1750 #endif
1751 
1752 static void addrconf_add_lroute(struct net_device *dev)
1753 {
1754 	struct in6_addr addr;
1755 
1756 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1757 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
1758 }
1759 
1760 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1761 {
1762 	struct inet6_dev *idev;
1763 
1764 	ASSERT_RTNL();
1765 
1766 	idev = ipv6_find_idev(dev);
1767 	if (!idev)
1768 		return ERR_PTR(-ENOBUFS);
1769 
1770 	if (idev->cnf.disable_ipv6)
1771 		return ERR_PTR(-EACCES);
1772 
1773 	/* Add default multicast route */
1774 	addrconf_add_mroute(dev);
1775 
1776 	/* Add link local route */
1777 	addrconf_add_lroute(dev);
1778 	return idev;
1779 }
1780 
1781 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1782 {
1783 	struct prefix_info *pinfo;
1784 	__u32 valid_lft;
1785 	__u32 prefered_lft;
1786 	int addr_type;
1787 	struct inet6_dev *in6_dev;
1788 	struct net *net = dev_net(dev);
1789 
1790 	pinfo = (struct prefix_info *) opt;
1791 
1792 	if (len < sizeof(struct prefix_info)) {
1793 		ADBG(("addrconf: prefix option too short\n"));
1794 		return;
1795 	}
1796 
1797 	/*
1798 	 *	Validation checks ([ADDRCONF], page 19)
1799 	 */
1800 
1801 	addr_type = ipv6_addr_type(&pinfo->prefix);
1802 
1803 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1804 		return;
1805 
1806 	valid_lft = ntohl(pinfo->valid);
1807 	prefered_lft = ntohl(pinfo->prefered);
1808 
1809 	if (prefered_lft > valid_lft) {
1810 		if (net_ratelimit())
1811 			printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1812 		return;
1813 	}
1814 
1815 	in6_dev = in6_dev_get(dev);
1816 
1817 	if (in6_dev == NULL) {
1818 		if (net_ratelimit())
1819 			printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1820 		return;
1821 	}
1822 
1823 	/*
1824 	 *	Two things going on here:
1825 	 *	1) Add routes for on-link prefixes
1826 	 *	2) Configure prefixes with the auto flag set
1827 	 */
1828 
1829 	if (pinfo->onlink) {
1830 		struct rt6_info *rt;
1831 		unsigned long rt_expires;
1832 
1833 		/* Avoid arithmetic overflow. Really, we could
1834 		 * save rt_expires in seconds, likely valid_lft,
1835 		 * but it would require division in fib gc, that it
1836 		 * not good.
1837 		 */
1838 		if (HZ > USER_HZ)
1839 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1840 		else
1841 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1842 
1843 		if (addrconf_finite_timeout(rt_expires))
1844 			rt_expires *= HZ;
1845 
1846 		rt = rt6_lookup(net, &pinfo->prefix, NULL,
1847 				dev->ifindex, 1);
1848 
1849 		if (rt && addrconf_is_prefix_route(rt)) {
1850 			/* Autoconf prefix route */
1851 			if (valid_lft == 0) {
1852 				ip6_del_rt(rt);
1853 				rt = NULL;
1854 			} else if (addrconf_finite_timeout(rt_expires)) {
1855 				/* not infinity */
1856 				rt->rt6i_expires = jiffies + rt_expires;
1857 				rt->rt6i_flags |= RTF_EXPIRES;
1858 			} else {
1859 				rt->rt6i_flags &= ~RTF_EXPIRES;
1860 				rt->rt6i_expires = 0;
1861 			}
1862 		} else if (valid_lft) {
1863 			clock_t expires = 0;
1864 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1865 			if (addrconf_finite_timeout(rt_expires)) {
1866 				/* not infinity */
1867 				flags |= RTF_EXPIRES;
1868 				expires = jiffies_to_clock_t(rt_expires);
1869 			}
1870 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1871 					      dev, expires, flags);
1872 		}
1873 		if (rt)
1874 			dst_release(&rt->dst);
1875 	}
1876 
1877 	/* Try to figure out our local address for this prefix */
1878 
1879 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1880 		struct inet6_ifaddr * ifp;
1881 		struct in6_addr addr;
1882 		int create = 0, update_lft = 0;
1883 
1884 		if (pinfo->prefix_len == 64) {
1885 			memcpy(&addr, &pinfo->prefix, 8);
1886 			if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1887 			    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1888 				in6_dev_put(in6_dev);
1889 				return;
1890 			}
1891 			goto ok;
1892 		}
1893 		if (net_ratelimit())
1894 			printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1895 			       pinfo->prefix_len);
1896 		in6_dev_put(in6_dev);
1897 		return;
1898 
1899 ok:
1900 
1901 		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1902 
1903 		if (ifp == NULL && valid_lft) {
1904 			int max_addresses = in6_dev->cnf.max_addresses;
1905 			u32 addr_flags = 0;
1906 
1907 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1908 			if (in6_dev->cnf.optimistic_dad &&
1909 			    !net->ipv6.devconf_all->forwarding)
1910 				addr_flags = IFA_F_OPTIMISTIC;
1911 #endif
1912 
1913 			/* Do not allow to create too much of autoconfigured
1914 			 * addresses; this would be too easy way to crash kernel.
1915 			 */
1916 			if (!max_addresses ||
1917 			    ipv6_count_addresses(in6_dev) < max_addresses)
1918 				ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1919 						    addr_type&IPV6_ADDR_SCOPE_MASK,
1920 						    addr_flags);
1921 
1922 			if (!ifp || IS_ERR(ifp)) {
1923 				in6_dev_put(in6_dev);
1924 				return;
1925 			}
1926 
1927 			update_lft = create = 1;
1928 			ifp->cstamp = jiffies;
1929 			addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1930 		}
1931 
1932 		if (ifp) {
1933 			int flags;
1934 			unsigned long now;
1935 #ifdef CONFIG_IPV6_PRIVACY
1936 			struct inet6_ifaddr *ift;
1937 #endif
1938 			u32 stored_lft;
1939 
1940 			/* update lifetime (RFC2462 5.5.3 e) */
1941 			spin_lock(&ifp->lock);
1942 			now = jiffies;
1943 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1944 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1945 			else
1946 				stored_lft = 0;
1947 			if (!update_lft && stored_lft) {
1948 				if (valid_lft > MIN_VALID_LIFETIME ||
1949 				    valid_lft > stored_lft)
1950 					update_lft = 1;
1951 				else if (stored_lft <= MIN_VALID_LIFETIME) {
1952 					/* valid_lft <= stored_lft is always true */
1953 					/*
1954 					 * RFC 4862 Section 5.5.3e:
1955 					 * "Note that the preferred lifetime of
1956 					 *  the corresponding address is always
1957 					 *  reset to the Preferred Lifetime in
1958 					 *  the received Prefix Information
1959 					 *  option, regardless of whether the
1960 					 *  valid lifetime is also reset or
1961 					 *  ignored."
1962 					 *
1963 					 *  So if the preferred lifetime in
1964 					 *  this advertisement is different
1965 					 *  than what we have stored, but the
1966 					 *  valid lifetime is invalid, just
1967 					 *  reset prefered_lft.
1968 					 *
1969 					 *  We must set the valid lifetime
1970 					 *  to the stored lifetime since we'll
1971 					 *  be updating the timestamp below,
1972 					 *  else we'll set it back to the
1973 					 *  minumum.
1974 					 */
1975 					if (prefered_lft != ifp->prefered_lft) {
1976 						valid_lft = stored_lft;
1977 						update_lft = 1;
1978 					}
1979 				} else {
1980 					valid_lft = MIN_VALID_LIFETIME;
1981 					if (valid_lft < prefered_lft)
1982 						prefered_lft = valid_lft;
1983 					update_lft = 1;
1984 				}
1985 			}
1986 
1987 			if (update_lft) {
1988 				ifp->valid_lft = valid_lft;
1989 				ifp->prefered_lft = prefered_lft;
1990 				ifp->tstamp = now;
1991 				flags = ifp->flags;
1992 				ifp->flags &= ~IFA_F_DEPRECATED;
1993 				spin_unlock(&ifp->lock);
1994 
1995 				if (!(flags&IFA_F_TENTATIVE))
1996 					ipv6_ifa_notify(0, ifp);
1997 			} else
1998 				spin_unlock(&ifp->lock);
1999 
2000 #ifdef CONFIG_IPV6_PRIVACY
2001 			read_lock_bh(&in6_dev->lock);
2002 			/* update all temporary addresses in the list */
2003 			list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2004 				/*
2005 				 * When adjusting the lifetimes of an existing
2006 				 * temporary address, only lower the lifetimes.
2007 				 * Implementations must not increase the
2008 				 * lifetimes of an existing temporary address
2009 				 * when processing a Prefix Information Option.
2010 				 */
2011 				if (ifp != ift->ifpub)
2012 					continue;
2013 
2014 				spin_lock(&ift->lock);
2015 				flags = ift->flags;
2016 				if (ift->valid_lft > valid_lft &&
2017 				    ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2018 					ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2019 				if (ift->prefered_lft > prefered_lft &&
2020 				    ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2021 					ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2022 				spin_unlock(&ift->lock);
2023 				if (!(flags&IFA_F_TENTATIVE))
2024 					ipv6_ifa_notify(0, ift);
2025 			}
2026 
2027 			if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2028 				/*
2029 				 * When a new public address is created as described in [ADDRCONF],
2030 				 * also create a new temporary address. Also create a temporary
2031 				 * address if it's enabled but no temporary address currently exists.
2032 				 */
2033 				read_unlock_bh(&in6_dev->lock);
2034 				ipv6_create_tempaddr(ifp, NULL);
2035 			} else {
2036 				read_unlock_bh(&in6_dev->lock);
2037 			}
2038 #endif
2039 			in6_ifa_put(ifp);
2040 			addrconf_verify(0);
2041 		}
2042 	}
2043 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2044 	in6_dev_put(in6_dev);
2045 }
2046 
2047 /*
2048  *	Set destination address.
2049  *	Special case for SIT interfaces where we create a new "virtual"
2050  *	device.
2051  */
2052 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2053 {
2054 	struct in6_ifreq ireq;
2055 	struct net_device *dev;
2056 	int err = -EINVAL;
2057 
2058 	rtnl_lock();
2059 
2060 	err = -EFAULT;
2061 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2062 		goto err_exit;
2063 
2064 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2065 
2066 	err = -ENODEV;
2067 	if (dev == NULL)
2068 		goto err_exit;
2069 
2070 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2071 	if (dev->type == ARPHRD_SIT) {
2072 		const struct net_device_ops *ops = dev->netdev_ops;
2073 		struct ifreq ifr;
2074 		struct ip_tunnel_parm p;
2075 
2076 		err = -EADDRNOTAVAIL;
2077 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2078 			goto err_exit;
2079 
2080 		memset(&p, 0, sizeof(p));
2081 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2082 		p.iph.saddr = 0;
2083 		p.iph.version = 4;
2084 		p.iph.ihl = 5;
2085 		p.iph.protocol = IPPROTO_IPV6;
2086 		p.iph.ttl = 64;
2087 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2088 
2089 		if (ops->ndo_do_ioctl) {
2090 			mm_segment_t oldfs = get_fs();
2091 
2092 			set_fs(KERNEL_DS);
2093 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2094 			set_fs(oldfs);
2095 		} else
2096 			err = -EOPNOTSUPP;
2097 
2098 		if (err == 0) {
2099 			err = -ENOBUFS;
2100 			dev = __dev_get_by_name(net, p.name);
2101 			if (!dev)
2102 				goto err_exit;
2103 			err = dev_open(dev);
2104 		}
2105 	}
2106 #endif
2107 
2108 err_exit:
2109 	rtnl_unlock();
2110 	return err;
2111 }
2112 
2113 /*
2114  *	Manual configuration of address on an interface
2115  */
2116 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2117 			  unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2118 			  __u32 valid_lft)
2119 {
2120 	struct inet6_ifaddr *ifp;
2121 	struct inet6_dev *idev;
2122 	struct net_device *dev;
2123 	int scope;
2124 	u32 flags;
2125 	clock_t expires;
2126 	unsigned long timeout;
2127 
2128 	ASSERT_RTNL();
2129 
2130 	if (plen > 128)
2131 		return -EINVAL;
2132 
2133 	/* check the lifetime */
2134 	if (!valid_lft || prefered_lft > valid_lft)
2135 		return -EINVAL;
2136 
2137 	dev = __dev_get_by_index(net, ifindex);
2138 	if (!dev)
2139 		return -ENODEV;
2140 
2141 	idev = addrconf_add_dev(dev);
2142 	if (IS_ERR(idev))
2143 		return PTR_ERR(idev);
2144 
2145 	scope = ipv6_addr_scope(pfx);
2146 
2147 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2148 	if (addrconf_finite_timeout(timeout)) {
2149 		expires = jiffies_to_clock_t(timeout * HZ);
2150 		valid_lft = timeout;
2151 		flags = RTF_EXPIRES;
2152 	} else {
2153 		expires = 0;
2154 		flags = 0;
2155 		ifa_flags |= IFA_F_PERMANENT;
2156 	}
2157 
2158 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2159 	if (addrconf_finite_timeout(timeout)) {
2160 		if (timeout == 0)
2161 			ifa_flags |= IFA_F_DEPRECATED;
2162 		prefered_lft = timeout;
2163 	}
2164 
2165 	ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2166 
2167 	if (!IS_ERR(ifp)) {
2168 		spin_lock_bh(&ifp->lock);
2169 		ifp->valid_lft = valid_lft;
2170 		ifp->prefered_lft = prefered_lft;
2171 		ifp->tstamp = jiffies;
2172 		spin_unlock_bh(&ifp->lock);
2173 
2174 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2175 				      expires, flags);
2176 		/*
2177 		 * Note that section 3.1 of RFC 4429 indicates
2178 		 * that the Optimistic flag should not be set for
2179 		 * manually configured addresses
2180 		 */
2181 		addrconf_dad_start(ifp, 0);
2182 		in6_ifa_put(ifp);
2183 		addrconf_verify(0);
2184 		return 0;
2185 	}
2186 
2187 	return PTR_ERR(ifp);
2188 }
2189 
2190 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2191 			  unsigned int plen)
2192 {
2193 	struct inet6_ifaddr *ifp;
2194 	struct inet6_dev *idev;
2195 	struct net_device *dev;
2196 
2197 	if (plen > 128)
2198 		return -EINVAL;
2199 
2200 	dev = __dev_get_by_index(net, ifindex);
2201 	if (!dev)
2202 		return -ENODEV;
2203 
2204 	if ((idev = __in6_dev_get(dev)) == NULL)
2205 		return -ENXIO;
2206 
2207 	read_lock_bh(&idev->lock);
2208 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2209 		if (ifp->prefix_len == plen &&
2210 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2211 			in6_ifa_hold(ifp);
2212 			read_unlock_bh(&idev->lock);
2213 
2214 			ipv6_del_addr(ifp);
2215 
2216 			/* If the last address is deleted administratively,
2217 			   disable IPv6 on this interface.
2218 			 */
2219 			if (list_empty(&idev->addr_list))
2220 				addrconf_ifdown(idev->dev, 1);
2221 			return 0;
2222 		}
2223 	}
2224 	read_unlock_bh(&idev->lock);
2225 	return -EADDRNOTAVAIL;
2226 }
2227 
2228 
2229 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2230 {
2231 	struct in6_ifreq ireq;
2232 	int err;
2233 
2234 	if (!capable(CAP_NET_ADMIN))
2235 		return -EPERM;
2236 
2237 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2238 		return -EFAULT;
2239 
2240 	rtnl_lock();
2241 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2242 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2243 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2244 	rtnl_unlock();
2245 	return err;
2246 }
2247 
2248 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2249 {
2250 	struct in6_ifreq ireq;
2251 	int err;
2252 
2253 	if (!capable(CAP_NET_ADMIN))
2254 		return -EPERM;
2255 
2256 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2257 		return -EFAULT;
2258 
2259 	rtnl_lock();
2260 	err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2261 			     ireq.ifr6_prefixlen);
2262 	rtnl_unlock();
2263 	return err;
2264 }
2265 
2266 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2267 		     int plen, int scope)
2268 {
2269 	struct inet6_ifaddr *ifp;
2270 
2271 	ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2272 	if (!IS_ERR(ifp)) {
2273 		spin_lock_bh(&ifp->lock);
2274 		ifp->flags &= ~IFA_F_TENTATIVE;
2275 		spin_unlock_bh(&ifp->lock);
2276 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2277 		in6_ifa_put(ifp);
2278 	}
2279 }
2280 
2281 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2282 static void sit_add_v4_addrs(struct inet6_dev *idev)
2283 {
2284 	struct in6_addr addr;
2285 	struct net_device *dev;
2286 	struct net *net = dev_net(idev->dev);
2287 	int scope;
2288 
2289 	ASSERT_RTNL();
2290 
2291 	memset(&addr, 0, sizeof(struct in6_addr));
2292 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2293 
2294 	if (idev->dev->flags&IFF_POINTOPOINT) {
2295 		addr.s6_addr32[0] = htonl(0xfe800000);
2296 		scope = IFA_LINK;
2297 	} else {
2298 		scope = IPV6_ADDR_COMPATv4;
2299 	}
2300 
2301 	if (addr.s6_addr32[3]) {
2302 		add_addr(idev, &addr, 128, scope);
2303 		return;
2304 	}
2305 
2306 	for_each_netdev(net, dev) {
2307 		struct in_device * in_dev = __in_dev_get_rtnl(dev);
2308 		if (in_dev && (dev->flags & IFF_UP)) {
2309 			struct in_ifaddr * ifa;
2310 
2311 			int flag = scope;
2312 
2313 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2314 				int plen;
2315 
2316 				addr.s6_addr32[3] = ifa->ifa_local;
2317 
2318 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2319 					continue;
2320 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2321 					if (idev->dev->flags&IFF_POINTOPOINT)
2322 						continue;
2323 					flag |= IFA_HOST;
2324 				}
2325 				if (idev->dev->flags&IFF_POINTOPOINT)
2326 					plen = 64;
2327 				else
2328 					plen = 96;
2329 
2330 				add_addr(idev, &addr, plen, flag);
2331 			}
2332 		}
2333 	}
2334 }
2335 #endif
2336 
2337 static void init_loopback(struct net_device *dev)
2338 {
2339 	struct inet6_dev  *idev;
2340 
2341 	/* ::1 */
2342 
2343 	ASSERT_RTNL();
2344 
2345 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2346 		printk(KERN_DEBUG "init loopback: add_dev failed\n");
2347 		return;
2348 	}
2349 
2350 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2351 }
2352 
2353 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2354 {
2355 	struct inet6_ifaddr * ifp;
2356 	u32 addr_flags = IFA_F_PERMANENT;
2357 
2358 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2359 	if (idev->cnf.optimistic_dad &&
2360 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2361 		addr_flags |= IFA_F_OPTIMISTIC;
2362 #endif
2363 
2364 
2365 	ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2366 	if (!IS_ERR(ifp)) {
2367 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2368 		addrconf_dad_start(ifp, 0);
2369 		in6_ifa_put(ifp);
2370 	}
2371 }
2372 
2373 static void addrconf_dev_config(struct net_device *dev)
2374 {
2375 	struct in6_addr addr;
2376 	struct inet6_dev    * idev;
2377 
2378 	ASSERT_RTNL();
2379 
2380 	if ((dev->type != ARPHRD_ETHER) &&
2381 	    (dev->type != ARPHRD_FDDI) &&
2382 	    (dev->type != ARPHRD_IEEE802_TR) &&
2383 	    (dev->type != ARPHRD_ARCNET) &&
2384 	    (dev->type != ARPHRD_INFINIBAND)) {
2385 		/* Alas, we support only Ethernet autoconfiguration. */
2386 		return;
2387 	}
2388 
2389 	idev = addrconf_add_dev(dev);
2390 	if (IS_ERR(idev))
2391 		return;
2392 
2393 	memset(&addr, 0, sizeof(struct in6_addr));
2394 	addr.s6_addr32[0] = htonl(0xFE800000);
2395 
2396 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2397 		addrconf_add_linklocal(idev, &addr);
2398 }
2399 
2400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2401 static void addrconf_sit_config(struct net_device *dev)
2402 {
2403 	struct inet6_dev *idev;
2404 
2405 	ASSERT_RTNL();
2406 
2407 	/*
2408 	 * Configure the tunnel with one of our IPv4
2409 	 * addresses... we should configure all of
2410 	 * our v4 addrs in the tunnel
2411 	 */
2412 
2413 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2414 		printk(KERN_DEBUG "init sit: add_dev failed\n");
2415 		return;
2416 	}
2417 
2418 	if (dev->priv_flags & IFF_ISATAP) {
2419 		struct in6_addr addr;
2420 
2421 		ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2422 		addrconf_prefix_route(&addr, 64, dev, 0, 0);
2423 		if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2424 			addrconf_add_linklocal(idev, &addr);
2425 		return;
2426 	}
2427 
2428 	sit_add_v4_addrs(idev);
2429 
2430 	if (dev->flags&IFF_POINTOPOINT) {
2431 		addrconf_add_mroute(dev);
2432 		addrconf_add_lroute(dev);
2433 	} else
2434 		sit_route_add(dev);
2435 }
2436 #endif
2437 
2438 static inline int
2439 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2440 {
2441 	struct in6_addr lladdr;
2442 
2443 	if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2444 		addrconf_add_linklocal(idev, &lladdr);
2445 		return 0;
2446 	}
2447 	return -1;
2448 }
2449 
2450 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2451 {
2452 	struct net_device *link_dev;
2453 	struct net *net = dev_net(idev->dev);
2454 
2455 	/* first try to inherit the link-local address from the link device */
2456 	if (idev->dev->iflink &&
2457 	    (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2458 		if (!ipv6_inherit_linklocal(idev, link_dev))
2459 			return;
2460 	}
2461 	/* then try to inherit it from any device */
2462 	for_each_netdev(net, link_dev) {
2463 		if (!ipv6_inherit_linklocal(idev, link_dev))
2464 			return;
2465 	}
2466 	printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2467 }
2468 
2469 /*
2470  * Autoconfigure tunnel with a link-local address so routing protocols,
2471  * DHCPv6, MLD etc. can be run over the virtual link
2472  */
2473 
2474 static void addrconf_ip6_tnl_config(struct net_device *dev)
2475 {
2476 	struct inet6_dev *idev;
2477 
2478 	ASSERT_RTNL();
2479 
2480 	idev = addrconf_add_dev(dev);
2481 	if (IS_ERR(idev)) {
2482 		printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2483 		return;
2484 	}
2485 	ip6_tnl_add_linklocal(idev);
2486 }
2487 
2488 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2489 			   void * data)
2490 {
2491 	struct net_device *dev = (struct net_device *) data;
2492 	struct inet6_dev *idev = __in6_dev_get(dev);
2493 	int run_pending = 0;
2494 	int err;
2495 
2496 	switch (event) {
2497 	case NETDEV_REGISTER:
2498 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2499 			idev = ipv6_add_dev(dev);
2500 			if (!idev)
2501 				return notifier_from_errno(-ENOMEM);
2502 		}
2503 		break;
2504 
2505 	case NETDEV_UP:
2506 	case NETDEV_CHANGE:
2507 		if (dev->flags & IFF_SLAVE)
2508 			break;
2509 
2510 		if (event == NETDEV_UP) {
2511 			if (!addrconf_qdisc_ok(dev)) {
2512 				/* device is not ready yet. */
2513 				printk(KERN_INFO
2514 					"ADDRCONF(NETDEV_UP): %s: "
2515 					"link is not ready\n",
2516 					dev->name);
2517 				break;
2518 			}
2519 
2520 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
2521 				idev = ipv6_add_dev(dev);
2522 
2523 			if (idev) {
2524 				idev->if_flags |= IF_READY;
2525 				run_pending = 1;
2526 			}
2527 		} else {
2528 			if (!addrconf_qdisc_ok(dev)) {
2529 				/* device is still not ready. */
2530 				break;
2531 			}
2532 
2533 			if (idev) {
2534 				if (idev->if_flags & IF_READY)
2535 					/* device is already configured. */
2536 					break;
2537 				idev->if_flags |= IF_READY;
2538 			}
2539 
2540 			printk(KERN_INFO
2541 					"ADDRCONF(NETDEV_CHANGE): %s: "
2542 					"link becomes ready\n",
2543 					dev->name);
2544 
2545 			run_pending = 1;
2546 		}
2547 
2548 		switch (dev->type) {
2549 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2550 		case ARPHRD_SIT:
2551 			addrconf_sit_config(dev);
2552 			break;
2553 #endif
2554 		case ARPHRD_TUNNEL6:
2555 			addrconf_ip6_tnl_config(dev);
2556 			break;
2557 		case ARPHRD_LOOPBACK:
2558 			init_loopback(dev);
2559 			break;
2560 
2561 		default:
2562 			addrconf_dev_config(dev);
2563 			break;
2564 		}
2565 
2566 		if (idev) {
2567 			if (run_pending)
2568 				addrconf_dad_run(idev);
2569 
2570 			/*
2571 			 * If the MTU changed during the interface down,
2572 			 * when the interface up, the changed MTU must be
2573 			 * reflected in the idev as well as routers.
2574 			 */
2575 			if (idev->cnf.mtu6 != dev->mtu &&
2576 			    dev->mtu >= IPV6_MIN_MTU) {
2577 				rt6_mtu_change(dev, dev->mtu);
2578 				idev->cnf.mtu6 = dev->mtu;
2579 			}
2580 			idev->tstamp = jiffies;
2581 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2582 
2583 			/*
2584 			 * If the changed mtu during down is lower than
2585 			 * IPV6_MIN_MTU stop IPv6 on this interface.
2586 			 */
2587 			if (dev->mtu < IPV6_MIN_MTU)
2588 				addrconf_ifdown(dev, 1);
2589 		}
2590 		break;
2591 
2592 	case NETDEV_CHANGEMTU:
2593 		if (idev && dev->mtu >= IPV6_MIN_MTU) {
2594 			rt6_mtu_change(dev, dev->mtu);
2595 			idev->cnf.mtu6 = dev->mtu;
2596 			break;
2597 		}
2598 
2599 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2600 			idev = ipv6_add_dev(dev);
2601 			if (idev)
2602 				break;
2603 		}
2604 
2605 		/*
2606 		 * MTU falled under IPV6_MIN_MTU.
2607 		 * Stop IPv6 on this interface.
2608 		 */
2609 
2610 	case NETDEV_DOWN:
2611 	case NETDEV_UNREGISTER:
2612 		/*
2613 		 *	Remove all addresses from this interface.
2614 		 */
2615 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2616 		break;
2617 
2618 	case NETDEV_CHANGENAME:
2619 		if (idev) {
2620 			snmp6_unregister_dev(idev);
2621 			addrconf_sysctl_unregister(idev);
2622 			addrconf_sysctl_register(idev);
2623 			err = snmp6_register_dev(idev);
2624 			if (err)
2625 				return notifier_from_errno(err);
2626 		}
2627 		break;
2628 
2629 	case NETDEV_PRE_TYPE_CHANGE:
2630 	case NETDEV_POST_TYPE_CHANGE:
2631 		addrconf_type_change(dev, event);
2632 		break;
2633 	}
2634 
2635 	return NOTIFY_OK;
2636 }
2637 
2638 /*
2639  *	addrconf module should be notified of a device going up
2640  */
2641 static struct notifier_block ipv6_dev_notf = {
2642 	.notifier_call = addrconf_notify,
2643 };
2644 
2645 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2646 {
2647 	struct inet6_dev *idev;
2648 	ASSERT_RTNL();
2649 
2650 	idev = __in6_dev_get(dev);
2651 
2652 	if (event == NETDEV_POST_TYPE_CHANGE)
2653 		ipv6_mc_remap(idev);
2654 	else if (event == NETDEV_PRE_TYPE_CHANGE)
2655 		ipv6_mc_unmap(idev);
2656 }
2657 
2658 static int addrconf_ifdown(struct net_device *dev, int how)
2659 {
2660 	struct net *net = dev_net(dev);
2661 	struct inet6_dev *idev;
2662 	struct inet6_ifaddr *ifa;
2663 	LIST_HEAD(keep_list);
2664 	int state;
2665 
2666 	ASSERT_RTNL();
2667 
2668 	rt6_ifdown(net, dev);
2669 	neigh_ifdown(&nd_tbl, dev);
2670 
2671 	idev = __in6_dev_get(dev);
2672 	if (idev == NULL)
2673 		return -ENODEV;
2674 
2675 	/*
2676 	 * Step 1: remove reference to ipv6 device from parent device.
2677 	 *	   Do not dev_put!
2678 	 */
2679 	if (how) {
2680 		idev->dead = 1;
2681 
2682 		/* protected by rtnl_lock */
2683 		rcu_assign_pointer(dev->ip6_ptr, NULL);
2684 
2685 		/* Step 1.5: remove snmp6 entry */
2686 		snmp6_unregister_dev(idev);
2687 
2688 	}
2689 
2690 	write_lock_bh(&idev->lock);
2691 
2692 	/* Step 2: clear flags for stateless addrconf */
2693 	if (!how)
2694 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2695 
2696 #ifdef CONFIG_IPV6_PRIVACY
2697 	if (how && del_timer(&idev->regen_timer))
2698 		in6_dev_put(idev);
2699 
2700 	/* Step 3: clear tempaddr list */
2701 	while (!list_empty(&idev->tempaddr_list)) {
2702 		ifa = list_first_entry(&idev->tempaddr_list,
2703 				       struct inet6_ifaddr, tmp_list);
2704 		list_del(&ifa->tmp_list);
2705 		write_unlock_bh(&idev->lock);
2706 		spin_lock_bh(&ifa->lock);
2707 
2708 		if (ifa->ifpub) {
2709 			in6_ifa_put(ifa->ifpub);
2710 			ifa->ifpub = NULL;
2711 		}
2712 		spin_unlock_bh(&ifa->lock);
2713 		in6_ifa_put(ifa);
2714 		write_lock_bh(&idev->lock);
2715 	}
2716 #endif
2717 
2718 	while (!list_empty(&idev->addr_list)) {
2719 		ifa = list_first_entry(&idev->addr_list,
2720 				       struct inet6_ifaddr, if_list);
2721 		addrconf_del_timer(ifa);
2722 
2723 		/* If just doing link down, and address is permanent
2724 		   and not link-local, then retain it. */
2725 		if (!how &&
2726 		    (ifa->flags&IFA_F_PERMANENT) &&
2727 		    !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2728 			list_move_tail(&ifa->if_list, &keep_list);
2729 
2730 			/* If not doing DAD on this address, just keep it. */
2731 			if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2732 			    idev->cnf.accept_dad <= 0 ||
2733 			    (ifa->flags & IFA_F_NODAD))
2734 				continue;
2735 
2736 			/* If it was tentative already, no need to notify */
2737 			if (ifa->flags & IFA_F_TENTATIVE)
2738 				continue;
2739 
2740 			/* Flag it for later restoration when link comes up */
2741 			ifa->flags |= IFA_F_TENTATIVE;
2742 			ifa->state = INET6_IFADDR_STATE_DAD;
2743 
2744 			write_unlock_bh(&idev->lock);
2745 
2746 			in6_ifa_hold(ifa);
2747 		} else {
2748 			list_del(&ifa->if_list);
2749 
2750 			/* clear hash table */
2751 			spin_lock_bh(&addrconf_hash_lock);
2752 			hlist_del_init_rcu(&ifa->addr_lst);
2753 			spin_unlock_bh(&addrconf_hash_lock);
2754 
2755 			write_unlock_bh(&idev->lock);
2756 			spin_lock_bh(&ifa->state_lock);
2757 			state = ifa->state;
2758 			ifa->state = INET6_IFADDR_STATE_DEAD;
2759 			spin_unlock_bh(&ifa->state_lock);
2760 
2761 			if (state == INET6_IFADDR_STATE_DEAD)
2762 				goto put_ifa;
2763 		}
2764 
2765 		__ipv6_ifa_notify(RTM_DELADDR, ifa);
2766 		if (ifa->state == INET6_IFADDR_STATE_DEAD)
2767 			atomic_notifier_call_chain(&inet6addr_chain,
2768 						   NETDEV_DOWN, ifa);
2769 
2770 put_ifa:
2771 		in6_ifa_put(ifa);
2772 
2773 		write_lock_bh(&idev->lock);
2774 	}
2775 
2776 	list_splice(&keep_list, &idev->addr_list);
2777 
2778 	write_unlock_bh(&idev->lock);
2779 
2780 	/* Step 5: Discard multicast list */
2781 	if (how)
2782 		ipv6_mc_destroy_dev(idev);
2783 	else
2784 		ipv6_mc_down(idev);
2785 
2786 	idev->tstamp = jiffies;
2787 
2788 	/* Last: Shot the device (if unregistered) */
2789 	if (how) {
2790 		addrconf_sysctl_unregister(idev);
2791 		neigh_parms_release(&nd_tbl, idev->nd_parms);
2792 		neigh_ifdown(&nd_tbl, dev);
2793 		in6_dev_put(idev);
2794 	}
2795 	return 0;
2796 }
2797 
2798 static void addrconf_rs_timer(unsigned long data)
2799 {
2800 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2801 	struct inet6_dev *idev = ifp->idev;
2802 
2803 	read_lock(&idev->lock);
2804 	if (idev->dead || !(idev->if_flags & IF_READY))
2805 		goto out;
2806 
2807 	if (idev->cnf.forwarding)
2808 		goto out;
2809 
2810 	/* Announcement received after solicitation was sent */
2811 	if (idev->if_flags & IF_RA_RCVD)
2812 		goto out;
2813 
2814 	spin_lock(&ifp->lock);
2815 	if (ifp->probes++ < idev->cnf.rtr_solicits) {
2816 		/* The wait after the last probe can be shorter */
2817 		addrconf_mod_timer(ifp, AC_RS,
2818 				   (ifp->probes == idev->cnf.rtr_solicits) ?
2819 				   idev->cnf.rtr_solicit_delay :
2820 				   idev->cnf.rtr_solicit_interval);
2821 		spin_unlock(&ifp->lock);
2822 
2823 		ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2824 	} else {
2825 		spin_unlock(&ifp->lock);
2826 		/*
2827 		 * Note: we do not support deprecated "all on-link"
2828 		 * assumption any longer.
2829 		 */
2830 		printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2831 		       idev->dev->name);
2832 	}
2833 
2834 out:
2835 	read_unlock(&idev->lock);
2836 	in6_ifa_put(ifp);
2837 }
2838 
2839 /*
2840  *	Duplicate Address Detection
2841  */
2842 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2843 {
2844 	unsigned long rand_num;
2845 	struct inet6_dev *idev = ifp->idev;
2846 
2847 	if (ifp->flags & IFA_F_OPTIMISTIC)
2848 		rand_num = 0;
2849 	else
2850 		rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2851 
2852 	ifp->probes = idev->cnf.dad_transmits;
2853 	addrconf_mod_timer(ifp, AC_DAD, rand_num);
2854 }
2855 
2856 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2857 {
2858 	struct inet6_dev *idev = ifp->idev;
2859 	struct net_device *dev = idev->dev;
2860 
2861 	addrconf_join_solict(dev, &ifp->addr);
2862 
2863 	net_srandom(ifp->addr.s6_addr32[3]);
2864 
2865 	read_lock_bh(&idev->lock);
2866 	spin_lock(&ifp->lock);
2867 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
2868 		goto out;
2869 
2870 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2871 	    idev->cnf.accept_dad < 1 ||
2872 	    !(ifp->flags&IFA_F_TENTATIVE) ||
2873 	    ifp->flags & IFA_F_NODAD) {
2874 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2875 		spin_unlock(&ifp->lock);
2876 		read_unlock_bh(&idev->lock);
2877 
2878 		addrconf_dad_completed(ifp);
2879 		return;
2880 	}
2881 
2882 	if (!(idev->if_flags & IF_READY)) {
2883 		spin_unlock(&ifp->lock);
2884 		read_unlock_bh(&idev->lock);
2885 		/*
2886 		 * If the device is not ready:
2887 		 * - keep it tentative if it is a permanent address.
2888 		 * - otherwise, kill it.
2889 		 */
2890 		in6_ifa_hold(ifp);
2891 		addrconf_dad_stop(ifp, 0);
2892 		return;
2893 	}
2894 
2895 	/*
2896 	 * Optimistic nodes can start receiving
2897 	 * Frames right away
2898 	 */
2899 	if (ifp->flags & IFA_F_OPTIMISTIC)
2900 		ip6_ins_rt(ifp->rt);
2901 
2902 	addrconf_dad_kick(ifp);
2903 out:
2904 	spin_unlock(&ifp->lock);
2905 	read_unlock_bh(&idev->lock);
2906 }
2907 
2908 static void addrconf_dad_timer(unsigned long data)
2909 {
2910 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2911 	struct inet6_dev *idev = ifp->idev;
2912 	struct in6_addr mcaddr;
2913 
2914 	if (!ifp->probes && addrconf_dad_end(ifp))
2915 		goto out;
2916 
2917 	read_lock(&idev->lock);
2918 	if (idev->dead || !(idev->if_flags & IF_READY)) {
2919 		read_unlock(&idev->lock);
2920 		goto out;
2921 	}
2922 
2923 	spin_lock(&ifp->lock);
2924 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2925 		spin_unlock(&ifp->lock);
2926 		read_unlock(&idev->lock);
2927 		goto out;
2928 	}
2929 
2930 	if (ifp->probes == 0) {
2931 		/*
2932 		 * DAD was successful
2933 		 */
2934 
2935 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2936 		spin_unlock(&ifp->lock);
2937 		read_unlock(&idev->lock);
2938 
2939 		addrconf_dad_completed(ifp);
2940 
2941 		goto out;
2942 	}
2943 
2944 	ifp->probes--;
2945 	addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2946 	spin_unlock(&ifp->lock);
2947 	read_unlock(&idev->lock);
2948 
2949 	/* send a neighbour solicitation for our addr */
2950 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2951 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2952 out:
2953 	in6_ifa_put(ifp);
2954 }
2955 
2956 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2957 {
2958 	struct net_device *dev = ifp->idev->dev;
2959 
2960 	/*
2961 	 *	Configure the address for reception. Now it is valid.
2962 	 */
2963 
2964 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
2965 
2966 	/* If added prefix is link local and forwarding is off,
2967 	   start sending router solicitations.
2968 	 */
2969 
2970 	if ((ifp->idev->cnf.forwarding == 0 ||
2971 	     ifp->idev->cnf.forwarding == 2) &&
2972 	    ifp->idev->cnf.rtr_solicits > 0 &&
2973 	    (dev->flags&IFF_LOOPBACK) == 0 &&
2974 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2975 		/*
2976 		 *	If a host as already performed a random delay
2977 		 *	[...] as part of DAD [...] there is no need
2978 		 *	to delay again before sending the first RS
2979 		 */
2980 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2981 
2982 		spin_lock_bh(&ifp->lock);
2983 		ifp->probes = 1;
2984 		ifp->idev->if_flags |= IF_RS_SENT;
2985 		addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2986 		spin_unlock_bh(&ifp->lock);
2987 	}
2988 }
2989 
2990 static void addrconf_dad_run(struct inet6_dev *idev)
2991 {
2992 	struct inet6_ifaddr *ifp;
2993 
2994 	read_lock_bh(&idev->lock);
2995 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2996 		spin_lock(&ifp->lock);
2997 		if (ifp->flags & IFA_F_TENTATIVE &&
2998 		    ifp->state == INET6_IFADDR_STATE_DAD)
2999 			addrconf_dad_kick(ifp);
3000 		spin_unlock(&ifp->lock);
3001 	}
3002 	read_unlock_bh(&idev->lock);
3003 }
3004 
3005 #ifdef CONFIG_PROC_FS
3006 struct if6_iter_state {
3007 	struct seq_net_private p;
3008 	int bucket;
3009 };
3010 
3011 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3012 {
3013 	struct inet6_ifaddr *ifa = NULL;
3014 	struct if6_iter_state *state = seq->private;
3015 	struct net *net = seq_file_net(seq);
3016 
3017 	for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3018 		struct hlist_node *n;
3019 		hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3020 					 addr_lst)
3021 			if (net_eq(dev_net(ifa->idev->dev), net))
3022 				return ifa;
3023 	}
3024 	return NULL;
3025 }
3026 
3027 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3028 					 struct inet6_ifaddr *ifa)
3029 {
3030 	struct if6_iter_state *state = seq->private;
3031 	struct net *net = seq_file_net(seq);
3032 	struct hlist_node *n = &ifa->addr_lst;
3033 
3034 	hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3035 		if (net_eq(dev_net(ifa->idev->dev), net))
3036 			return ifa;
3037 
3038 	while (++state->bucket < IN6_ADDR_HSIZE) {
3039 		hlist_for_each_entry_rcu_bh(ifa, n,
3040 				     &inet6_addr_lst[state->bucket], addr_lst) {
3041 			if (net_eq(dev_net(ifa->idev->dev), net))
3042 				return ifa;
3043 		}
3044 	}
3045 
3046 	return NULL;
3047 }
3048 
3049 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3050 {
3051 	struct inet6_ifaddr *ifa = if6_get_first(seq);
3052 
3053 	if (ifa)
3054 		while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3055 			--pos;
3056 	return pos ? NULL : ifa;
3057 }
3058 
3059 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3060 	__acquires(rcu_bh)
3061 {
3062 	rcu_read_lock_bh();
3063 	return if6_get_idx(seq, *pos);
3064 }
3065 
3066 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3067 {
3068 	struct inet6_ifaddr *ifa;
3069 
3070 	ifa = if6_get_next(seq, v);
3071 	++*pos;
3072 	return ifa;
3073 }
3074 
3075 static void if6_seq_stop(struct seq_file *seq, void *v)
3076 	__releases(rcu_bh)
3077 {
3078 	rcu_read_unlock_bh();
3079 }
3080 
3081 static int if6_seq_show(struct seq_file *seq, void *v)
3082 {
3083 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3084 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3085 		   &ifp->addr,
3086 		   ifp->idev->dev->ifindex,
3087 		   ifp->prefix_len,
3088 		   ifp->scope,
3089 		   ifp->flags,
3090 		   ifp->idev->dev->name);
3091 	return 0;
3092 }
3093 
3094 static const struct seq_operations if6_seq_ops = {
3095 	.start	= if6_seq_start,
3096 	.next	= if6_seq_next,
3097 	.show	= if6_seq_show,
3098 	.stop	= if6_seq_stop,
3099 };
3100 
3101 static int if6_seq_open(struct inode *inode, struct file *file)
3102 {
3103 	return seq_open_net(inode, file, &if6_seq_ops,
3104 			    sizeof(struct if6_iter_state));
3105 }
3106 
3107 static const struct file_operations if6_fops = {
3108 	.owner		= THIS_MODULE,
3109 	.open		= if6_seq_open,
3110 	.read		= seq_read,
3111 	.llseek		= seq_lseek,
3112 	.release	= seq_release_net,
3113 };
3114 
3115 static int __net_init if6_proc_net_init(struct net *net)
3116 {
3117 	if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3118 		return -ENOMEM;
3119 	return 0;
3120 }
3121 
3122 static void __net_exit if6_proc_net_exit(struct net *net)
3123 {
3124        proc_net_remove(net, "if_inet6");
3125 }
3126 
3127 static struct pernet_operations if6_proc_net_ops = {
3128        .init = if6_proc_net_init,
3129        .exit = if6_proc_net_exit,
3130 };
3131 
3132 int __init if6_proc_init(void)
3133 {
3134 	return register_pernet_subsys(&if6_proc_net_ops);
3135 }
3136 
3137 void if6_proc_exit(void)
3138 {
3139 	unregister_pernet_subsys(&if6_proc_net_ops);
3140 }
3141 #endif	/* CONFIG_PROC_FS */
3142 
3143 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3144 /* Check if address is a home address configured on any interface. */
3145 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3146 {
3147 	int ret = 0;
3148 	struct inet6_ifaddr *ifp = NULL;
3149 	struct hlist_node *n;
3150 	unsigned int hash = ipv6_addr_hash(addr);
3151 
3152 	rcu_read_lock_bh();
3153 	hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3154 		if (!net_eq(dev_net(ifp->idev->dev), net))
3155 			continue;
3156 		if (ipv6_addr_equal(&ifp->addr, addr) &&
3157 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3158 			ret = 1;
3159 			break;
3160 		}
3161 	}
3162 	rcu_read_unlock_bh();
3163 	return ret;
3164 }
3165 #endif
3166 
3167 /*
3168  *	Periodic address status verification
3169  */
3170 
3171 static void addrconf_verify(unsigned long foo)
3172 {
3173 	unsigned long now, next, next_sec, next_sched;
3174 	struct inet6_ifaddr *ifp;
3175 	struct hlist_node *node;
3176 	int i;
3177 
3178 	rcu_read_lock_bh();
3179 	spin_lock(&addrconf_verify_lock);
3180 	now = jiffies;
3181 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3182 
3183 	del_timer(&addr_chk_timer);
3184 
3185 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3186 restart:
3187 		hlist_for_each_entry_rcu_bh(ifp, node,
3188 					 &inet6_addr_lst[i], addr_lst) {
3189 			unsigned long age;
3190 
3191 			if (ifp->flags & IFA_F_PERMANENT)
3192 				continue;
3193 
3194 			spin_lock(&ifp->lock);
3195 			/* We try to batch several events at once. */
3196 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3197 
3198 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3199 			    age >= ifp->valid_lft) {
3200 				spin_unlock(&ifp->lock);
3201 				in6_ifa_hold(ifp);
3202 				ipv6_del_addr(ifp);
3203 				goto restart;
3204 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3205 				spin_unlock(&ifp->lock);
3206 				continue;
3207 			} else if (age >= ifp->prefered_lft) {
3208 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3209 				int deprecate = 0;
3210 
3211 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3212 					deprecate = 1;
3213 					ifp->flags |= IFA_F_DEPRECATED;
3214 				}
3215 
3216 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3217 					next = ifp->tstamp + ifp->valid_lft * HZ;
3218 
3219 				spin_unlock(&ifp->lock);
3220 
3221 				if (deprecate) {
3222 					in6_ifa_hold(ifp);
3223 
3224 					ipv6_ifa_notify(0, ifp);
3225 					in6_ifa_put(ifp);
3226 					goto restart;
3227 				}
3228 #ifdef CONFIG_IPV6_PRIVACY
3229 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3230 				   !(ifp->flags&IFA_F_TENTATIVE)) {
3231 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3232 					ifp->idev->cnf.dad_transmits *
3233 					ifp->idev->nd_parms->retrans_time / HZ;
3234 
3235 				if (age >= ifp->prefered_lft - regen_advance) {
3236 					struct inet6_ifaddr *ifpub = ifp->ifpub;
3237 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3238 						next = ifp->tstamp + ifp->prefered_lft * HZ;
3239 					if (!ifp->regen_count && ifpub) {
3240 						ifp->regen_count++;
3241 						in6_ifa_hold(ifp);
3242 						in6_ifa_hold(ifpub);
3243 						spin_unlock(&ifp->lock);
3244 
3245 						spin_lock(&ifpub->lock);
3246 						ifpub->regen_count = 0;
3247 						spin_unlock(&ifpub->lock);
3248 						ipv6_create_tempaddr(ifpub, ifp);
3249 						in6_ifa_put(ifpub);
3250 						in6_ifa_put(ifp);
3251 						goto restart;
3252 					}
3253 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3254 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3255 				spin_unlock(&ifp->lock);
3256 #endif
3257 			} else {
3258 				/* ifp->prefered_lft <= ifp->valid_lft */
3259 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3260 					next = ifp->tstamp + ifp->prefered_lft * HZ;
3261 				spin_unlock(&ifp->lock);
3262 			}
3263 		}
3264 	}
3265 
3266 	next_sec = round_jiffies_up(next);
3267 	next_sched = next;
3268 
3269 	/* If rounded timeout is accurate enough, accept it. */
3270 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3271 		next_sched = next_sec;
3272 
3273 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3274 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3275 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3276 
3277 	ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3278 	      now, next, next_sec, next_sched));
3279 
3280 	addr_chk_timer.expires = next_sched;
3281 	add_timer(&addr_chk_timer);
3282 	spin_unlock(&addrconf_verify_lock);
3283 	rcu_read_unlock_bh();
3284 }
3285 
3286 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3287 {
3288 	struct in6_addr *pfx = NULL;
3289 
3290 	if (addr)
3291 		pfx = nla_data(addr);
3292 
3293 	if (local) {
3294 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3295 			pfx = NULL;
3296 		else
3297 			pfx = nla_data(local);
3298 	}
3299 
3300 	return pfx;
3301 }
3302 
3303 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3304 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
3305 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
3306 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
3307 };
3308 
3309 static int
3310 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3311 {
3312 	struct net *net = sock_net(skb->sk);
3313 	struct ifaddrmsg *ifm;
3314 	struct nlattr *tb[IFA_MAX+1];
3315 	struct in6_addr *pfx;
3316 	int err;
3317 
3318 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3319 	if (err < 0)
3320 		return err;
3321 
3322 	ifm = nlmsg_data(nlh);
3323 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3324 	if (pfx == NULL)
3325 		return -EINVAL;
3326 
3327 	return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3328 }
3329 
3330 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3331 			     u32 prefered_lft, u32 valid_lft)
3332 {
3333 	u32 flags;
3334 	clock_t expires;
3335 	unsigned long timeout;
3336 
3337 	if (!valid_lft || (prefered_lft > valid_lft))
3338 		return -EINVAL;
3339 
3340 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
3341 	if (addrconf_finite_timeout(timeout)) {
3342 		expires = jiffies_to_clock_t(timeout * HZ);
3343 		valid_lft = timeout;
3344 		flags = RTF_EXPIRES;
3345 	} else {
3346 		expires = 0;
3347 		flags = 0;
3348 		ifa_flags |= IFA_F_PERMANENT;
3349 	}
3350 
3351 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3352 	if (addrconf_finite_timeout(timeout)) {
3353 		if (timeout == 0)
3354 			ifa_flags |= IFA_F_DEPRECATED;
3355 		prefered_lft = timeout;
3356 	}
3357 
3358 	spin_lock_bh(&ifp->lock);
3359 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3360 	ifp->tstamp = jiffies;
3361 	ifp->valid_lft = valid_lft;
3362 	ifp->prefered_lft = prefered_lft;
3363 
3364 	spin_unlock_bh(&ifp->lock);
3365 	if (!(ifp->flags&IFA_F_TENTATIVE))
3366 		ipv6_ifa_notify(0, ifp);
3367 
3368 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3369 			      expires, flags);
3370 	addrconf_verify(0);
3371 
3372 	return 0;
3373 }
3374 
3375 static int
3376 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3377 {
3378 	struct net *net = sock_net(skb->sk);
3379 	struct ifaddrmsg *ifm;
3380 	struct nlattr *tb[IFA_MAX+1];
3381 	struct in6_addr *pfx;
3382 	struct inet6_ifaddr *ifa;
3383 	struct net_device *dev;
3384 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3385 	u8 ifa_flags;
3386 	int err;
3387 
3388 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3389 	if (err < 0)
3390 		return err;
3391 
3392 	ifm = nlmsg_data(nlh);
3393 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3394 	if (pfx == NULL)
3395 		return -EINVAL;
3396 
3397 	if (tb[IFA_CACHEINFO]) {
3398 		struct ifa_cacheinfo *ci;
3399 
3400 		ci = nla_data(tb[IFA_CACHEINFO]);
3401 		valid_lft = ci->ifa_valid;
3402 		preferred_lft = ci->ifa_prefered;
3403 	} else {
3404 		preferred_lft = INFINITY_LIFE_TIME;
3405 		valid_lft = INFINITY_LIFE_TIME;
3406 	}
3407 
3408 	dev =  __dev_get_by_index(net, ifm->ifa_index);
3409 	if (dev == NULL)
3410 		return -ENODEV;
3411 
3412 	/* We ignore other flags so far. */
3413 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3414 
3415 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3416 	if (ifa == NULL) {
3417 		/*
3418 		 * It would be best to check for !NLM_F_CREATE here but
3419 		 * userspace alreay relies on not having to provide this.
3420 		 */
3421 		return inet6_addr_add(net, ifm->ifa_index, pfx,
3422 				      ifm->ifa_prefixlen, ifa_flags,
3423 				      preferred_lft, valid_lft);
3424 	}
3425 
3426 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3427 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3428 		err = -EEXIST;
3429 	else
3430 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3431 
3432 	in6_ifa_put(ifa);
3433 
3434 	return err;
3435 }
3436 
3437 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3438 			  u8 scope, int ifindex)
3439 {
3440 	struct ifaddrmsg *ifm;
3441 
3442 	ifm = nlmsg_data(nlh);
3443 	ifm->ifa_family = AF_INET6;
3444 	ifm->ifa_prefixlen = prefixlen;
3445 	ifm->ifa_flags = flags;
3446 	ifm->ifa_scope = scope;
3447 	ifm->ifa_index = ifindex;
3448 }
3449 
3450 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3451 			 unsigned long tstamp, u32 preferred, u32 valid)
3452 {
3453 	struct ifa_cacheinfo ci;
3454 
3455 	ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3456 			+ TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3457 	ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3458 			+ TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3459 	ci.ifa_prefered = preferred;
3460 	ci.ifa_valid = valid;
3461 
3462 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3463 }
3464 
3465 static inline int rt_scope(int ifa_scope)
3466 {
3467 	if (ifa_scope & IFA_HOST)
3468 		return RT_SCOPE_HOST;
3469 	else if (ifa_scope & IFA_LINK)
3470 		return RT_SCOPE_LINK;
3471 	else if (ifa_scope & IFA_SITE)
3472 		return RT_SCOPE_SITE;
3473 	else
3474 		return RT_SCOPE_UNIVERSE;
3475 }
3476 
3477 static inline int inet6_ifaddr_msgsize(void)
3478 {
3479 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3480 	       + nla_total_size(16) /* IFA_ADDRESS */
3481 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3482 }
3483 
3484 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3485 			     u32 pid, u32 seq, int event, unsigned int flags)
3486 {
3487 	struct nlmsghdr  *nlh;
3488 	u32 preferred, valid;
3489 
3490 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3491 	if (nlh == NULL)
3492 		return -EMSGSIZE;
3493 
3494 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3495 		      ifa->idev->dev->ifindex);
3496 
3497 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3498 		preferred = ifa->prefered_lft;
3499 		valid = ifa->valid_lft;
3500 		if (preferred != INFINITY_LIFE_TIME) {
3501 			long tval = (jiffies - ifa->tstamp)/HZ;
3502 			if (preferred > tval)
3503 				preferred -= tval;
3504 			else
3505 				preferred = 0;
3506 			if (valid != INFINITY_LIFE_TIME) {
3507 				if (valid > tval)
3508 					valid -= tval;
3509 				else
3510 					valid = 0;
3511 			}
3512 		}
3513 	} else {
3514 		preferred = INFINITY_LIFE_TIME;
3515 		valid = INFINITY_LIFE_TIME;
3516 	}
3517 
3518 	if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3519 	    put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3520 		nlmsg_cancel(skb, nlh);
3521 		return -EMSGSIZE;
3522 	}
3523 
3524 	return nlmsg_end(skb, nlh);
3525 }
3526 
3527 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3528 				u32 pid, u32 seq, int event, u16 flags)
3529 {
3530 	struct nlmsghdr  *nlh;
3531 	u8 scope = RT_SCOPE_UNIVERSE;
3532 	int ifindex = ifmca->idev->dev->ifindex;
3533 
3534 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3535 		scope = RT_SCOPE_SITE;
3536 
3537 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3538 	if (nlh == NULL)
3539 		return -EMSGSIZE;
3540 
3541 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3542 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3543 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3544 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3545 		nlmsg_cancel(skb, nlh);
3546 		return -EMSGSIZE;
3547 	}
3548 
3549 	return nlmsg_end(skb, nlh);
3550 }
3551 
3552 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3553 				u32 pid, u32 seq, int event, unsigned int flags)
3554 {
3555 	struct nlmsghdr  *nlh;
3556 	u8 scope = RT_SCOPE_UNIVERSE;
3557 	int ifindex = ifaca->aca_idev->dev->ifindex;
3558 
3559 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3560 		scope = RT_SCOPE_SITE;
3561 
3562 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3563 	if (nlh == NULL)
3564 		return -EMSGSIZE;
3565 
3566 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3567 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3568 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3569 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3570 		nlmsg_cancel(skb, nlh);
3571 		return -EMSGSIZE;
3572 	}
3573 
3574 	return nlmsg_end(skb, nlh);
3575 }
3576 
3577 enum addr_type_t {
3578 	UNICAST_ADDR,
3579 	MULTICAST_ADDR,
3580 	ANYCAST_ADDR,
3581 };
3582 
3583 /* called with rcu_read_lock() */
3584 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3585 			  struct netlink_callback *cb, enum addr_type_t type,
3586 			  int s_ip_idx, int *p_ip_idx)
3587 {
3588 	struct ifmcaddr6 *ifmca;
3589 	struct ifacaddr6 *ifaca;
3590 	int err = 1;
3591 	int ip_idx = *p_ip_idx;
3592 
3593 	read_lock_bh(&idev->lock);
3594 	switch (type) {
3595 	case UNICAST_ADDR: {
3596 		struct inet6_ifaddr *ifa;
3597 
3598 		/* unicast address incl. temp addr */
3599 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
3600 			if (++ip_idx < s_ip_idx)
3601 				continue;
3602 			err = inet6_fill_ifaddr(skb, ifa,
3603 						NETLINK_CB(cb->skb).pid,
3604 						cb->nlh->nlmsg_seq,
3605 						RTM_NEWADDR,
3606 						NLM_F_MULTI);
3607 			if (err <= 0)
3608 				break;
3609 		}
3610 		break;
3611 	}
3612 	case MULTICAST_ADDR:
3613 		/* multicast address */
3614 		for (ifmca = idev->mc_list; ifmca;
3615 		     ifmca = ifmca->next, ip_idx++) {
3616 			if (ip_idx < s_ip_idx)
3617 				continue;
3618 			err = inet6_fill_ifmcaddr(skb, ifmca,
3619 						  NETLINK_CB(cb->skb).pid,
3620 						  cb->nlh->nlmsg_seq,
3621 						  RTM_GETMULTICAST,
3622 						  NLM_F_MULTI);
3623 			if (err <= 0)
3624 				break;
3625 		}
3626 		break;
3627 	case ANYCAST_ADDR:
3628 		/* anycast address */
3629 		for (ifaca = idev->ac_list; ifaca;
3630 		     ifaca = ifaca->aca_next, ip_idx++) {
3631 			if (ip_idx < s_ip_idx)
3632 				continue;
3633 			err = inet6_fill_ifacaddr(skb, ifaca,
3634 						  NETLINK_CB(cb->skb).pid,
3635 						  cb->nlh->nlmsg_seq,
3636 						  RTM_GETANYCAST,
3637 						  NLM_F_MULTI);
3638 			if (err <= 0)
3639 				break;
3640 		}
3641 		break;
3642 	default:
3643 		break;
3644 	}
3645 	read_unlock_bh(&idev->lock);
3646 	*p_ip_idx = ip_idx;
3647 	return err;
3648 }
3649 
3650 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3651 			   enum addr_type_t type)
3652 {
3653 	struct net *net = sock_net(skb->sk);
3654 	int h, s_h;
3655 	int idx, ip_idx;
3656 	int s_idx, s_ip_idx;
3657 	struct net_device *dev;
3658 	struct inet6_dev *idev;
3659 	struct hlist_head *head;
3660 	struct hlist_node *node;
3661 
3662 	s_h = cb->args[0];
3663 	s_idx = idx = cb->args[1];
3664 	s_ip_idx = ip_idx = cb->args[2];
3665 
3666 	rcu_read_lock();
3667 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3668 		idx = 0;
3669 		head = &net->dev_index_head[h];
3670 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3671 			if (idx < s_idx)
3672 				goto cont;
3673 			if (h > s_h || idx > s_idx)
3674 				s_ip_idx = 0;
3675 			ip_idx = 0;
3676 			idev = __in6_dev_get(dev);
3677 			if (!idev)
3678 				goto cont;
3679 
3680 			if (in6_dump_addrs(idev, skb, cb, type,
3681 					   s_ip_idx, &ip_idx) <= 0)
3682 				goto done;
3683 cont:
3684 			idx++;
3685 		}
3686 	}
3687 done:
3688 	rcu_read_unlock();
3689 	cb->args[0] = h;
3690 	cb->args[1] = idx;
3691 	cb->args[2] = ip_idx;
3692 
3693 	return skb->len;
3694 }
3695 
3696 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3697 {
3698 	enum addr_type_t type = UNICAST_ADDR;
3699 
3700 	return inet6_dump_addr(skb, cb, type);
3701 }
3702 
3703 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3704 {
3705 	enum addr_type_t type = MULTICAST_ADDR;
3706 
3707 	return inet6_dump_addr(skb, cb, type);
3708 }
3709 
3710 
3711 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3712 {
3713 	enum addr_type_t type = ANYCAST_ADDR;
3714 
3715 	return inet6_dump_addr(skb, cb, type);
3716 }
3717 
3718 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3719 			     void *arg)
3720 {
3721 	struct net *net = sock_net(in_skb->sk);
3722 	struct ifaddrmsg *ifm;
3723 	struct nlattr *tb[IFA_MAX+1];
3724 	struct in6_addr *addr = NULL;
3725 	struct net_device *dev = NULL;
3726 	struct inet6_ifaddr *ifa;
3727 	struct sk_buff *skb;
3728 	int err;
3729 
3730 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3731 	if (err < 0)
3732 		goto errout;
3733 
3734 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3735 	if (addr == NULL) {
3736 		err = -EINVAL;
3737 		goto errout;
3738 	}
3739 
3740 	ifm = nlmsg_data(nlh);
3741 	if (ifm->ifa_index)
3742 		dev = __dev_get_by_index(net, ifm->ifa_index);
3743 
3744 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3745 	if (!ifa) {
3746 		err = -EADDRNOTAVAIL;
3747 		goto errout;
3748 	}
3749 
3750 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3751 	if (!skb) {
3752 		err = -ENOBUFS;
3753 		goto errout_ifa;
3754 	}
3755 
3756 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3757 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
3758 	if (err < 0) {
3759 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3760 		WARN_ON(err == -EMSGSIZE);
3761 		kfree_skb(skb);
3762 		goto errout_ifa;
3763 	}
3764 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3765 errout_ifa:
3766 	in6_ifa_put(ifa);
3767 errout:
3768 	return err;
3769 }
3770 
3771 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3772 {
3773 	struct sk_buff *skb;
3774 	struct net *net = dev_net(ifa->idev->dev);
3775 	int err = -ENOBUFS;
3776 
3777 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3778 	if (skb == NULL)
3779 		goto errout;
3780 
3781 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3782 	if (err < 0) {
3783 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3784 		WARN_ON(err == -EMSGSIZE);
3785 		kfree_skb(skb);
3786 		goto errout;
3787 	}
3788 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3789 	return;
3790 errout:
3791 	if (err < 0)
3792 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3793 }
3794 
3795 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3796 				__s32 *array, int bytes)
3797 {
3798 	BUG_ON(bytes < (DEVCONF_MAX * 4));
3799 
3800 	memset(array, 0, bytes);
3801 	array[DEVCONF_FORWARDING] = cnf->forwarding;
3802 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3803 	array[DEVCONF_MTU6] = cnf->mtu6;
3804 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3805 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3806 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
3807 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3808 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3809 	array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3810 	array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3811 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3812 #ifdef CONFIG_IPV6_PRIVACY
3813 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3814 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3815 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3816 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3817 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3818 #endif
3819 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3820 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3821 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3822 #ifdef CONFIG_IPV6_ROUTER_PREF
3823 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3824 	array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3825 #ifdef CONFIG_IPV6_ROUTE_INFO
3826 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3827 #endif
3828 #endif
3829 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3830 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3831 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3832 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3833 #endif
3834 #ifdef CONFIG_IPV6_MROUTE
3835 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3836 #endif
3837 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3838 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3839 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3840 }
3841 
3842 static inline size_t inet6_if_nlmsg_size(void)
3843 {
3844 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3845 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3846 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3847 	       + nla_total_size(4) /* IFLA_MTU */
3848 	       + nla_total_size(4) /* IFLA_LINK */
3849 	       + nla_total_size( /* IFLA_PROTINFO */
3850 			nla_total_size(4) /* IFLA_INET6_FLAGS */
3851 			+ nla_total_size(sizeof(struct ifla_cacheinfo))
3852 			+ nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3853 			+ nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3854 			+ nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3855 		 );
3856 }
3857 
3858 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3859 				      int items, int bytes)
3860 {
3861 	int i;
3862 	int pad = bytes - sizeof(u64) * items;
3863 	BUG_ON(pad < 0);
3864 
3865 	/* Use put_unaligned() because stats may not be aligned for u64. */
3866 	put_unaligned(items, &stats[0]);
3867 	for (i = 1; i < items; i++)
3868 		put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3869 
3870 	memset(&stats[items], 0, pad);
3871 }
3872 
3873 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3874 				      int items, int bytes, size_t syncpoff)
3875 {
3876 	int i;
3877 	int pad = bytes - sizeof(u64) * items;
3878 	BUG_ON(pad < 0);
3879 
3880 	/* Use put_unaligned() because stats may not be aligned for u64. */
3881 	put_unaligned(items, &stats[0]);
3882 	for (i = 1; i < items; i++)
3883 		put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3884 
3885 	memset(&stats[items], 0, pad);
3886 }
3887 
3888 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3889 			     int bytes)
3890 {
3891 	switch (attrtype) {
3892 	case IFLA_INET6_STATS:
3893 		__snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3894 				     IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3895 		break;
3896 	case IFLA_INET6_ICMP6STATS:
3897 		__snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3898 		break;
3899 	}
3900 }
3901 
3902 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3903 			     u32 pid, u32 seq, int event, unsigned int flags)
3904 {
3905 	struct net_device *dev = idev->dev;
3906 	struct nlattr *nla;
3907 	struct ifinfomsg *hdr;
3908 	struct nlmsghdr *nlh;
3909 	void *protoinfo;
3910 	struct ifla_cacheinfo ci;
3911 
3912 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3913 	if (nlh == NULL)
3914 		return -EMSGSIZE;
3915 
3916 	hdr = nlmsg_data(nlh);
3917 	hdr->ifi_family = AF_INET6;
3918 	hdr->__ifi_pad = 0;
3919 	hdr->ifi_type = dev->type;
3920 	hdr->ifi_index = dev->ifindex;
3921 	hdr->ifi_flags = dev_get_flags(dev);
3922 	hdr->ifi_change = 0;
3923 
3924 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3925 
3926 	if (dev->addr_len)
3927 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3928 
3929 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3930 	if (dev->ifindex != dev->iflink)
3931 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3932 
3933 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3934 	if (protoinfo == NULL)
3935 		goto nla_put_failure;
3936 
3937 	NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3938 
3939 	ci.max_reasm_len = IPV6_MAXPLEN;
3940 	ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3941 		    + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3942 	ci.reachable_time = idev->nd_parms->reachable_time;
3943 	ci.retrans_time = idev->nd_parms->retrans_time;
3944 	NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3945 
3946 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3947 	if (nla == NULL)
3948 		goto nla_put_failure;
3949 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3950 
3951 	/* XXX - MC not implemented */
3952 
3953 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3954 	if (nla == NULL)
3955 		goto nla_put_failure;
3956 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3957 
3958 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3959 	if (nla == NULL)
3960 		goto nla_put_failure;
3961 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3962 
3963 	nla_nest_end(skb, protoinfo);
3964 	return nlmsg_end(skb, nlh);
3965 
3966 nla_put_failure:
3967 	nlmsg_cancel(skb, nlh);
3968 	return -EMSGSIZE;
3969 }
3970 
3971 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3972 {
3973 	struct net *net = sock_net(skb->sk);
3974 	int h, s_h;
3975 	int idx = 0, s_idx;
3976 	struct net_device *dev;
3977 	struct inet6_dev *idev;
3978 	struct hlist_head *head;
3979 	struct hlist_node *node;
3980 
3981 	s_h = cb->args[0];
3982 	s_idx = cb->args[1];
3983 
3984 	rcu_read_lock();
3985 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3986 		idx = 0;
3987 		head = &net->dev_index_head[h];
3988 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3989 			if (idx < s_idx)
3990 				goto cont;
3991 			idev = __in6_dev_get(dev);
3992 			if (!idev)
3993 				goto cont;
3994 			if (inet6_fill_ifinfo(skb, idev,
3995 					      NETLINK_CB(cb->skb).pid,
3996 					      cb->nlh->nlmsg_seq,
3997 					      RTM_NEWLINK, NLM_F_MULTI) <= 0)
3998 				goto out;
3999 cont:
4000 			idx++;
4001 		}
4002 	}
4003 out:
4004 	rcu_read_unlock();
4005 	cb->args[1] = idx;
4006 	cb->args[0] = h;
4007 
4008 	return skb->len;
4009 }
4010 
4011 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4012 {
4013 	struct sk_buff *skb;
4014 	struct net *net = dev_net(idev->dev);
4015 	int err = -ENOBUFS;
4016 
4017 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4018 	if (skb == NULL)
4019 		goto errout;
4020 
4021 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4022 	if (err < 0) {
4023 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4024 		WARN_ON(err == -EMSGSIZE);
4025 		kfree_skb(skb);
4026 		goto errout;
4027 	}
4028 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4029 	return;
4030 errout:
4031 	if (err < 0)
4032 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4033 }
4034 
4035 static inline size_t inet6_prefix_nlmsg_size(void)
4036 {
4037 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
4038 	       + nla_total_size(sizeof(struct in6_addr))
4039 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
4040 }
4041 
4042 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4043 			     struct prefix_info *pinfo, u32 pid, u32 seq,
4044 			     int event, unsigned int flags)
4045 {
4046 	struct prefixmsg *pmsg;
4047 	struct nlmsghdr *nlh;
4048 	struct prefix_cacheinfo	ci;
4049 
4050 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4051 	if (nlh == NULL)
4052 		return -EMSGSIZE;
4053 
4054 	pmsg = nlmsg_data(nlh);
4055 	pmsg->prefix_family = AF_INET6;
4056 	pmsg->prefix_pad1 = 0;
4057 	pmsg->prefix_pad2 = 0;
4058 	pmsg->prefix_ifindex = idev->dev->ifindex;
4059 	pmsg->prefix_len = pinfo->prefix_len;
4060 	pmsg->prefix_type = pinfo->type;
4061 	pmsg->prefix_pad3 = 0;
4062 	pmsg->prefix_flags = 0;
4063 	if (pinfo->onlink)
4064 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4065 	if (pinfo->autoconf)
4066 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4067 
4068 	NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4069 
4070 	ci.preferred_time = ntohl(pinfo->prefered);
4071 	ci.valid_time = ntohl(pinfo->valid);
4072 	NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4073 
4074 	return nlmsg_end(skb, nlh);
4075 
4076 nla_put_failure:
4077 	nlmsg_cancel(skb, nlh);
4078 	return -EMSGSIZE;
4079 }
4080 
4081 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4082 			 struct prefix_info *pinfo)
4083 {
4084 	struct sk_buff *skb;
4085 	struct net *net = dev_net(idev->dev);
4086 	int err = -ENOBUFS;
4087 
4088 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4089 	if (skb == NULL)
4090 		goto errout;
4091 
4092 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4093 	if (err < 0) {
4094 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4095 		WARN_ON(err == -EMSGSIZE);
4096 		kfree_skb(skb);
4097 		goto errout;
4098 	}
4099 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4100 	return;
4101 errout:
4102 	if (err < 0)
4103 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4104 }
4105 
4106 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4107 {
4108 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4109 
4110 	switch (event) {
4111 	case RTM_NEWADDR:
4112 		/*
4113 		 * If the address was optimistic
4114 		 * we inserted the route at the start of
4115 		 * our DAD process, so we don't need
4116 		 * to do it again
4117 		 */
4118 		if (!(ifp->rt->rt6i_node))
4119 			ip6_ins_rt(ifp->rt);
4120 		if (ifp->idev->cnf.forwarding)
4121 			addrconf_join_anycast(ifp);
4122 		break;
4123 	case RTM_DELADDR:
4124 		if (ifp->idev->cnf.forwarding)
4125 			addrconf_leave_anycast(ifp);
4126 		addrconf_leave_solict(ifp->idev, &ifp->addr);
4127 		dst_hold(&ifp->rt->dst);
4128 
4129 		if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4130 		    ip6_del_rt(ifp->rt))
4131 			dst_free(&ifp->rt->dst);
4132 		break;
4133 	}
4134 }
4135 
4136 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4137 {
4138 	rcu_read_lock_bh();
4139 	if (likely(ifp->idev->dead == 0))
4140 		__ipv6_ifa_notify(event, ifp);
4141 	rcu_read_unlock_bh();
4142 }
4143 
4144 #ifdef CONFIG_SYSCTL
4145 
4146 static
4147 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4148 			   void __user *buffer, size_t *lenp, loff_t *ppos)
4149 {
4150 	int *valp = ctl->data;
4151 	int val = *valp;
4152 	loff_t pos = *ppos;
4153 	int ret;
4154 
4155 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4156 
4157 	if (write)
4158 		ret = addrconf_fixup_forwarding(ctl, valp, val);
4159 	if (ret)
4160 		*ppos = pos;
4161 	return ret;
4162 }
4163 
4164 static void dev_disable_change(struct inet6_dev *idev)
4165 {
4166 	if (!idev || !idev->dev)
4167 		return;
4168 
4169 	if (idev->cnf.disable_ipv6)
4170 		addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4171 	else
4172 		addrconf_notify(NULL, NETDEV_UP, idev->dev);
4173 }
4174 
4175 static void addrconf_disable_change(struct net *net, __s32 newf)
4176 {
4177 	struct net_device *dev;
4178 	struct inet6_dev *idev;
4179 
4180 	rcu_read_lock();
4181 	for_each_netdev_rcu(net, dev) {
4182 		idev = __in6_dev_get(dev);
4183 		if (idev) {
4184 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4185 			idev->cnf.disable_ipv6 = newf;
4186 			if (changed)
4187 				dev_disable_change(idev);
4188 		}
4189 	}
4190 	rcu_read_unlock();
4191 }
4192 
4193 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4194 {
4195 	struct net *net;
4196 
4197 	net = (struct net *)table->extra2;
4198 
4199 	if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4200 		return 0;
4201 
4202 	if (!rtnl_trylock()) {
4203 		/* Restore the original values before restarting */
4204 		*p = old;
4205 		return restart_syscall();
4206 	}
4207 
4208 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
4209 		__s32 newf = net->ipv6.devconf_all->disable_ipv6;
4210 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
4211 		addrconf_disable_change(net, newf);
4212 	} else if ((!*p) ^ (!old))
4213 		dev_disable_change((struct inet6_dev *)table->extra1);
4214 
4215 	rtnl_unlock();
4216 	return 0;
4217 }
4218 
4219 static
4220 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4221 			    void __user *buffer, size_t *lenp, loff_t *ppos)
4222 {
4223 	int *valp = ctl->data;
4224 	int val = *valp;
4225 	loff_t pos = *ppos;
4226 	int ret;
4227 
4228 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4229 
4230 	if (write)
4231 		ret = addrconf_disable_ipv6(ctl, valp, val);
4232 	if (ret)
4233 		*ppos = pos;
4234 	return ret;
4235 }
4236 
4237 static struct addrconf_sysctl_table
4238 {
4239 	struct ctl_table_header *sysctl_header;
4240 	ctl_table addrconf_vars[DEVCONF_MAX+1];
4241 	char *dev_name;
4242 } addrconf_sysctl __read_mostly = {
4243 	.sysctl_header = NULL,
4244 	.addrconf_vars = {
4245 		{
4246 			.procname	= "forwarding",
4247 			.data		= &ipv6_devconf.forwarding,
4248 			.maxlen		= sizeof(int),
4249 			.mode		= 0644,
4250 			.proc_handler	= addrconf_sysctl_forward,
4251 		},
4252 		{
4253 			.procname	= "hop_limit",
4254 			.data		= &ipv6_devconf.hop_limit,
4255 			.maxlen		= sizeof(int),
4256 			.mode		= 0644,
4257 			.proc_handler	= proc_dointvec,
4258 		},
4259 		{
4260 			.procname	= "mtu",
4261 			.data		= &ipv6_devconf.mtu6,
4262 			.maxlen		= sizeof(int),
4263 			.mode		= 0644,
4264 			.proc_handler	= proc_dointvec,
4265 		},
4266 		{
4267 			.procname	= "accept_ra",
4268 			.data		= &ipv6_devconf.accept_ra,
4269 			.maxlen		= sizeof(int),
4270 			.mode		= 0644,
4271 			.proc_handler	= proc_dointvec,
4272 		},
4273 		{
4274 			.procname	= "accept_redirects",
4275 			.data		= &ipv6_devconf.accept_redirects,
4276 			.maxlen		= sizeof(int),
4277 			.mode		= 0644,
4278 			.proc_handler	= proc_dointvec,
4279 		},
4280 		{
4281 			.procname	= "autoconf",
4282 			.data		= &ipv6_devconf.autoconf,
4283 			.maxlen		= sizeof(int),
4284 			.mode		= 0644,
4285 			.proc_handler	= proc_dointvec,
4286 		},
4287 		{
4288 			.procname	= "dad_transmits",
4289 			.data		= &ipv6_devconf.dad_transmits,
4290 			.maxlen		= sizeof(int),
4291 			.mode		= 0644,
4292 			.proc_handler	= proc_dointvec,
4293 		},
4294 		{
4295 			.procname	= "router_solicitations",
4296 			.data		= &ipv6_devconf.rtr_solicits,
4297 			.maxlen		= sizeof(int),
4298 			.mode		= 0644,
4299 			.proc_handler	= proc_dointvec,
4300 		},
4301 		{
4302 			.procname	= "router_solicitation_interval",
4303 			.data		= &ipv6_devconf.rtr_solicit_interval,
4304 			.maxlen		= sizeof(int),
4305 			.mode		= 0644,
4306 			.proc_handler	= proc_dointvec_jiffies,
4307 		},
4308 		{
4309 			.procname	= "router_solicitation_delay",
4310 			.data		= &ipv6_devconf.rtr_solicit_delay,
4311 			.maxlen		= sizeof(int),
4312 			.mode		= 0644,
4313 			.proc_handler	= proc_dointvec_jiffies,
4314 		},
4315 		{
4316 			.procname	= "force_mld_version",
4317 			.data		= &ipv6_devconf.force_mld_version,
4318 			.maxlen		= sizeof(int),
4319 			.mode		= 0644,
4320 			.proc_handler	= proc_dointvec,
4321 		},
4322 #ifdef CONFIG_IPV6_PRIVACY
4323 		{
4324 			.procname	= "use_tempaddr",
4325 			.data		= &ipv6_devconf.use_tempaddr,
4326 			.maxlen		= sizeof(int),
4327 			.mode		= 0644,
4328 			.proc_handler	= proc_dointvec,
4329 		},
4330 		{
4331 			.procname	= "temp_valid_lft",
4332 			.data		= &ipv6_devconf.temp_valid_lft,
4333 			.maxlen		= sizeof(int),
4334 			.mode		= 0644,
4335 			.proc_handler	= proc_dointvec,
4336 		},
4337 		{
4338 			.procname	= "temp_prefered_lft",
4339 			.data		= &ipv6_devconf.temp_prefered_lft,
4340 			.maxlen		= sizeof(int),
4341 			.mode		= 0644,
4342 			.proc_handler	= proc_dointvec,
4343 		},
4344 		{
4345 			.procname	= "regen_max_retry",
4346 			.data		= &ipv6_devconf.regen_max_retry,
4347 			.maxlen		= sizeof(int),
4348 			.mode		= 0644,
4349 			.proc_handler	= proc_dointvec,
4350 		},
4351 		{
4352 			.procname	= "max_desync_factor",
4353 			.data		= &ipv6_devconf.max_desync_factor,
4354 			.maxlen		= sizeof(int),
4355 			.mode		= 0644,
4356 			.proc_handler	= proc_dointvec,
4357 		},
4358 #endif
4359 		{
4360 			.procname	= "max_addresses",
4361 			.data		= &ipv6_devconf.max_addresses,
4362 			.maxlen		= sizeof(int),
4363 			.mode		= 0644,
4364 			.proc_handler	= proc_dointvec,
4365 		},
4366 		{
4367 			.procname	= "accept_ra_defrtr",
4368 			.data		= &ipv6_devconf.accept_ra_defrtr,
4369 			.maxlen		= sizeof(int),
4370 			.mode		= 0644,
4371 			.proc_handler	= proc_dointvec,
4372 		},
4373 		{
4374 			.procname	= "accept_ra_pinfo",
4375 			.data		= &ipv6_devconf.accept_ra_pinfo,
4376 			.maxlen		= sizeof(int),
4377 			.mode		= 0644,
4378 			.proc_handler	= proc_dointvec,
4379 		},
4380 #ifdef CONFIG_IPV6_ROUTER_PREF
4381 		{
4382 			.procname	= "accept_ra_rtr_pref",
4383 			.data		= &ipv6_devconf.accept_ra_rtr_pref,
4384 			.maxlen		= sizeof(int),
4385 			.mode		= 0644,
4386 			.proc_handler	= proc_dointvec,
4387 		},
4388 		{
4389 			.procname	= "router_probe_interval",
4390 			.data		= &ipv6_devconf.rtr_probe_interval,
4391 			.maxlen		= sizeof(int),
4392 			.mode		= 0644,
4393 			.proc_handler	= proc_dointvec_jiffies,
4394 		},
4395 #ifdef CONFIG_IPV6_ROUTE_INFO
4396 		{
4397 			.procname	= "accept_ra_rt_info_max_plen",
4398 			.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
4399 			.maxlen		= sizeof(int),
4400 			.mode		= 0644,
4401 			.proc_handler	= proc_dointvec,
4402 		},
4403 #endif
4404 #endif
4405 		{
4406 			.procname	= "proxy_ndp",
4407 			.data		= &ipv6_devconf.proxy_ndp,
4408 			.maxlen		= sizeof(int),
4409 			.mode		= 0644,
4410 			.proc_handler	= proc_dointvec,
4411 		},
4412 		{
4413 			.procname	= "accept_source_route",
4414 			.data		= &ipv6_devconf.accept_source_route,
4415 			.maxlen		= sizeof(int),
4416 			.mode		= 0644,
4417 			.proc_handler	= proc_dointvec,
4418 		},
4419 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4420 		{
4421 			.procname       = "optimistic_dad",
4422 			.data           = &ipv6_devconf.optimistic_dad,
4423 			.maxlen         = sizeof(int),
4424 			.mode           = 0644,
4425 			.proc_handler   = proc_dointvec,
4426 
4427 		},
4428 #endif
4429 #ifdef CONFIG_IPV6_MROUTE
4430 		{
4431 			.procname	= "mc_forwarding",
4432 			.data		= &ipv6_devconf.mc_forwarding,
4433 			.maxlen		= sizeof(int),
4434 			.mode		= 0444,
4435 			.proc_handler	= proc_dointvec,
4436 		},
4437 #endif
4438 		{
4439 			.procname	= "disable_ipv6",
4440 			.data		= &ipv6_devconf.disable_ipv6,
4441 			.maxlen		= sizeof(int),
4442 			.mode		= 0644,
4443 			.proc_handler	= addrconf_sysctl_disable,
4444 		},
4445 		{
4446 			.procname	= "accept_dad",
4447 			.data		= &ipv6_devconf.accept_dad,
4448 			.maxlen		= sizeof(int),
4449 			.mode		= 0644,
4450 			.proc_handler	= proc_dointvec,
4451 		},
4452 		{
4453 			.procname       = "force_tllao",
4454 			.data           = &ipv6_devconf.force_tllao,
4455 			.maxlen         = sizeof(int),
4456 			.mode           = 0644,
4457 			.proc_handler   = proc_dointvec
4458 		},
4459 		{
4460 			/* sentinel */
4461 		}
4462 	},
4463 };
4464 
4465 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4466 		struct inet6_dev *idev, struct ipv6_devconf *p)
4467 {
4468 	int i;
4469 	struct addrconf_sysctl_table *t;
4470 
4471 #define ADDRCONF_CTL_PATH_DEV	3
4472 
4473 	struct ctl_path addrconf_ctl_path[] = {
4474 		{ .procname = "net", },
4475 		{ .procname = "ipv6", },
4476 		{ .procname = "conf", },
4477 		{ /* to be set */ },
4478 		{ },
4479 	};
4480 
4481 
4482 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4483 	if (t == NULL)
4484 		goto out;
4485 
4486 	for (i = 0; t->addrconf_vars[i].data; i++) {
4487 		t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4488 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4489 		t->addrconf_vars[i].extra2 = net;
4490 	}
4491 
4492 	/*
4493 	 * Make a copy of dev_name, because '.procname' is regarded as const
4494 	 * by sysctl and we wouldn't want anyone to change it under our feet
4495 	 * (see SIOCSIFNAME).
4496 	 */
4497 	t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4498 	if (!t->dev_name)
4499 		goto free;
4500 
4501 	addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4502 
4503 	t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4504 			t->addrconf_vars);
4505 	if (t->sysctl_header == NULL)
4506 		goto free_procname;
4507 
4508 	p->sysctl = t;
4509 	return 0;
4510 
4511 free_procname:
4512 	kfree(t->dev_name);
4513 free:
4514 	kfree(t);
4515 out:
4516 	return -ENOBUFS;
4517 }
4518 
4519 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4520 {
4521 	struct addrconf_sysctl_table *t;
4522 
4523 	if (p->sysctl == NULL)
4524 		return;
4525 
4526 	t = p->sysctl;
4527 	p->sysctl = NULL;
4528 	unregister_sysctl_table(t->sysctl_header);
4529 	kfree(t->dev_name);
4530 	kfree(t);
4531 }
4532 
4533 static void addrconf_sysctl_register(struct inet6_dev *idev)
4534 {
4535 	neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4536 			      &ndisc_ifinfo_sysctl_change);
4537 	__addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4538 					idev, &idev->cnf);
4539 }
4540 
4541 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4542 {
4543 	__addrconf_sysctl_unregister(&idev->cnf);
4544 	neigh_sysctl_unregister(idev->nd_parms);
4545 }
4546 
4547 
4548 #endif
4549 
4550 static int __net_init addrconf_init_net(struct net *net)
4551 {
4552 	int err;
4553 	struct ipv6_devconf *all, *dflt;
4554 
4555 	err = -ENOMEM;
4556 	all = &ipv6_devconf;
4557 	dflt = &ipv6_devconf_dflt;
4558 
4559 	if (!net_eq(net, &init_net)) {
4560 		all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4561 		if (all == NULL)
4562 			goto err_alloc_all;
4563 
4564 		dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4565 		if (dflt == NULL)
4566 			goto err_alloc_dflt;
4567 	} else {
4568 		/* these will be inherited by all namespaces */
4569 		dflt->autoconf = ipv6_defaults.autoconf;
4570 		dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4571 	}
4572 
4573 	net->ipv6.devconf_all = all;
4574 	net->ipv6.devconf_dflt = dflt;
4575 
4576 #ifdef CONFIG_SYSCTL
4577 	err = __addrconf_sysctl_register(net, "all", NULL, all);
4578 	if (err < 0)
4579 		goto err_reg_all;
4580 
4581 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4582 	if (err < 0)
4583 		goto err_reg_dflt;
4584 #endif
4585 	return 0;
4586 
4587 #ifdef CONFIG_SYSCTL
4588 err_reg_dflt:
4589 	__addrconf_sysctl_unregister(all);
4590 err_reg_all:
4591 	kfree(dflt);
4592 #endif
4593 err_alloc_dflt:
4594 	kfree(all);
4595 err_alloc_all:
4596 	return err;
4597 }
4598 
4599 static void __net_exit addrconf_exit_net(struct net *net)
4600 {
4601 #ifdef CONFIG_SYSCTL
4602 	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4603 	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
4604 #endif
4605 	if (!net_eq(net, &init_net)) {
4606 		kfree(net->ipv6.devconf_dflt);
4607 		kfree(net->ipv6.devconf_all);
4608 	}
4609 }
4610 
4611 static struct pernet_operations addrconf_ops = {
4612 	.init = addrconf_init_net,
4613 	.exit = addrconf_exit_net,
4614 };
4615 
4616 /*
4617  *      Device notifier
4618  */
4619 
4620 int register_inet6addr_notifier(struct notifier_block *nb)
4621 {
4622 	return atomic_notifier_chain_register(&inet6addr_chain, nb);
4623 }
4624 EXPORT_SYMBOL(register_inet6addr_notifier);
4625 
4626 int unregister_inet6addr_notifier(struct notifier_block *nb)
4627 {
4628 	return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4629 }
4630 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4631 
4632 /*
4633  *	Init / cleanup code
4634  */
4635 
4636 int __init addrconf_init(void)
4637 {
4638 	int i, err;
4639 
4640 	err = ipv6_addr_label_init();
4641 	if (err < 0) {
4642 		printk(KERN_CRIT "IPv6 Addrconf:"
4643 		       " cannot initialize default policy table: %d.\n", err);
4644 		goto out;
4645 	}
4646 
4647 	err = register_pernet_subsys(&addrconf_ops);
4648 	if (err < 0)
4649 		goto out_addrlabel;
4650 
4651 	/* The addrconf netdev notifier requires that loopback_dev
4652 	 * has it's ipv6 private information allocated and setup
4653 	 * before it can bring up and give link-local addresses
4654 	 * to other devices which are up.
4655 	 *
4656 	 * Unfortunately, loopback_dev is not necessarily the first
4657 	 * entry in the global dev_base list of net devices.  In fact,
4658 	 * it is likely to be the very last entry on that list.
4659 	 * So this causes the notifier registry below to try and
4660 	 * give link-local addresses to all devices besides loopback_dev
4661 	 * first, then loopback_dev, which cases all the non-loopback_dev
4662 	 * devices to fail to get a link-local address.
4663 	 *
4664 	 * So, as a temporary fix, allocate the ipv6 structure for
4665 	 * loopback_dev first by hand.
4666 	 * Longer term, all of the dependencies ipv6 has upon the loopback
4667 	 * device and it being up should be removed.
4668 	 */
4669 	rtnl_lock();
4670 	if (!ipv6_add_dev(init_net.loopback_dev))
4671 		err = -ENOMEM;
4672 	rtnl_unlock();
4673 	if (err)
4674 		goto errlo;
4675 
4676 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4677 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4678 
4679 	register_netdevice_notifier(&ipv6_dev_notf);
4680 
4681 	addrconf_verify(0);
4682 
4683 	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4684 	if (err < 0)
4685 		goto errout;
4686 
4687 	/* Only the first call to __rtnl_register can fail */
4688 	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4689 	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4690 	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4691 	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4692 	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4693 
4694 	ipv6_addr_label_rtnl_register();
4695 
4696 	return 0;
4697 errout:
4698 	unregister_netdevice_notifier(&ipv6_dev_notf);
4699 errlo:
4700 	unregister_pernet_subsys(&addrconf_ops);
4701 out_addrlabel:
4702 	ipv6_addr_label_cleanup();
4703 out:
4704 	return err;
4705 }
4706 
4707 void addrconf_cleanup(void)
4708 {
4709 	struct net_device *dev;
4710 	int i;
4711 
4712 	unregister_netdevice_notifier(&ipv6_dev_notf);
4713 	unregister_pernet_subsys(&addrconf_ops);
4714 	ipv6_addr_label_cleanup();
4715 
4716 	rtnl_lock();
4717 
4718 	/* clean dev list */
4719 	for_each_netdev(&init_net, dev) {
4720 		if (__in6_dev_get(dev) == NULL)
4721 			continue;
4722 		addrconf_ifdown(dev, 1);
4723 	}
4724 	addrconf_ifdown(init_net.loopback_dev, 2);
4725 
4726 	/*
4727 	 *	Check hash table.
4728 	 */
4729 	spin_lock_bh(&addrconf_hash_lock);
4730 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4731 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4732 	spin_unlock_bh(&addrconf_hash_lock);
4733 
4734 	del_timer(&addr_chk_timer);
4735 	rtnl_unlock();
4736 }
4737