xref: /openbmc/linux/net/ipv6/addrconf.c (revision 08f10aff)
1 /*
2  *	IPv6 Address [auto]configuration
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14 
15 /*
16  *	Changes:
17  *
18  *	Janos Farkas			:	delete timer on ifdown
19  *	<chexum@bankinf.banki.hu>
20  *	Andi Kleen			:	kill double kfree on module
21  *						unload.
22  *	Maciej W. Rozycki		:	FDDI support
23  *	sekiya@USAGI			:	Don't send too many RS
24  *						packets.
25  *	yoshfuji@USAGI			:       Fixed interval between DAD
26  *						packets.
27  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
28  *						address validation timer.
29  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
30  *						support.
31  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
32  *						address on a same interface.
33  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
34  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
35  *						seq_file.
36  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
37  *						selection; consider scope,
38  *						status etc.
39  */
40 
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68 
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80 
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84 
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
87 
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90 #include <linux/export.h>
91 
92 /* Set to 3 to get tracing... */
93 #define ACONF_DEBUG 2
94 
95 #if ACONF_DEBUG >= 3
96 #define ADBG(x) printk x
97 #else
98 #define ADBG(x)
99 #endif
100 
101 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
102 
103 static inline u32 cstamp_delta(unsigned long cstamp)
104 {
105 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
106 }
107 
108 #define ADDRCONF_TIMER_FUZZ_MINUS	(HZ > 50 ? HZ/50 : 1)
109 #define ADDRCONF_TIMER_FUZZ		(HZ / 4)
110 #define ADDRCONF_TIMER_FUZZ_MAX		(HZ)
111 
112 #ifdef CONFIG_SYSCTL
113 static void addrconf_sysctl_register(struct inet6_dev *idev);
114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 #else
116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
117 {
118 }
119 
120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
121 {
122 }
123 #endif
124 
125 #ifdef CONFIG_IPV6_PRIVACY
126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
128 static void ipv6_regen_rndid(unsigned long data);
129 #endif
130 
131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
132 static int ipv6_count_addresses(struct inet6_dev *idev);
133 
134 /*
135  *	Configured unicast address hash table
136  */
137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
138 static DEFINE_SPINLOCK(addrconf_hash_lock);
139 
140 static void addrconf_verify(unsigned long);
141 
142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
143 static DEFINE_SPINLOCK(addrconf_verify_lock);
144 
145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
147 
148 static void addrconf_type_change(struct net_device *dev,
149 				 unsigned long event);
150 static int addrconf_ifdown(struct net_device *dev, int how);
151 
152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
153 static void addrconf_dad_timer(unsigned long data);
154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
155 static void addrconf_dad_run(struct inet6_dev *idev);
156 static void addrconf_rs_timer(unsigned long data);
157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
159 
160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
161 				struct prefix_info *pinfo);
162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
163 			       struct net_device *dev);
164 
165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
166 
167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168 	.forwarding		= 0,
169 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
170 	.mtu6			= IPV6_MIN_MTU,
171 	.accept_ra		= 1,
172 	.accept_redirects	= 1,
173 	.autoconf		= 1,
174 	.force_mld_version	= 0,
175 	.dad_transmits		= 1,
176 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
177 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
178 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
179 #ifdef CONFIG_IPV6_PRIVACY
180 	.use_tempaddr 		= 0,
181 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
182 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
183 	.regen_max_retry	= REGEN_MAX_RETRY,
184 	.max_desync_factor	= MAX_DESYNC_FACTOR,
185 #endif
186 	.max_addresses		= IPV6_MAX_ADDRESSES,
187 	.accept_ra_defrtr	= 1,
188 	.accept_ra_pinfo	= 1,
189 #ifdef CONFIG_IPV6_ROUTER_PREF
190 	.accept_ra_rtr_pref	= 1,
191 	.rtr_probe_interval	= 60 * HZ,
192 #ifdef CONFIG_IPV6_ROUTE_INFO
193 	.accept_ra_rt_info_max_plen = 0,
194 #endif
195 #endif
196 	.proxy_ndp		= 0,
197 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
198 	.disable_ipv6		= 0,
199 	.accept_dad		= 1,
200 };
201 
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 	.forwarding		= 0,
204 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
205 	.mtu6			= IPV6_MIN_MTU,
206 	.accept_ra		= 1,
207 	.accept_redirects	= 1,
208 	.autoconf		= 1,
209 	.dad_transmits		= 1,
210 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
211 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
212 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
213 #ifdef CONFIG_IPV6_PRIVACY
214 	.use_tempaddr		= 0,
215 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
216 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
217 	.regen_max_retry	= REGEN_MAX_RETRY,
218 	.max_desync_factor	= MAX_DESYNC_FACTOR,
219 #endif
220 	.max_addresses		= IPV6_MAX_ADDRESSES,
221 	.accept_ra_defrtr	= 1,
222 	.accept_ra_pinfo	= 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 	.accept_ra_rtr_pref	= 1,
225 	.rtr_probe_interval	= 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 	.accept_ra_rt_info_max_plen = 0,
228 #endif
229 #endif
230 	.proxy_ndp		= 0,
231 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
232 	.disable_ipv6		= 0,
233 	.accept_dad		= 1,
234 };
235 
236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
241 
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
244 {
245 	return !qdisc_tx_is_noop(dev);
246 }
247 
248 /* Check if a route is valid prefix route */
249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
250 {
251 	return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
252 }
253 
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
255 {
256 	if (del_timer(&ifp->timer))
257 		__in6_ifa_put(ifp);
258 }
259 
260 enum addrconf_timer_t {
261 	AC_NONE,
262 	AC_DAD,
263 	AC_RS,
264 };
265 
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267 			       enum addrconf_timer_t what,
268 			       unsigned long when)
269 {
270 	if (!del_timer(&ifp->timer))
271 		in6_ifa_hold(ifp);
272 
273 	switch (what) {
274 	case AC_DAD:
275 		ifp->timer.function = addrconf_dad_timer;
276 		break;
277 	case AC_RS:
278 		ifp->timer.function = addrconf_rs_timer;
279 		break;
280 	default:
281 		break;
282 	}
283 	ifp->timer.expires = jiffies + when;
284 	add_timer(&ifp->timer);
285 }
286 
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
288 {
289 	if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290 			  sizeof(struct ipstats_mib),
291 			  __alignof__(struct ipstats_mib)) < 0)
292 		goto err_ip;
293 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294 					GFP_KERNEL);
295 	if (!idev->stats.icmpv6dev)
296 		goto err_icmp;
297 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298 					   GFP_KERNEL);
299 	if (!idev->stats.icmpv6msgdev)
300 		goto err_icmpmsg;
301 
302 	return 0;
303 
304 err_icmpmsg:
305 	kfree(idev->stats.icmpv6dev);
306 err_icmp:
307 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 err_ip:
309 	return -ENOMEM;
310 }
311 
312 static void snmp6_free_dev(struct inet6_dev *idev)
313 {
314 	kfree(idev->stats.icmpv6msgdev);
315 	kfree(idev->stats.icmpv6dev);
316 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
317 }
318 
319 /* Nobody refers to this device, we may destroy it. */
320 
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
322 {
323 	struct net_device *dev = idev->dev;
324 
325 	WARN_ON(!list_empty(&idev->addr_list));
326 	WARN_ON(idev->mc_list != NULL);
327 
328 #ifdef NET_REFCNT_DEBUG
329 	printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
330 #endif
331 	dev_put(dev);
332 	if (!idev->dead) {
333 		pr_warning("Freeing alive inet6 device %p\n", idev);
334 		return;
335 	}
336 	snmp6_free_dev(idev);
337 	kfree_rcu(idev, rcu);
338 }
339 EXPORT_SYMBOL(in6_dev_finish_destroy);
340 
341 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
342 {
343 	struct inet6_dev *ndev;
344 
345 	ASSERT_RTNL();
346 
347 	if (dev->mtu < IPV6_MIN_MTU)
348 		return NULL;
349 
350 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
351 
352 	if (ndev == NULL)
353 		return NULL;
354 
355 	rwlock_init(&ndev->lock);
356 	ndev->dev = dev;
357 	INIT_LIST_HEAD(&ndev->addr_list);
358 
359 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
360 	ndev->cnf.mtu6 = dev->mtu;
361 	ndev->cnf.sysctl = NULL;
362 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
363 	if (ndev->nd_parms == NULL) {
364 		kfree(ndev);
365 		return NULL;
366 	}
367 	if (ndev->cnf.forwarding)
368 		dev_disable_lro(dev);
369 	/* We refer to the device */
370 	dev_hold(dev);
371 
372 	if (snmp6_alloc_dev(ndev) < 0) {
373 		ADBG((KERN_WARNING
374 			"%s(): cannot allocate memory for statistics; dev=%s.\n",
375 			__func__, dev->name));
376 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
377 		dev_put(dev);
378 		kfree(ndev);
379 		return NULL;
380 	}
381 
382 	if (snmp6_register_dev(ndev) < 0) {
383 		ADBG((KERN_WARNING
384 			"%s(): cannot create /proc/net/dev_snmp6/%s\n",
385 			__func__, dev->name));
386 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
387 		ndev->dead = 1;
388 		in6_dev_finish_destroy(ndev);
389 		return NULL;
390 	}
391 
392 	/* One reference from device.  We must do this before
393 	 * we invoke __ipv6_regen_rndid().
394 	 */
395 	in6_dev_hold(ndev);
396 
397 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
398 		ndev->cnf.accept_dad = -1;
399 
400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
401 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
402 		printk(KERN_INFO
403 		       "%s: Disabled Multicast RS\n",
404 		       dev->name);
405 		ndev->cnf.rtr_solicits = 0;
406 	}
407 #endif
408 
409 #ifdef CONFIG_IPV6_PRIVACY
410 	INIT_LIST_HEAD(&ndev->tempaddr_list);
411 	setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
412 	if ((dev->flags&IFF_LOOPBACK) ||
413 	    dev->type == ARPHRD_TUNNEL ||
414 	    dev->type == ARPHRD_TUNNEL6 ||
415 	    dev->type == ARPHRD_SIT ||
416 	    dev->type == ARPHRD_NONE) {
417 		ndev->cnf.use_tempaddr = -1;
418 	} else {
419 		in6_dev_hold(ndev);
420 		ipv6_regen_rndid((unsigned long) ndev);
421 	}
422 #endif
423 
424 	if (netif_running(dev) && addrconf_qdisc_ok(dev))
425 		ndev->if_flags |= IF_READY;
426 
427 	ipv6_mc_init_dev(ndev);
428 	ndev->tstamp = jiffies;
429 	addrconf_sysctl_register(ndev);
430 	/* protected by rtnl_lock */
431 	rcu_assign_pointer(dev->ip6_ptr, ndev);
432 
433 	/* Join all-node multicast group */
434 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
435 
436 	/* Join all-router multicast group if forwarding is set */
437 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
438 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
439 
440 	return ndev;
441 }
442 
443 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
444 {
445 	struct inet6_dev *idev;
446 
447 	ASSERT_RTNL();
448 
449 	idev = __in6_dev_get(dev);
450 	if (!idev) {
451 		idev = ipv6_add_dev(dev);
452 		if (!idev)
453 			return NULL;
454 	}
455 
456 	if (dev->flags&IFF_UP)
457 		ipv6_mc_up(idev);
458 	return idev;
459 }
460 
461 #ifdef CONFIG_SYSCTL
462 static void dev_forward_change(struct inet6_dev *idev)
463 {
464 	struct net_device *dev;
465 	struct inet6_ifaddr *ifa;
466 
467 	if (!idev)
468 		return;
469 	dev = idev->dev;
470 	if (idev->cnf.forwarding)
471 		dev_disable_lro(dev);
472 	if (dev && (dev->flags & IFF_MULTICAST)) {
473 		if (idev->cnf.forwarding)
474 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
475 		else
476 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
477 	}
478 
479 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
480 		if (ifa->flags&IFA_F_TENTATIVE)
481 			continue;
482 		if (idev->cnf.forwarding)
483 			addrconf_join_anycast(ifa);
484 		else
485 			addrconf_leave_anycast(ifa);
486 	}
487 }
488 
489 
490 static void addrconf_forward_change(struct net *net, __s32 newf)
491 {
492 	struct net_device *dev;
493 	struct inet6_dev *idev;
494 
495 	rcu_read_lock();
496 	for_each_netdev_rcu(net, dev) {
497 		idev = __in6_dev_get(dev);
498 		if (idev) {
499 			int changed = (!idev->cnf.forwarding) ^ (!newf);
500 			idev->cnf.forwarding = newf;
501 			if (changed)
502 				dev_forward_change(idev);
503 		}
504 	}
505 	rcu_read_unlock();
506 }
507 
508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
509 {
510 	struct net *net;
511 	int old;
512 
513 	if (!rtnl_trylock())
514 		return restart_syscall();
515 
516 	net = (struct net *)table->extra2;
517 	old = *p;
518 	*p = newf;
519 
520 	if (p == &net->ipv6.devconf_dflt->forwarding) {
521 		rtnl_unlock();
522 		return 0;
523 	}
524 
525 	if (p == &net->ipv6.devconf_all->forwarding) {
526 		net->ipv6.devconf_dflt->forwarding = newf;
527 		addrconf_forward_change(net, newf);
528 	} else if ((!newf) ^ (!old))
529 		dev_forward_change((struct inet6_dev *)table->extra1);
530 	rtnl_unlock();
531 
532 	if (newf)
533 		rt6_purge_dflt_routers(net);
534 	return 1;
535 }
536 #endif
537 
538 /* Nobody refers to this ifaddr, destroy it */
539 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
540 {
541 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
542 
543 #ifdef NET_REFCNT_DEBUG
544 	printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
545 #endif
546 
547 	in6_dev_put(ifp->idev);
548 
549 	if (del_timer(&ifp->timer))
550 		pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
551 
552 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
553 		pr_warning("Freeing alive inet6 address %p\n", ifp);
554 		return;
555 	}
556 	dst_release(&ifp->rt->dst);
557 
558 	kfree_rcu(ifp, rcu);
559 }
560 
561 static void
562 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
563 {
564 	struct list_head *p;
565 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
566 
567 	/*
568 	 * Each device address list is sorted in order of scope -
569 	 * global before linklocal.
570 	 */
571 	list_for_each(p, &idev->addr_list) {
572 		struct inet6_ifaddr *ifa
573 			= list_entry(p, struct inet6_ifaddr, if_list);
574 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
575 			break;
576 	}
577 
578 	list_add_tail(&ifp->if_list, p);
579 }
580 
581 static u32 ipv6_addr_hash(const struct in6_addr *addr)
582 {
583 	/*
584 	 * We perform the hash function over the last 64 bits of the address
585 	 * This will include the IEEE address token on links that support it.
586 	 */
587 	return jhash_2words((__force u32)addr->s6_addr32[2],
588 			    (__force u32)addr->s6_addr32[3], 0)
589 		& (IN6_ADDR_HSIZE - 1);
590 }
591 
592 /* On success it returns ifp with increased reference count */
593 
594 static struct inet6_ifaddr *
595 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
596 	      int scope, u32 flags)
597 {
598 	struct inet6_ifaddr *ifa = NULL;
599 	struct rt6_info *rt;
600 	unsigned int hash;
601 	int err = 0;
602 	int addr_type = ipv6_addr_type(addr);
603 
604 	if (addr_type == IPV6_ADDR_ANY ||
605 	    addr_type & IPV6_ADDR_MULTICAST ||
606 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
607 	     addr_type & IPV6_ADDR_LOOPBACK))
608 		return ERR_PTR(-EADDRNOTAVAIL);
609 
610 	rcu_read_lock_bh();
611 	if (idev->dead) {
612 		err = -ENODEV;			/*XXX*/
613 		goto out2;
614 	}
615 
616 	if (idev->cnf.disable_ipv6) {
617 		err = -EACCES;
618 		goto out2;
619 	}
620 
621 	spin_lock(&addrconf_hash_lock);
622 
623 	/* Ignore adding duplicate addresses on an interface */
624 	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
625 		ADBG(("ipv6_add_addr: already assigned\n"));
626 		err = -EEXIST;
627 		goto out;
628 	}
629 
630 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
631 
632 	if (ifa == NULL) {
633 		ADBG(("ipv6_add_addr: malloc failed\n"));
634 		err = -ENOBUFS;
635 		goto out;
636 	}
637 
638 	rt = addrconf_dst_alloc(idev, addr, false);
639 	if (IS_ERR(rt)) {
640 		err = PTR_ERR(rt);
641 		goto out;
642 	}
643 
644 	ifa->addr = *addr;
645 
646 	spin_lock_init(&ifa->lock);
647 	spin_lock_init(&ifa->state_lock);
648 	init_timer(&ifa->timer);
649 	INIT_HLIST_NODE(&ifa->addr_lst);
650 	ifa->timer.data = (unsigned long) ifa;
651 	ifa->scope = scope;
652 	ifa->prefix_len = pfxlen;
653 	ifa->flags = flags | IFA_F_TENTATIVE;
654 	ifa->cstamp = ifa->tstamp = jiffies;
655 
656 	ifa->rt = rt;
657 
658 	ifa->idev = idev;
659 	in6_dev_hold(idev);
660 	/* For caller */
661 	in6_ifa_hold(ifa);
662 
663 	/* Add to big hash table */
664 	hash = ipv6_addr_hash(addr);
665 
666 	hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
667 	spin_unlock(&addrconf_hash_lock);
668 
669 	write_lock(&idev->lock);
670 	/* Add to inet6_dev unicast addr list. */
671 	ipv6_link_dev_addr(idev, ifa);
672 
673 #ifdef CONFIG_IPV6_PRIVACY
674 	if (ifa->flags&IFA_F_TEMPORARY) {
675 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
676 		in6_ifa_hold(ifa);
677 	}
678 #endif
679 
680 	in6_ifa_hold(ifa);
681 	write_unlock(&idev->lock);
682 out2:
683 	rcu_read_unlock_bh();
684 
685 	if (likely(err == 0))
686 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
687 	else {
688 		kfree(ifa);
689 		ifa = ERR_PTR(err);
690 	}
691 
692 	return ifa;
693 out:
694 	spin_unlock(&addrconf_hash_lock);
695 	goto out2;
696 }
697 
698 /* This function wants to get referenced ifp and releases it before return */
699 
700 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
701 {
702 	struct inet6_ifaddr *ifa, *ifn;
703 	struct inet6_dev *idev = ifp->idev;
704 	int state;
705 	int deleted = 0, onlink = 0;
706 	unsigned long expires = jiffies;
707 
708 	spin_lock_bh(&ifp->state_lock);
709 	state = ifp->state;
710 	ifp->state = INET6_IFADDR_STATE_DEAD;
711 	spin_unlock_bh(&ifp->state_lock);
712 
713 	if (state == INET6_IFADDR_STATE_DEAD)
714 		goto out;
715 
716 	spin_lock_bh(&addrconf_hash_lock);
717 	hlist_del_init_rcu(&ifp->addr_lst);
718 	spin_unlock_bh(&addrconf_hash_lock);
719 
720 	write_lock_bh(&idev->lock);
721 #ifdef CONFIG_IPV6_PRIVACY
722 	if (ifp->flags&IFA_F_TEMPORARY) {
723 		list_del(&ifp->tmp_list);
724 		if (ifp->ifpub) {
725 			in6_ifa_put(ifp->ifpub);
726 			ifp->ifpub = NULL;
727 		}
728 		__in6_ifa_put(ifp);
729 	}
730 #endif
731 
732 	list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
733 		if (ifa == ifp) {
734 			list_del_init(&ifp->if_list);
735 			__in6_ifa_put(ifp);
736 
737 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
738 				break;
739 			deleted = 1;
740 			continue;
741 		} else if (ifp->flags & IFA_F_PERMANENT) {
742 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
743 					      ifp->prefix_len)) {
744 				if (ifa->flags & IFA_F_PERMANENT) {
745 					onlink = 1;
746 					if (deleted)
747 						break;
748 				} else {
749 					unsigned long lifetime;
750 
751 					if (!onlink)
752 						onlink = -1;
753 
754 					spin_lock(&ifa->lock);
755 
756 					lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
757 					/*
758 					 * Note: Because this address is
759 					 * not permanent, lifetime <
760 					 * LONG_MAX / HZ here.
761 					 */
762 					if (time_before(expires,
763 							ifa->tstamp + lifetime * HZ))
764 						expires = ifa->tstamp + lifetime * HZ;
765 					spin_unlock(&ifa->lock);
766 				}
767 			}
768 		}
769 	}
770 	write_unlock_bh(&idev->lock);
771 
772 	addrconf_del_timer(ifp);
773 
774 	ipv6_ifa_notify(RTM_DELADDR, ifp);
775 
776 	atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
777 
778 	/*
779 	 * Purge or update corresponding prefix
780 	 *
781 	 * 1) we don't purge prefix here if address was not permanent.
782 	 *    prefix is managed by its own lifetime.
783 	 * 2) if there're no addresses, delete prefix.
784 	 * 3) if there're still other permanent address(es),
785 	 *    corresponding prefix is still permanent.
786 	 * 4) otherwise, update prefix lifetime to the
787 	 *    longest valid lifetime among the corresponding
788 	 *    addresses on the device.
789 	 *    Note: subsequent RA will update lifetime.
790 	 *
791 	 * --yoshfuji
792 	 */
793 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
794 		struct in6_addr prefix;
795 		struct rt6_info *rt;
796 		struct net *net = dev_net(ifp->idev->dev);
797 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
798 		rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
799 
800 		if (rt && addrconf_is_prefix_route(rt)) {
801 			if (onlink == 0) {
802 				ip6_del_rt(rt);
803 				rt = NULL;
804 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
805 				rt->dst.expires = expires;
806 				rt->rt6i_flags |= RTF_EXPIRES;
807 			}
808 		}
809 		dst_release(&rt->dst);
810 	}
811 
812 	/* clean up prefsrc entries */
813 	rt6_remove_prefsrc(ifp);
814 out:
815 	in6_ifa_put(ifp);
816 }
817 
818 #ifdef CONFIG_IPV6_PRIVACY
819 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
820 {
821 	struct inet6_dev *idev = ifp->idev;
822 	struct in6_addr addr, *tmpaddr;
823 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
824 	unsigned long regen_advance;
825 	int tmp_plen;
826 	int ret = 0;
827 	int max_addresses;
828 	u32 addr_flags;
829 	unsigned long now = jiffies;
830 
831 	write_lock(&idev->lock);
832 	if (ift) {
833 		spin_lock_bh(&ift->lock);
834 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
835 		spin_unlock_bh(&ift->lock);
836 		tmpaddr = &addr;
837 	} else {
838 		tmpaddr = NULL;
839 	}
840 retry:
841 	in6_dev_hold(idev);
842 	if (idev->cnf.use_tempaddr <= 0) {
843 		write_unlock(&idev->lock);
844 		printk(KERN_INFO
845 			"ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
846 		in6_dev_put(idev);
847 		ret = -1;
848 		goto out;
849 	}
850 	spin_lock_bh(&ifp->lock);
851 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
852 		idev->cnf.use_tempaddr = -1;	/*XXX*/
853 		spin_unlock_bh(&ifp->lock);
854 		write_unlock(&idev->lock);
855 		printk(KERN_WARNING
856 			"ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
857 		in6_dev_put(idev);
858 		ret = -1;
859 		goto out;
860 	}
861 	in6_ifa_hold(ifp);
862 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
863 	if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
864 		spin_unlock_bh(&ifp->lock);
865 		write_unlock(&idev->lock);
866 		printk(KERN_WARNING
867 			"ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
868 		in6_ifa_put(ifp);
869 		in6_dev_put(idev);
870 		ret = -1;
871 		goto out;
872 	}
873 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
874 	age = (now - ifp->tstamp) / HZ;
875 	tmp_valid_lft = min_t(__u32,
876 			      ifp->valid_lft,
877 			      idev->cnf.temp_valid_lft + age);
878 	tmp_prefered_lft = min_t(__u32,
879 				 ifp->prefered_lft,
880 				 idev->cnf.temp_prefered_lft + age -
881 				 idev->cnf.max_desync_factor);
882 	tmp_plen = ifp->prefix_len;
883 	max_addresses = idev->cnf.max_addresses;
884 	tmp_tstamp = ifp->tstamp;
885 	spin_unlock_bh(&ifp->lock);
886 
887 	regen_advance = idev->cnf.regen_max_retry *
888 	                idev->cnf.dad_transmits *
889 	                idev->nd_parms->retrans_time / HZ;
890 	write_unlock(&idev->lock);
891 
892 	/* A temporary address is created only if this calculated Preferred
893 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
894 	 * an implementation must not create a temporary address with a zero
895 	 * Preferred Lifetime.
896 	 */
897 	if (tmp_prefered_lft <= regen_advance) {
898 		in6_ifa_put(ifp);
899 		in6_dev_put(idev);
900 		ret = -1;
901 		goto out;
902 	}
903 
904 	addr_flags = IFA_F_TEMPORARY;
905 	/* set in addrconf_prefix_rcv() */
906 	if (ifp->flags & IFA_F_OPTIMISTIC)
907 		addr_flags |= IFA_F_OPTIMISTIC;
908 
909 	ift = !max_addresses ||
910 	      ipv6_count_addresses(idev) < max_addresses ?
911 		ipv6_add_addr(idev, &addr, tmp_plen,
912 			      ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
913 			      addr_flags) : NULL;
914 	if (!ift || IS_ERR(ift)) {
915 		in6_ifa_put(ifp);
916 		in6_dev_put(idev);
917 		printk(KERN_INFO
918 			"ipv6_create_tempaddr(): retry temporary address regeneration.\n");
919 		tmpaddr = &addr;
920 		write_lock(&idev->lock);
921 		goto retry;
922 	}
923 
924 	spin_lock_bh(&ift->lock);
925 	ift->ifpub = ifp;
926 	ift->valid_lft = tmp_valid_lft;
927 	ift->prefered_lft = tmp_prefered_lft;
928 	ift->cstamp = now;
929 	ift->tstamp = tmp_tstamp;
930 	spin_unlock_bh(&ift->lock);
931 
932 	addrconf_dad_start(ift, 0);
933 	in6_ifa_put(ift);
934 	in6_dev_put(idev);
935 out:
936 	return ret;
937 }
938 #endif
939 
940 /*
941  *	Choose an appropriate source address (RFC3484)
942  */
943 enum {
944 	IPV6_SADDR_RULE_INIT = 0,
945 	IPV6_SADDR_RULE_LOCAL,
946 	IPV6_SADDR_RULE_SCOPE,
947 	IPV6_SADDR_RULE_PREFERRED,
948 #ifdef CONFIG_IPV6_MIP6
949 	IPV6_SADDR_RULE_HOA,
950 #endif
951 	IPV6_SADDR_RULE_OIF,
952 	IPV6_SADDR_RULE_LABEL,
953 #ifdef CONFIG_IPV6_PRIVACY
954 	IPV6_SADDR_RULE_PRIVACY,
955 #endif
956 	IPV6_SADDR_RULE_ORCHID,
957 	IPV6_SADDR_RULE_PREFIX,
958 	IPV6_SADDR_RULE_MAX
959 };
960 
961 struct ipv6_saddr_score {
962 	int			rule;
963 	int			addr_type;
964 	struct inet6_ifaddr	*ifa;
965 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
966 	int			scopedist;
967 	int			matchlen;
968 };
969 
970 struct ipv6_saddr_dst {
971 	const struct in6_addr *addr;
972 	int ifindex;
973 	int scope;
974 	int label;
975 	unsigned int prefs;
976 };
977 
978 static inline int ipv6_saddr_preferred(int type)
979 {
980 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
981 		return 1;
982 	return 0;
983 }
984 
985 static int ipv6_get_saddr_eval(struct net *net,
986 			       struct ipv6_saddr_score *score,
987 			       struct ipv6_saddr_dst *dst,
988 			       int i)
989 {
990 	int ret;
991 
992 	if (i <= score->rule) {
993 		switch (i) {
994 		case IPV6_SADDR_RULE_SCOPE:
995 			ret = score->scopedist;
996 			break;
997 		case IPV6_SADDR_RULE_PREFIX:
998 			ret = score->matchlen;
999 			break;
1000 		default:
1001 			ret = !!test_bit(i, score->scorebits);
1002 		}
1003 		goto out;
1004 	}
1005 
1006 	switch (i) {
1007 	case IPV6_SADDR_RULE_INIT:
1008 		/* Rule 0: remember if hiscore is not ready yet */
1009 		ret = !!score->ifa;
1010 		break;
1011 	case IPV6_SADDR_RULE_LOCAL:
1012 		/* Rule 1: Prefer same address */
1013 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1014 		break;
1015 	case IPV6_SADDR_RULE_SCOPE:
1016 		/* Rule 2: Prefer appropriate scope
1017 		 *
1018 		 *      ret
1019 		 *       ^
1020 		 *    -1 |  d 15
1021 		 *    ---+--+-+---> scope
1022 		 *       |
1023 		 *       |             d is scope of the destination.
1024 		 *  B-d  |  \
1025 		 *       |   \      <- smaller scope is better if
1026 		 *  B-15 |    \        if scope is enough for destinaion.
1027 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1028 		 * d-C-1 | /
1029 		 *       |/         <- greater is better
1030 		 *   -C  /             if scope is not enough for destination.
1031 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1032 		 *
1033 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1034 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1035 		 * Assume B = 0 and we get C > 29.
1036 		 */
1037 		ret = __ipv6_addr_src_scope(score->addr_type);
1038 		if (ret >= dst->scope)
1039 			ret = -ret;
1040 		else
1041 			ret -= 128;	/* 30 is enough */
1042 		score->scopedist = ret;
1043 		break;
1044 	case IPV6_SADDR_RULE_PREFERRED:
1045 		/* Rule 3: Avoid deprecated and optimistic addresses */
1046 		ret = ipv6_saddr_preferred(score->addr_type) ||
1047 		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1048 		break;
1049 #ifdef CONFIG_IPV6_MIP6
1050 	case IPV6_SADDR_RULE_HOA:
1051 	    {
1052 		/* Rule 4: Prefer home address */
1053 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1054 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1055 		break;
1056 	    }
1057 #endif
1058 	case IPV6_SADDR_RULE_OIF:
1059 		/* Rule 5: Prefer outgoing interface */
1060 		ret = (!dst->ifindex ||
1061 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1062 		break;
1063 	case IPV6_SADDR_RULE_LABEL:
1064 		/* Rule 6: Prefer matching label */
1065 		ret = ipv6_addr_label(net,
1066 				      &score->ifa->addr, score->addr_type,
1067 				      score->ifa->idev->dev->ifindex) == dst->label;
1068 		break;
1069 #ifdef CONFIG_IPV6_PRIVACY
1070 	case IPV6_SADDR_RULE_PRIVACY:
1071 	    {
1072 		/* Rule 7: Prefer public address
1073 		 * Note: prefer temporary address if use_tempaddr >= 2
1074 		 */
1075 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1076 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1077 				score->ifa->idev->cnf.use_tempaddr >= 2;
1078 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1079 		break;
1080 	    }
1081 #endif
1082 	case IPV6_SADDR_RULE_ORCHID:
1083 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1084 		 *	    non-ORCHID vs non-ORCHID
1085 		 */
1086 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1087 			ipv6_addr_orchid(dst->addr));
1088 		break;
1089 	case IPV6_SADDR_RULE_PREFIX:
1090 		/* Rule 8: Use longest matching prefix */
1091 		score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1092 						       dst->addr);
1093 		break;
1094 	default:
1095 		ret = 0;
1096 	}
1097 
1098 	if (ret)
1099 		__set_bit(i, score->scorebits);
1100 	score->rule = i;
1101 out:
1102 	return ret;
1103 }
1104 
1105 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1106 		       const struct in6_addr *daddr, unsigned int prefs,
1107 		       struct in6_addr *saddr)
1108 {
1109 	struct ipv6_saddr_score scores[2],
1110 				*score = &scores[0], *hiscore = &scores[1];
1111 	struct ipv6_saddr_dst dst;
1112 	struct net_device *dev;
1113 	int dst_type;
1114 
1115 	dst_type = __ipv6_addr_type(daddr);
1116 	dst.addr = daddr;
1117 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1118 	dst.scope = __ipv6_addr_src_scope(dst_type);
1119 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1120 	dst.prefs = prefs;
1121 
1122 	hiscore->rule = -1;
1123 	hiscore->ifa = NULL;
1124 
1125 	rcu_read_lock();
1126 
1127 	for_each_netdev_rcu(net, dev) {
1128 		struct inet6_dev *idev;
1129 
1130 		/* Candidate Source Address (section 4)
1131 		 *  - multicast and link-local destination address,
1132 		 *    the set of candidate source address MUST only
1133 		 *    include addresses assigned to interfaces
1134 		 *    belonging to the same link as the outgoing
1135 		 *    interface.
1136 		 * (- For site-local destination addresses, the
1137 		 *    set of candidate source addresses MUST only
1138 		 *    include addresses assigned to interfaces
1139 		 *    belonging to the same site as the outgoing
1140 		 *    interface.)
1141 		 */
1142 		if (((dst_type & IPV6_ADDR_MULTICAST) ||
1143 		     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1144 		    dst.ifindex && dev->ifindex != dst.ifindex)
1145 			continue;
1146 
1147 		idev = __in6_dev_get(dev);
1148 		if (!idev)
1149 			continue;
1150 
1151 		read_lock_bh(&idev->lock);
1152 		list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1153 			int i;
1154 
1155 			/*
1156 			 * - Tentative Address (RFC2462 section 5.4)
1157 			 *  - A tentative address is not considered
1158 			 *    "assigned to an interface" in the traditional
1159 			 *    sense, unless it is also flagged as optimistic.
1160 			 * - Candidate Source Address (section 4)
1161 			 *  - In any case, anycast addresses, multicast
1162 			 *    addresses, and the unspecified address MUST
1163 			 *    NOT be included in a candidate set.
1164 			 */
1165 			if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1166 			    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1167 				continue;
1168 
1169 			score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1170 
1171 			if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1172 				     score->addr_type & IPV6_ADDR_MULTICAST)) {
1173 				LIMIT_NETDEBUG(KERN_DEBUG
1174 					       "ADDRCONF: unspecified / multicast address "
1175 					       "assigned as unicast address on %s",
1176 					       dev->name);
1177 				continue;
1178 			}
1179 
1180 			score->rule = -1;
1181 			bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1182 
1183 			for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1184 				int minihiscore, miniscore;
1185 
1186 				minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1187 				miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1188 
1189 				if (minihiscore > miniscore) {
1190 					if (i == IPV6_SADDR_RULE_SCOPE &&
1191 					    score->scopedist > 0) {
1192 						/*
1193 						 * special case:
1194 						 * each remaining entry
1195 						 * has too small (not enough)
1196 						 * scope, because ifa entries
1197 						 * are sorted by their scope
1198 						 * values.
1199 						 */
1200 						goto try_nextdev;
1201 					}
1202 					break;
1203 				} else if (minihiscore < miniscore) {
1204 					if (hiscore->ifa)
1205 						in6_ifa_put(hiscore->ifa);
1206 
1207 					in6_ifa_hold(score->ifa);
1208 
1209 					swap(hiscore, score);
1210 
1211 					/* restore our iterator */
1212 					score->ifa = hiscore->ifa;
1213 
1214 					break;
1215 				}
1216 			}
1217 		}
1218 try_nextdev:
1219 		read_unlock_bh(&idev->lock);
1220 	}
1221 	rcu_read_unlock();
1222 
1223 	if (!hiscore->ifa)
1224 		return -EADDRNOTAVAIL;
1225 
1226 	*saddr = hiscore->ifa->addr;
1227 	in6_ifa_put(hiscore->ifa);
1228 	return 0;
1229 }
1230 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1231 
1232 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1233 		    unsigned char banned_flags)
1234 {
1235 	struct inet6_dev *idev;
1236 	int err = -EADDRNOTAVAIL;
1237 
1238 	rcu_read_lock();
1239 	idev = __in6_dev_get(dev);
1240 	if (idev) {
1241 		struct inet6_ifaddr *ifp;
1242 
1243 		read_lock_bh(&idev->lock);
1244 		list_for_each_entry(ifp, &idev->addr_list, if_list) {
1245 			if (ifp->scope == IFA_LINK &&
1246 			    !(ifp->flags & banned_flags)) {
1247 				*addr = ifp->addr;
1248 				err = 0;
1249 				break;
1250 			}
1251 		}
1252 		read_unlock_bh(&idev->lock);
1253 	}
1254 	rcu_read_unlock();
1255 	return err;
1256 }
1257 
1258 static int ipv6_count_addresses(struct inet6_dev *idev)
1259 {
1260 	int cnt = 0;
1261 	struct inet6_ifaddr *ifp;
1262 
1263 	read_lock_bh(&idev->lock);
1264 	list_for_each_entry(ifp, &idev->addr_list, if_list)
1265 		cnt++;
1266 	read_unlock_bh(&idev->lock);
1267 	return cnt;
1268 }
1269 
1270 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1271 		  struct net_device *dev, int strict)
1272 {
1273 	struct inet6_ifaddr *ifp;
1274 	struct hlist_node *node;
1275 	unsigned int hash = ipv6_addr_hash(addr);
1276 
1277 	rcu_read_lock_bh();
1278 	hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1279 		if (!net_eq(dev_net(ifp->idev->dev), net))
1280 			continue;
1281 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1282 		    !(ifp->flags&IFA_F_TENTATIVE) &&
1283 		    (dev == NULL || ifp->idev->dev == dev ||
1284 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1285 			rcu_read_unlock_bh();
1286 			return 1;
1287 		}
1288 	}
1289 
1290 	rcu_read_unlock_bh();
1291 	return 0;
1292 }
1293 EXPORT_SYMBOL(ipv6_chk_addr);
1294 
1295 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1296 			       struct net_device *dev)
1297 {
1298 	unsigned int hash = ipv6_addr_hash(addr);
1299 	struct inet6_ifaddr *ifp;
1300 	struct hlist_node *node;
1301 
1302 	hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1303 		if (!net_eq(dev_net(ifp->idev->dev), net))
1304 			continue;
1305 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1306 			if (dev == NULL || ifp->idev->dev == dev)
1307 				return true;
1308 		}
1309 	}
1310 	return false;
1311 }
1312 
1313 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1314 {
1315 	struct inet6_dev *idev;
1316 	struct inet6_ifaddr *ifa;
1317 	int	onlink;
1318 
1319 	onlink = 0;
1320 	rcu_read_lock();
1321 	idev = __in6_dev_get(dev);
1322 	if (idev) {
1323 		read_lock_bh(&idev->lock);
1324 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1325 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1326 						   ifa->prefix_len);
1327 			if (onlink)
1328 				break;
1329 		}
1330 		read_unlock_bh(&idev->lock);
1331 	}
1332 	rcu_read_unlock();
1333 	return onlink;
1334 }
1335 EXPORT_SYMBOL(ipv6_chk_prefix);
1336 
1337 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1338 				     struct net_device *dev, int strict)
1339 {
1340 	struct inet6_ifaddr *ifp, *result = NULL;
1341 	unsigned int hash = ipv6_addr_hash(addr);
1342 	struct hlist_node *node;
1343 
1344 	rcu_read_lock_bh();
1345 	hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1346 		if (!net_eq(dev_net(ifp->idev->dev), net))
1347 			continue;
1348 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1349 			if (dev == NULL || ifp->idev->dev == dev ||
1350 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1351 				result = ifp;
1352 				in6_ifa_hold(ifp);
1353 				break;
1354 			}
1355 		}
1356 	}
1357 	rcu_read_unlock_bh();
1358 
1359 	return result;
1360 }
1361 
1362 /* Gets referenced address, destroys ifaddr */
1363 
1364 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1365 {
1366 	if (ifp->flags&IFA_F_PERMANENT) {
1367 		spin_lock_bh(&ifp->lock);
1368 		addrconf_del_timer(ifp);
1369 		ifp->flags |= IFA_F_TENTATIVE;
1370 		if (dad_failed)
1371 			ifp->flags |= IFA_F_DADFAILED;
1372 		spin_unlock_bh(&ifp->lock);
1373 		if (dad_failed)
1374 			ipv6_ifa_notify(0, ifp);
1375 		in6_ifa_put(ifp);
1376 #ifdef CONFIG_IPV6_PRIVACY
1377 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1378 		struct inet6_ifaddr *ifpub;
1379 		spin_lock_bh(&ifp->lock);
1380 		ifpub = ifp->ifpub;
1381 		if (ifpub) {
1382 			in6_ifa_hold(ifpub);
1383 			spin_unlock_bh(&ifp->lock);
1384 			ipv6_create_tempaddr(ifpub, ifp);
1385 			in6_ifa_put(ifpub);
1386 		} else {
1387 			spin_unlock_bh(&ifp->lock);
1388 		}
1389 		ipv6_del_addr(ifp);
1390 #endif
1391 	} else
1392 		ipv6_del_addr(ifp);
1393 }
1394 
1395 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1396 {
1397 	int err = -ENOENT;
1398 
1399 	spin_lock(&ifp->state_lock);
1400 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
1401 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
1402 		err = 0;
1403 	}
1404 	spin_unlock(&ifp->state_lock);
1405 
1406 	return err;
1407 }
1408 
1409 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1410 {
1411 	struct inet6_dev *idev = ifp->idev;
1412 
1413 	if (addrconf_dad_end(ifp)) {
1414 		in6_ifa_put(ifp);
1415 		return;
1416 	}
1417 
1418 	if (net_ratelimit())
1419 		printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1420 			ifp->idev->dev->name, &ifp->addr);
1421 
1422 	if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1423 		struct in6_addr addr;
1424 
1425 		addr.s6_addr32[0] = htonl(0xfe800000);
1426 		addr.s6_addr32[1] = 0;
1427 
1428 		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1429 		    ipv6_addr_equal(&ifp->addr, &addr)) {
1430 			/* DAD failed for link-local based on MAC address */
1431 			idev->cnf.disable_ipv6 = 1;
1432 
1433 			printk(KERN_INFO "%s: IPv6 being disabled!\n",
1434 				ifp->idev->dev->name);
1435 		}
1436 	}
1437 
1438 	addrconf_dad_stop(ifp, 1);
1439 }
1440 
1441 /* Join to solicited addr multicast group. */
1442 
1443 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1444 {
1445 	struct in6_addr maddr;
1446 
1447 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1448 		return;
1449 
1450 	addrconf_addr_solict_mult(addr, &maddr);
1451 	ipv6_dev_mc_inc(dev, &maddr);
1452 }
1453 
1454 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1455 {
1456 	struct in6_addr maddr;
1457 
1458 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1459 		return;
1460 
1461 	addrconf_addr_solict_mult(addr, &maddr);
1462 	__ipv6_dev_mc_dec(idev, &maddr);
1463 }
1464 
1465 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1466 {
1467 	struct in6_addr addr;
1468 	if (ifp->prefix_len == 127) /* RFC 6164 */
1469 		return;
1470 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1471 	if (ipv6_addr_any(&addr))
1472 		return;
1473 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1474 }
1475 
1476 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1477 {
1478 	struct in6_addr addr;
1479 	if (ifp->prefix_len == 127) /* RFC 6164 */
1480 		return;
1481 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1482 	if (ipv6_addr_any(&addr))
1483 		return;
1484 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1485 }
1486 
1487 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1488 {
1489 	if (dev->addr_len != ETH_ALEN)
1490 		return -1;
1491 	memcpy(eui, dev->dev_addr, 3);
1492 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1493 
1494 	/*
1495 	 * The zSeries OSA network cards can be shared among various
1496 	 * OS instances, but the OSA cards have only one MAC address.
1497 	 * This leads to duplicate address conflicts in conjunction
1498 	 * with IPv6 if more than one instance uses the same card.
1499 	 *
1500 	 * The driver for these cards can deliver a unique 16-bit
1501 	 * identifier for each instance sharing the same card.  It is
1502 	 * placed instead of 0xFFFE in the interface identifier.  The
1503 	 * "u" bit of the interface identifier is not inverted in this
1504 	 * case.  Hence the resulting interface identifier has local
1505 	 * scope according to RFC2373.
1506 	 */
1507 	if (dev->dev_id) {
1508 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1509 		eui[4] = dev->dev_id & 0xFF;
1510 	} else {
1511 		eui[3] = 0xFF;
1512 		eui[4] = 0xFE;
1513 		eui[0] ^= 2;
1514 	}
1515 	return 0;
1516 }
1517 
1518 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1519 {
1520 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1521 	if (dev->addr_len != ARCNET_ALEN)
1522 		return -1;
1523 	memset(eui, 0, 7);
1524 	eui[7] = *(u8 *)dev->dev_addr;
1525 	return 0;
1526 }
1527 
1528 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1529 {
1530 	if (dev->addr_len != INFINIBAND_ALEN)
1531 		return -1;
1532 	memcpy(eui, dev->dev_addr + 12, 8);
1533 	eui[0] |= 2;
1534 	return 0;
1535 }
1536 
1537 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1538 {
1539 	if (addr == 0)
1540 		return -1;
1541 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1542 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1543 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1544 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1545 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1546 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1547 	eui[1] = 0;
1548 	eui[2] = 0x5E;
1549 	eui[3] = 0xFE;
1550 	memcpy(eui + 4, &addr, 4);
1551 	return 0;
1552 }
1553 
1554 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1555 {
1556 	if (dev->priv_flags & IFF_ISATAP)
1557 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1558 	return -1;
1559 }
1560 
1561 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1562 {
1563 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1564 }
1565 
1566 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1567 {
1568 	switch (dev->type) {
1569 	case ARPHRD_ETHER:
1570 	case ARPHRD_FDDI:
1571 	case ARPHRD_IEEE802_TR:
1572 		return addrconf_ifid_eui48(eui, dev);
1573 	case ARPHRD_ARCNET:
1574 		return addrconf_ifid_arcnet(eui, dev);
1575 	case ARPHRD_INFINIBAND:
1576 		return addrconf_ifid_infiniband(eui, dev);
1577 	case ARPHRD_SIT:
1578 		return addrconf_ifid_sit(eui, dev);
1579 	case ARPHRD_IPGRE:
1580 		return addrconf_ifid_gre(eui, dev);
1581 	}
1582 	return -1;
1583 }
1584 
1585 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1586 {
1587 	int err = -1;
1588 	struct inet6_ifaddr *ifp;
1589 
1590 	read_lock_bh(&idev->lock);
1591 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
1592 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1593 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1594 			err = 0;
1595 			break;
1596 		}
1597 	}
1598 	read_unlock_bh(&idev->lock);
1599 	return err;
1600 }
1601 
1602 #ifdef CONFIG_IPV6_PRIVACY
1603 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1604 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1605 {
1606 regen:
1607 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1608 	idev->rndid[0] &= ~0x02;
1609 
1610 	/*
1611 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1612 	 * check if generated address is not inappropriate
1613 	 *
1614 	 *  - Reserved subnet anycast (RFC 2526)
1615 	 *	11111101 11....11 1xxxxxxx
1616 	 *  - ISATAP (RFC4214) 6.1
1617 	 *	00-00-5E-FE-xx-xx-xx-xx
1618 	 *  - value 0
1619 	 *  - XXX: already assigned to an address on the device
1620 	 */
1621 	if (idev->rndid[0] == 0xfd &&
1622 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1623 	    (idev->rndid[7]&0x80))
1624 		goto regen;
1625 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1626 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1627 			goto regen;
1628 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1629 			goto regen;
1630 	}
1631 
1632 	return 0;
1633 }
1634 
1635 static void ipv6_regen_rndid(unsigned long data)
1636 {
1637 	struct inet6_dev *idev = (struct inet6_dev *) data;
1638 	unsigned long expires;
1639 
1640 	rcu_read_lock_bh();
1641 	write_lock_bh(&idev->lock);
1642 
1643 	if (idev->dead)
1644 		goto out;
1645 
1646 	if (__ipv6_regen_rndid(idev) < 0)
1647 		goto out;
1648 
1649 	expires = jiffies +
1650 		idev->cnf.temp_prefered_lft * HZ -
1651 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1652 		idev->cnf.max_desync_factor * HZ;
1653 	if (time_before(expires, jiffies)) {
1654 		printk(KERN_WARNING
1655 			"ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1656 			idev->dev->name);
1657 		goto out;
1658 	}
1659 
1660 	if (!mod_timer(&idev->regen_timer, expires))
1661 		in6_dev_hold(idev);
1662 
1663 out:
1664 	write_unlock_bh(&idev->lock);
1665 	rcu_read_unlock_bh();
1666 	in6_dev_put(idev);
1667 }
1668 
1669 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1670 {
1671 	int ret = 0;
1672 
1673 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1674 		ret = __ipv6_regen_rndid(idev);
1675 	return ret;
1676 }
1677 #endif
1678 
1679 /*
1680  *	Add prefix route.
1681  */
1682 
1683 static void
1684 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1685 		      unsigned long expires, u32 flags)
1686 {
1687 	struct fib6_config cfg = {
1688 		.fc_table = RT6_TABLE_PREFIX,
1689 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1690 		.fc_ifindex = dev->ifindex,
1691 		.fc_expires = expires,
1692 		.fc_dst_len = plen,
1693 		.fc_flags = RTF_UP | flags,
1694 		.fc_nlinfo.nl_net = dev_net(dev),
1695 		.fc_protocol = RTPROT_KERNEL,
1696 	};
1697 
1698 	cfg.fc_dst = *pfx;
1699 
1700 	/* Prevent useless cloning on PtP SIT.
1701 	   This thing is done here expecting that the whole
1702 	   class of non-broadcast devices need not cloning.
1703 	 */
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1706 		cfg.fc_flags |= RTF_NONEXTHOP;
1707 #endif
1708 
1709 	ip6_route_add(&cfg);
1710 }
1711 
1712 
1713 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1714 						  int plen,
1715 						  const struct net_device *dev,
1716 						  u32 flags, u32 noflags)
1717 {
1718 	struct fib6_node *fn;
1719 	struct rt6_info *rt = NULL;
1720 	struct fib6_table *table;
1721 
1722 	table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1723 	if (table == NULL)
1724 		return NULL;
1725 
1726 	write_lock_bh(&table->tb6_lock);
1727 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1728 	if (!fn)
1729 		goto out;
1730 	for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1731 		if (rt->dst.dev->ifindex != dev->ifindex)
1732 			continue;
1733 		if ((rt->rt6i_flags & flags) != flags)
1734 			continue;
1735 		if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1736 			continue;
1737 		dst_hold(&rt->dst);
1738 		break;
1739 	}
1740 out:
1741 	write_unlock_bh(&table->tb6_lock);
1742 	return rt;
1743 }
1744 
1745 
1746 /* Create "default" multicast route to the interface */
1747 
1748 static void addrconf_add_mroute(struct net_device *dev)
1749 {
1750 	struct fib6_config cfg = {
1751 		.fc_table = RT6_TABLE_LOCAL,
1752 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1753 		.fc_ifindex = dev->ifindex,
1754 		.fc_dst_len = 8,
1755 		.fc_flags = RTF_UP,
1756 		.fc_nlinfo.nl_net = dev_net(dev),
1757 	};
1758 
1759 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1760 
1761 	ip6_route_add(&cfg);
1762 }
1763 
1764 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1765 static void sit_route_add(struct net_device *dev)
1766 {
1767 	struct fib6_config cfg = {
1768 		.fc_table = RT6_TABLE_MAIN,
1769 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1770 		.fc_ifindex = dev->ifindex,
1771 		.fc_dst_len = 96,
1772 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
1773 		.fc_nlinfo.nl_net = dev_net(dev),
1774 	};
1775 
1776 	/* prefix length - 96 bits "::d.d.d.d" */
1777 	ip6_route_add(&cfg);
1778 }
1779 #endif
1780 
1781 static void addrconf_add_lroute(struct net_device *dev)
1782 {
1783 	struct in6_addr addr;
1784 
1785 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1786 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
1787 }
1788 
1789 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1790 {
1791 	struct inet6_dev *idev;
1792 
1793 	ASSERT_RTNL();
1794 
1795 	idev = ipv6_find_idev(dev);
1796 	if (!idev)
1797 		return ERR_PTR(-ENOBUFS);
1798 
1799 	if (idev->cnf.disable_ipv6)
1800 		return ERR_PTR(-EACCES);
1801 
1802 	/* Add default multicast route */
1803 	if (!(dev->flags & IFF_LOOPBACK))
1804 		addrconf_add_mroute(dev);
1805 
1806 	/* Add link local route */
1807 	addrconf_add_lroute(dev);
1808 	return idev;
1809 }
1810 
1811 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1812 {
1813 	struct prefix_info *pinfo;
1814 	__u32 valid_lft;
1815 	__u32 prefered_lft;
1816 	int addr_type;
1817 	struct inet6_dev *in6_dev;
1818 	struct net *net = dev_net(dev);
1819 
1820 	pinfo = (struct prefix_info *) opt;
1821 
1822 	if (len < sizeof(struct prefix_info)) {
1823 		ADBG(("addrconf: prefix option too short\n"));
1824 		return;
1825 	}
1826 
1827 	/*
1828 	 *	Validation checks ([ADDRCONF], page 19)
1829 	 */
1830 
1831 	addr_type = ipv6_addr_type(&pinfo->prefix);
1832 
1833 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1834 		return;
1835 
1836 	valid_lft = ntohl(pinfo->valid);
1837 	prefered_lft = ntohl(pinfo->prefered);
1838 
1839 	if (prefered_lft > valid_lft) {
1840 		if (net_ratelimit())
1841 			printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1842 		return;
1843 	}
1844 
1845 	in6_dev = in6_dev_get(dev);
1846 
1847 	if (in6_dev == NULL) {
1848 		if (net_ratelimit())
1849 			printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1850 		return;
1851 	}
1852 
1853 	/*
1854 	 *	Two things going on here:
1855 	 *	1) Add routes for on-link prefixes
1856 	 *	2) Configure prefixes with the auto flag set
1857 	 */
1858 
1859 	if (pinfo->onlink) {
1860 		struct rt6_info *rt;
1861 		unsigned long rt_expires;
1862 
1863 		/* Avoid arithmetic overflow. Really, we could
1864 		 * save rt_expires in seconds, likely valid_lft,
1865 		 * but it would require division in fib gc, that it
1866 		 * not good.
1867 		 */
1868 		if (HZ > USER_HZ)
1869 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1870 		else
1871 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1872 
1873 		if (addrconf_finite_timeout(rt_expires))
1874 			rt_expires *= HZ;
1875 
1876 		rt = addrconf_get_prefix_route(&pinfo->prefix,
1877 					       pinfo->prefix_len,
1878 					       dev,
1879 					       RTF_ADDRCONF | RTF_PREFIX_RT,
1880 					       RTF_GATEWAY | RTF_DEFAULT);
1881 
1882 		if (rt) {
1883 			/* Autoconf prefix route */
1884 			if (valid_lft == 0) {
1885 				ip6_del_rt(rt);
1886 				rt = NULL;
1887 			} else if (addrconf_finite_timeout(rt_expires)) {
1888 				/* not infinity */
1889 				rt->dst.expires = jiffies + rt_expires;
1890 				rt->rt6i_flags |= RTF_EXPIRES;
1891 			} else {
1892 				rt->rt6i_flags &= ~RTF_EXPIRES;
1893 				rt->dst.expires = 0;
1894 			}
1895 		} else if (valid_lft) {
1896 			clock_t expires = 0;
1897 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1898 			if (addrconf_finite_timeout(rt_expires)) {
1899 				/* not infinity */
1900 				flags |= RTF_EXPIRES;
1901 				expires = jiffies_to_clock_t(rt_expires);
1902 			}
1903 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1904 					      dev, expires, flags);
1905 		}
1906 		if (rt)
1907 			dst_release(&rt->dst);
1908 	}
1909 
1910 	/* Try to figure out our local address for this prefix */
1911 
1912 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1913 		struct inet6_ifaddr *ifp;
1914 		struct in6_addr addr;
1915 		int create = 0, update_lft = 0;
1916 
1917 		if (pinfo->prefix_len == 64) {
1918 			memcpy(&addr, &pinfo->prefix, 8);
1919 			if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1920 			    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1921 				in6_dev_put(in6_dev);
1922 				return;
1923 			}
1924 			goto ok;
1925 		}
1926 		if (net_ratelimit())
1927 			printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1928 			       pinfo->prefix_len);
1929 		in6_dev_put(in6_dev);
1930 		return;
1931 
1932 ok:
1933 
1934 		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1935 
1936 		if (ifp == NULL && valid_lft) {
1937 			int max_addresses = in6_dev->cnf.max_addresses;
1938 			u32 addr_flags = 0;
1939 
1940 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1941 			if (in6_dev->cnf.optimistic_dad &&
1942 			    !net->ipv6.devconf_all->forwarding && sllao)
1943 				addr_flags = IFA_F_OPTIMISTIC;
1944 #endif
1945 
1946 			/* Do not allow to create too much of autoconfigured
1947 			 * addresses; this would be too easy way to crash kernel.
1948 			 */
1949 			if (!max_addresses ||
1950 			    ipv6_count_addresses(in6_dev) < max_addresses)
1951 				ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1952 						    addr_type&IPV6_ADDR_SCOPE_MASK,
1953 						    addr_flags);
1954 
1955 			if (!ifp || IS_ERR(ifp)) {
1956 				in6_dev_put(in6_dev);
1957 				return;
1958 			}
1959 
1960 			update_lft = create = 1;
1961 			ifp->cstamp = jiffies;
1962 			addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1963 		}
1964 
1965 		if (ifp) {
1966 			int flags;
1967 			unsigned long now;
1968 #ifdef CONFIG_IPV6_PRIVACY
1969 			struct inet6_ifaddr *ift;
1970 #endif
1971 			u32 stored_lft;
1972 
1973 			/* update lifetime (RFC2462 5.5.3 e) */
1974 			spin_lock(&ifp->lock);
1975 			now = jiffies;
1976 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1977 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1978 			else
1979 				stored_lft = 0;
1980 			if (!update_lft && stored_lft) {
1981 				if (valid_lft > MIN_VALID_LIFETIME ||
1982 				    valid_lft > stored_lft)
1983 					update_lft = 1;
1984 				else if (stored_lft <= MIN_VALID_LIFETIME) {
1985 					/* valid_lft <= stored_lft is always true */
1986 					/*
1987 					 * RFC 4862 Section 5.5.3e:
1988 					 * "Note that the preferred lifetime of
1989 					 *  the corresponding address is always
1990 					 *  reset to the Preferred Lifetime in
1991 					 *  the received Prefix Information
1992 					 *  option, regardless of whether the
1993 					 *  valid lifetime is also reset or
1994 					 *  ignored."
1995 					 *
1996 					 *  So if the preferred lifetime in
1997 					 *  this advertisement is different
1998 					 *  than what we have stored, but the
1999 					 *  valid lifetime is invalid, just
2000 					 *  reset prefered_lft.
2001 					 *
2002 					 *  We must set the valid lifetime
2003 					 *  to the stored lifetime since we'll
2004 					 *  be updating the timestamp below,
2005 					 *  else we'll set it back to the
2006 					 *  minimum.
2007 					 */
2008 					if (prefered_lft != ifp->prefered_lft) {
2009 						valid_lft = stored_lft;
2010 						update_lft = 1;
2011 					}
2012 				} else {
2013 					valid_lft = MIN_VALID_LIFETIME;
2014 					if (valid_lft < prefered_lft)
2015 						prefered_lft = valid_lft;
2016 					update_lft = 1;
2017 				}
2018 			}
2019 
2020 			if (update_lft) {
2021 				ifp->valid_lft = valid_lft;
2022 				ifp->prefered_lft = prefered_lft;
2023 				ifp->tstamp = now;
2024 				flags = ifp->flags;
2025 				ifp->flags &= ~IFA_F_DEPRECATED;
2026 				spin_unlock(&ifp->lock);
2027 
2028 				if (!(flags&IFA_F_TENTATIVE))
2029 					ipv6_ifa_notify(0, ifp);
2030 			} else
2031 				spin_unlock(&ifp->lock);
2032 
2033 #ifdef CONFIG_IPV6_PRIVACY
2034 			read_lock_bh(&in6_dev->lock);
2035 			/* update all temporary addresses in the list */
2036 			list_for_each_entry(ift, &in6_dev->tempaddr_list,
2037 					    tmp_list) {
2038 				int age, max_valid, max_prefered;
2039 
2040 				if (ifp != ift->ifpub)
2041 					continue;
2042 
2043 				/*
2044 				 * RFC 4941 section 3.3:
2045 				 * If a received option will extend the lifetime
2046 				 * of a public address, the lifetimes of
2047 				 * temporary addresses should be extended,
2048 				 * subject to the overall constraint that no
2049 				 * temporary addresses should ever remain
2050 				 * "valid" or "preferred" for a time longer than
2051 				 * (TEMP_VALID_LIFETIME) or
2052 				 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2053 				 * respectively.
2054 				 */
2055 				age = (now - ift->cstamp) / HZ;
2056 				max_valid = in6_dev->cnf.temp_valid_lft - age;
2057 				if (max_valid < 0)
2058 					max_valid = 0;
2059 
2060 				max_prefered = in6_dev->cnf.temp_prefered_lft -
2061 					       in6_dev->cnf.max_desync_factor -
2062 					       age;
2063 				if (max_prefered < 0)
2064 					max_prefered = 0;
2065 
2066 				if (valid_lft > max_valid)
2067 					valid_lft = max_valid;
2068 
2069 				if (prefered_lft > max_prefered)
2070 					prefered_lft = max_prefered;
2071 
2072 				spin_lock(&ift->lock);
2073 				flags = ift->flags;
2074 				ift->valid_lft = valid_lft;
2075 				ift->prefered_lft = prefered_lft;
2076 				ift->tstamp = now;
2077 				if (prefered_lft > 0)
2078 					ift->flags &= ~IFA_F_DEPRECATED;
2079 
2080 				spin_unlock(&ift->lock);
2081 				if (!(flags&IFA_F_TENTATIVE))
2082 					ipv6_ifa_notify(0, ift);
2083 			}
2084 
2085 			if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2086 				/*
2087 				 * When a new public address is created as
2088 				 * described in [ADDRCONF], also create a new
2089 				 * temporary address. Also create a temporary
2090 				 * address if it's enabled but no temporary
2091 				 * address currently exists.
2092 				 */
2093 				read_unlock_bh(&in6_dev->lock);
2094 				ipv6_create_tempaddr(ifp, NULL);
2095 			} else {
2096 				read_unlock_bh(&in6_dev->lock);
2097 			}
2098 #endif
2099 			in6_ifa_put(ifp);
2100 			addrconf_verify(0);
2101 		}
2102 	}
2103 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2104 	in6_dev_put(in6_dev);
2105 }
2106 
2107 /*
2108  *	Set destination address.
2109  *	Special case for SIT interfaces where we create a new "virtual"
2110  *	device.
2111  */
2112 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2113 {
2114 	struct in6_ifreq ireq;
2115 	struct net_device *dev;
2116 	int err = -EINVAL;
2117 
2118 	rtnl_lock();
2119 
2120 	err = -EFAULT;
2121 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2122 		goto err_exit;
2123 
2124 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2125 
2126 	err = -ENODEV;
2127 	if (dev == NULL)
2128 		goto err_exit;
2129 
2130 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2131 	if (dev->type == ARPHRD_SIT) {
2132 		const struct net_device_ops *ops = dev->netdev_ops;
2133 		struct ifreq ifr;
2134 		struct ip_tunnel_parm p;
2135 
2136 		err = -EADDRNOTAVAIL;
2137 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2138 			goto err_exit;
2139 
2140 		memset(&p, 0, sizeof(p));
2141 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2142 		p.iph.saddr = 0;
2143 		p.iph.version = 4;
2144 		p.iph.ihl = 5;
2145 		p.iph.protocol = IPPROTO_IPV6;
2146 		p.iph.ttl = 64;
2147 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2148 
2149 		if (ops->ndo_do_ioctl) {
2150 			mm_segment_t oldfs = get_fs();
2151 
2152 			set_fs(KERNEL_DS);
2153 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2154 			set_fs(oldfs);
2155 		} else
2156 			err = -EOPNOTSUPP;
2157 
2158 		if (err == 0) {
2159 			err = -ENOBUFS;
2160 			dev = __dev_get_by_name(net, p.name);
2161 			if (!dev)
2162 				goto err_exit;
2163 			err = dev_open(dev);
2164 		}
2165 	}
2166 #endif
2167 
2168 err_exit:
2169 	rtnl_unlock();
2170 	return err;
2171 }
2172 
2173 /*
2174  *	Manual configuration of address on an interface
2175  */
2176 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2177 			  unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2178 			  __u32 valid_lft)
2179 {
2180 	struct inet6_ifaddr *ifp;
2181 	struct inet6_dev *idev;
2182 	struct net_device *dev;
2183 	int scope;
2184 	u32 flags;
2185 	clock_t expires;
2186 	unsigned long timeout;
2187 
2188 	ASSERT_RTNL();
2189 
2190 	if (plen > 128)
2191 		return -EINVAL;
2192 
2193 	/* check the lifetime */
2194 	if (!valid_lft || prefered_lft > valid_lft)
2195 		return -EINVAL;
2196 
2197 	dev = __dev_get_by_index(net, ifindex);
2198 	if (!dev)
2199 		return -ENODEV;
2200 
2201 	idev = addrconf_add_dev(dev);
2202 	if (IS_ERR(idev))
2203 		return PTR_ERR(idev);
2204 
2205 	scope = ipv6_addr_scope(pfx);
2206 
2207 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2208 	if (addrconf_finite_timeout(timeout)) {
2209 		expires = jiffies_to_clock_t(timeout * HZ);
2210 		valid_lft = timeout;
2211 		flags = RTF_EXPIRES;
2212 	} else {
2213 		expires = 0;
2214 		flags = 0;
2215 		ifa_flags |= IFA_F_PERMANENT;
2216 	}
2217 
2218 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2219 	if (addrconf_finite_timeout(timeout)) {
2220 		if (timeout == 0)
2221 			ifa_flags |= IFA_F_DEPRECATED;
2222 		prefered_lft = timeout;
2223 	}
2224 
2225 	ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2226 
2227 	if (!IS_ERR(ifp)) {
2228 		spin_lock_bh(&ifp->lock);
2229 		ifp->valid_lft = valid_lft;
2230 		ifp->prefered_lft = prefered_lft;
2231 		ifp->tstamp = jiffies;
2232 		spin_unlock_bh(&ifp->lock);
2233 
2234 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2235 				      expires, flags);
2236 		/*
2237 		 * Note that section 3.1 of RFC 4429 indicates
2238 		 * that the Optimistic flag should not be set for
2239 		 * manually configured addresses
2240 		 */
2241 		addrconf_dad_start(ifp, 0);
2242 		in6_ifa_put(ifp);
2243 		addrconf_verify(0);
2244 		return 0;
2245 	}
2246 
2247 	return PTR_ERR(ifp);
2248 }
2249 
2250 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2251 			  unsigned int plen)
2252 {
2253 	struct inet6_ifaddr *ifp;
2254 	struct inet6_dev *idev;
2255 	struct net_device *dev;
2256 
2257 	if (plen > 128)
2258 		return -EINVAL;
2259 
2260 	dev = __dev_get_by_index(net, ifindex);
2261 	if (!dev)
2262 		return -ENODEV;
2263 
2264 	if ((idev = __in6_dev_get(dev)) == NULL)
2265 		return -ENXIO;
2266 
2267 	read_lock_bh(&idev->lock);
2268 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2269 		if (ifp->prefix_len == plen &&
2270 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2271 			in6_ifa_hold(ifp);
2272 			read_unlock_bh(&idev->lock);
2273 
2274 			ipv6_del_addr(ifp);
2275 
2276 			/* If the last address is deleted administratively,
2277 			   disable IPv6 on this interface.
2278 			 */
2279 			if (list_empty(&idev->addr_list))
2280 				addrconf_ifdown(idev->dev, 1);
2281 			return 0;
2282 		}
2283 	}
2284 	read_unlock_bh(&idev->lock);
2285 	return -EADDRNOTAVAIL;
2286 }
2287 
2288 
2289 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2290 {
2291 	struct in6_ifreq ireq;
2292 	int err;
2293 
2294 	if (!capable(CAP_NET_ADMIN))
2295 		return -EPERM;
2296 
2297 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2298 		return -EFAULT;
2299 
2300 	rtnl_lock();
2301 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2302 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2303 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2304 	rtnl_unlock();
2305 	return err;
2306 }
2307 
2308 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2309 {
2310 	struct in6_ifreq ireq;
2311 	int err;
2312 
2313 	if (!capable(CAP_NET_ADMIN))
2314 		return -EPERM;
2315 
2316 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2317 		return -EFAULT;
2318 
2319 	rtnl_lock();
2320 	err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2321 			     ireq.ifr6_prefixlen);
2322 	rtnl_unlock();
2323 	return err;
2324 }
2325 
2326 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2327 		     int plen, int scope)
2328 {
2329 	struct inet6_ifaddr *ifp;
2330 
2331 	ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2332 	if (!IS_ERR(ifp)) {
2333 		spin_lock_bh(&ifp->lock);
2334 		ifp->flags &= ~IFA_F_TENTATIVE;
2335 		spin_unlock_bh(&ifp->lock);
2336 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2337 		in6_ifa_put(ifp);
2338 	}
2339 }
2340 
2341 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2342 static void sit_add_v4_addrs(struct inet6_dev *idev)
2343 {
2344 	struct in6_addr addr;
2345 	struct net_device *dev;
2346 	struct net *net = dev_net(idev->dev);
2347 	int scope;
2348 
2349 	ASSERT_RTNL();
2350 
2351 	memset(&addr, 0, sizeof(struct in6_addr));
2352 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2353 
2354 	if (idev->dev->flags&IFF_POINTOPOINT) {
2355 		addr.s6_addr32[0] = htonl(0xfe800000);
2356 		scope = IFA_LINK;
2357 	} else {
2358 		scope = IPV6_ADDR_COMPATv4;
2359 	}
2360 
2361 	if (addr.s6_addr32[3]) {
2362 		add_addr(idev, &addr, 128, scope);
2363 		return;
2364 	}
2365 
2366 	for_each_netdev(net, dev) {
2367 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
2368 		if (in_dev && (dev->flags & IFF_UP)) {
2369 			struct in_ifaddr *ifa;
2370 
2371 			int flag = scope;
2372 
2373 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2374 				int plen;
2375 
2376 				addr.s6_addr32[3] = ifa->ifa_local;
2377 
2378 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2379 					continue;
2380 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2381 					if (idev->dev->flags&IFF_POINTOPOINT)
2382 						continue;
2383 					flag |= IFA_HOST;
2384 				}
2385 				if (idev->dev->flags&IFF_POINTOPOINT)
2386 					plen = 64;
2387 				else
2388 					plen = 96;
2389 
2390 				add_addr(idev, &addr, plen, flag);
2391 			}
2392 		}
2393 	}
2394 }
2395 #endif
2396 
2397 static void init_loopback(struct net_device *dev)
2398 {
2399 	struct inet6_dev  *idev;
2400 
2401 	/* ::1 */
2402 
2403 	ASSERT_RTNL();
2404 
2405 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2406 		printk(KERN_DEBUG "init loopback: add_dev failed\n");
2407 		return;
2408 	}
2409 
2410 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2411 }
2412 
2413 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2414 {
2415 	struct inet6_ifaddr *ifp;
2416 	u32 addr_flags = IFA_F_PERMANENT;
2417 
2418 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2419 	if (idev->cnf.optimistic_dad &&
2420 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2421 		addr_flags |= IFA_F_OPTIMISTIC;
2422 #endif
2423 
2424 
2425 	ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2426 	if (!IS_ERR(ifp)) {
2427 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2428 		addrconf_dad_start(ifp, 0);
2429 		in6_ifa_put(ifp);
2430 	}
2431 }
2432 
2433 static void addrconf_dev_config(struct net_device *dev)
2434 {
2435 	struct in6_addr addr;
2436 	struct inet6_dev *idev;
2437 
2438 	ASSERT_RTNL();
2439 
2440 	if ((dev->type != ARPHRD_ETHER) &&
2441 	    (dev->type != ARPHRD_FDDI) &&
2442 	    (dev->type != ARPHRD_IEEE802_TR) &&
2443 	    (dev->type != ARPHRD_ARCNET) &&
2444 	    (dev->type != ARPHRD_INFINIBAND)) {
2445 		/* Alas, we support only Ethernet autoconfiguration. */
2446 		return;
2447 	}
2448 
2449 	idev = addrconf_add_dev(dev);
2450 	if (IS_ERR(idev))
2451 		return;
2452 
2453 	memset(&addr, 0, sizeof(struct in6_addr));
2454 	addr.s6_addr32[0] = htonl(0xFE800000);
2455 
2456 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2457 		addrconf_add_linklocal(idev, &addr);
2458 }
2459 
2460 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2461 static void addrconf_sit_config(struct net_device *dev)
2462 {
2463 	struct inet6_dev *idev;
2464 
2465 	ASSERT_RTNL();
2466 
2467 	/*
2468 	 * Configure the tunnel with one of our IPv4
2469 	 * addresses... we should configure all of
2470 	 * our v4 addrs in the tunnel
2471 	 */
2472 
2473 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2474 		printk(KERN_DEBUG "init sit: add_dev failed\n");
2475 		return;
2476 	}
2477 
2478 	if (dev->priv_flags & IFF_ISATAP) {
2479 		struct in6_addr addr;
2480 
2481 		ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2482 		addrconf_prefix_route(&addr, 64, dev, 0, 0);
2483 		if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2484 			addrconf_add_linklocal(idev, &addr);
2485 		return;
2486 	}
2487 
2488 	sit_add_v4_addrs(idev);
2489 
2490 	if (dev->flags&IFF_POINTOPOINT) {
2491 		addrconf_add_mroute(dev);
2492 		addrconf_add_lroute(dev);
2493 	} else
2494 		sit_route_add(dev);
2495 }
2496 #endif
2497 
2498 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2499 static void addrconf_gre_config(struct net_device *dev)
2500 {
2501 	struct inet6_dev *idev;
2502 	struct in6_addr addr;
2503 
2504 	pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2505 
2506 	ASSERT_RTNL();
2507 
2508 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2509 		printk(KERN_DEBUG "init gre: add_dev failed\n");
2510 		return;
2511 	}
2512 
2513 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2514 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
2515 
2516 	if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2517 		addrconf_add_linklocal(idev, &addr);
2518 }
2519 #endif
2520 
2521 static inline int
2522 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2523 {
2524 	struct in6_addr lladdr;
2525 
2526 	if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2527 		addrconf_add_linklocal(idev, &lladdr);
2528 		return 0;
2529 	}
2530 	return -1;
2531 }
2532 
2533 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2534 {
2535 	struct net_device *link_dev;
2536 	struct net *net = dev_net(idev->dev);
2537 
2538 	/* first try to inherit the link-local address from the link device */
2539 	if (idev->dev->iflink &&
2540 	    (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2541 		if (!ipv6_inherit_linklocal(idev, link_dev))
2542 			return;
2543 	}
2544 	/* then try to inherit it from any device */
2545 	for_each_netdev(net, link_dev) {
2546 		if (!ipv6_inherit_linklocal(idev, link_dev))
2547 			return;
2548 	}
2549 	printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2550 }
2551 
2552 /*
2553  * Autoconfigure tunnel with a link-local address so routing protocols,
2554  * DHCPv6, MLD etc. can be run over the virtual link
2555  */
2556 
2557 static void addrconf_ip6_tnl_config(struct net_device *dev)
2558 {
2559 	struct inet6_dev *idev;
2560 
2561 	ASSERT_RTNL();
2562 
2563 	idev = addrconf_add_dev(dev);
2564 	if (IS_ERR(idev)) {
2565 		printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2566 		return;
2567 	}
2568 	ip6_tnl_add_linklocal(idev);
2569 }
2570 
2571 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2572 			   void *data)
2573 {
2574 	struct net_device *dev = (struct net_device *) data;
2575 	struct inet6_dev *idev = __in6_dev_get(dev);
2576 	int run_pending = 0;
2577 	int err;
2578 
2579 	switch (event) {
2580 	case NETDEV_REGISTER:
2581 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2582 			idev = ipv6_add_dev(dev);
2583 			if (!idev)
2584 				return notifier_from_errno(-ENOMEM);
2585 		}
2586 		break;
2587 
2588 	case NETDEV_UP:
2589 	case NETDEV_CHANGE:
2590 		if (dev->flags & IFF_SLAVE)
2591 			break;
2592 
2593 		if (event == NETDEV_UP) {
2594 			if (!addrconf_qdisc_ok(dev)) {
2595 				/* device is not ready yet. */
2596 				printk(KERN_INFO
2597 					"ADDRCONF(NETDEV_UP): %s: "
2598 					"link is not ready\n",
2599 					dev->name);
2600 				break;
2601 			}
2602 
2603 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
2604 				idev = ipv6_add_dev(dev);
2605 
2606 			if (idev) {
2607 				idev->if_flags |= IF_READY;
2608 				run_pending = 1;
2609 			}
2610 		} else {
2611 			if (!addrconf_qdisc_ok(dev)) {
2612 				/* device is still not ready. */
2613 				break;
2614 			}
2615 
2616 			if (idev) {
2617 				if (idev->if_flags & IF_READY)
2618 					/* device is already configured. */
2619 					break;
2620 				idev->if_flags |= IF_READY;
2621 			}
2622 
2623 			printk(KERN_INFO
2624 					"ADDRCONF(NETDEV_CHANGE): %s: "
2625 					"link becomes ready\n",
2626 					dev->name);
2627 
2628 			run_pending = 1;
2629 		}
2630 
2631 		switch (dev->type) {
2632 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2633 		case ARPHRD_SIT:
2634 			addrconf_sit_config(dev);
2635 			break;
2636 #endif
2637 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2638 		case ARPHRD_IPGRE:
2639 			addrconf_gre_config(dev);
2640 			break;
2641 #endif
2642 		case ARPHRD_TUNNEL6:
2643 			addrconf_ip6_tnl_config(dev);
2644 			break;
2645 		case ARPHRD_LOOPBACK:
2646 			init_loopback(dev);
2647 			break;
2648 
2649 		default:
2650 			addrconf_dev_config(dev);
2651 			break;
2652 		}
2653 
2654 		if (idev) {
2655 			if (run_pending)
2656 				addrconf_dad_run(idev);
2657 
2658 			/*
2659 			 * If the MTU changed during the interface down,
2660 			 * when the interface up, the changed MTU must be
2661 			 * reflected in the idev as well as routers.
2662 			 */
2663 			if (idev->cnf.mtu6 != dev->mtu &&
2664 			    dev->mtu >= IPV6_MIN_MTU) {
2665 				rt6_mtu_change(dev, dev->mtu);
2666 				idev->cnf.mtu6 = dev->mtu;
2667 			}
2668 			idev->tstamp = jiffies;
2669 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2670 
2671 			/*
2672 			 * If the changed mtu during down is lower than
2673 			 * IPV6_MIN_MTU stop IPv6 on this interface.
2674 			 */
2675 			if (dev->mtu < IPV6_MIN_MTU)
2676 				addrconf_ifdown(dev, 1);
2677 		}
2678 		break;
2679 
2680 	case NETDEV_CHANGEMTU:
2681 		if (idev && dev->mtu >= IPV6_MIN_MTU) {
2682 			rt6_mtu_change(dev, dev->mtu);
2683 			idev->cnf.mtu6 = dev->mtu;
2684 			break;
2685 		}
2686 
2687 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2688 			idev = ipv6_add_dev(dev);
2689 			if (idev)
2690 				break;
2691 		}
2692 
2693 		/*
2694 		 * MTU falled under IPV6_MIN_MTU.
2695 		 * Stop IPv6 on this interface.
2696 		 */
2697 
2698 	case NETDEV_DOWN:
2699 	case NETDEV_UNREGISTER:
2700 		/*
2701 		 *	Remove all addresses from this interface.
2702 		 */
2703 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2704 		break;
2705 
2706 	case NETDEV_CHANGENAME:
2707 		if (idev) {
2708 			snmp6_unregister_dev(idev);
2709 			addrconf_sysctl_unregister(idev);
2710 			addrconf_sysctl_register(idev);
2711 			err = snmp6_register_dev(idev);
2712 			if (err)
2713 				return notifier_from_errno(err);
2714 		}
2715 		break;
2716 
2717 	case NETDEV_PRE_TYPE_CHANGE:
2718 	case NETDEV_POST_TYPE_CHANGE:
2719 		addrconf_type_change(dev, event);
2720 		break;
2721 	}
2722 
2723 	return NOTIFY_OK;
2724 }
2725 
2726 /*
2727  *	addrconf module should be notified of a device going up
2728  */
2729 static struct notifier_block ipv6_dev_notf = {
2730 	.notifier_call = addrconf_notify,
2731 };
2732 
2733 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2734 {
2735 	struct inet6_dev *idev;
2736 	ASSERT_RTNL();
2737 
2738 	idev = __in6_dev_get(dev);
2739 
2740 	if (event == NETDEV_POST_TYPE_CHANGE)
2741 		ipv6_mc_remap(idev);
2742 	else if (event == NETDEV_PRE_TYPE_CHANGE)
2743 		ipv6_mc_unmap(idev);
2744 }
2745 
2746 static int addrconf_ifdown(struct net_device *dev, int how)
2747 {
2748 	struct net *net = dev_net(dev);
2749 	struct inet6_dev *idev;
2750 	struct inet6_ifaddr *ifa;
2751 	int state, i;
2752 
2753 	ASSERT_RTNL();
2754 
2755 	rt6_ifdown(net, dev);
2756 	neigh_ifdown(&nd_tbl, dev);
2757 
2758 	idev = __in6_dev_get(dev);
2759 	if (idev == NULL)
2760 		return -ENODEV;
2761 
2762 	/*
2763 	 * Step 1: remove reference to ipv6 device from parent device.
2764 	 *	   Do not dev_put!
2765 	 */
2766 	if (how) {
2767 		idev->dead = 1;
2768 
2769 		/* protected by rtnl_lock */
2770 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2771 
2772 		/* Step 1.5: remove snmp6 entry */
2773 		snmp6_unregister_dev(idev);
2774 
2775 	}
2776 
2777 	/* Step 2: clear hash table */
2778 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2779 		struct hlist_head *h = &inet6_addr_lst[i];
2780 		struct hlist_node *n;
2781 
2782 		spin_lock_bh(&addrconf_hash_lock);
2783 	restart:
2784 		hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2785 			if (ifa->idev == idev) {
2786 				hlist_del_init_rcu(&ifa->addr_lst);
2787 				addrconf_del_timer(ifa);
2788 				goto restart;
2789 			}
2790 		}
2791 		spin_unlock_bh(&addrconf_hash_lock);
2792 	}
2793 
2794 	write_lock_bh(&idev->lock);
2795 
2796 	/* Step 2: clear flags for stateless addrconf */
2797 	if (!how)
2798 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2799 
2800 #ifdef CONFIG_IPV6_PRIVACY
2801 	if (how && del_timer(&idev->regen_timer))
2802 		in6_dev_put(idev);
2803 
2804 	/* Step 3: clear tempaddr list */
2805 	while (!list_empty(&idev->tempaddr_list)) {
2806 		ifa = list_first_entry(&idev->tempaddr_list,
2807 				       struct inet6_ifaddr, tmp_list);
2808 		list_del(&ifa->tmp_list);
2809 		write_unlock_bh(&idev->lock);
2810 		spin_lock_bh(&ifa->lock);
2811 
2812 		if (ifa->ifpub) {
2813 			in6_ifa_put(ifa->ifpub);
2814 			ifa->ifpub = NULL;
2815 		}
2816 		spin_unlock_bh(&ifa->lock);
2817 		in6_ifa_put(ifa);
2818 		write_lock_bh(&idev->lock);
2819 	}
2820 #endif
2821 
2822 	while (!list_empty(&idev->addr_list)) {
2823 		ifa = list_first_entry(&idev->addr_list,
2824 				       struct inet6_ifaddr, if_list);
2825 		addrconf_del_timer(ifa);
2826 
2827 		list_del(&ifa->if_list);
2828 
2829 		write_unlock_bh(&idev->lock);
2830 
2831 		spin_lock_bh(&ifa->state_lock);
2832 		state = ifa->state;
2833 		ifa->state = INET6_IFADDR_STATE_DEAD;
2834 		spin_unlock_bh(&ifa->state_lock);
2835 
2836 		if (state != INET6_IFADDR_STATE_DEAD) {
2837 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
2838 			atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2839 		}
2840 		in6_ifa_put(ifa);
2841 
2842 		write_lock_bh(&idev->lock);
2843 	}
2844 
2845 	write_unlock_bh(&idev->lock);
2846 
2847 	/* Step 5: Discard multicast list */
2848 	if (how)
2849 		ipv6_mc_destroy_dev(idev);
2850 	else
2851 		ipv6_mc_down(idev);
2852 
2853 	idev->tstamp = jiffies;
2854 
2855 	/* Last: Shot the device (if unregistered) */
2856 	if (how) {
2857 		addrconf_sysctl_unregister(idev);
2858 		neigh_parms_release(&nd_tbl, idev->nd_parms);
2859 		neigh_ifdown(&nd_tbl, dev);
2860 		in6_dev_put(idev);
2861 	}
2862 	return 0;
2863 }
2864 
2865 static void addrconf_rs_timer(unsigned long data)
2866 {
2867 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2868 	struct inet6_dev *idev = ifp->idev;
2869 
2870 	read_lock(&idev->lock);
2871 	if (idev->dead || !(idev->if_flags & IF_READY))
2872 		goto out;
2873 
2874 	if (idev->cnf.forwarding)
2875 		goto out;
2876 
2877 	/* Announcement received after solicitation was sent */
2878 	if (idev->if_flags & IF_RA_RCVD)
2879 		goto out;
2880 
2881 	spin_lock(&ifp->lock);
2882 	if (ifp->probes++ < idev->cnf.rtr_solicits) {
2883 		/* The wait after the last probe can be shorter */
2884 		addrconf_mod_timer(ifp, AC_RS,
2885 				   (ifp->probes == idev->cnf.rtr_solicits) ?
2886 				   idev->cnf.rtr_solicit_delay :
2887 				   idev->cnf.rtr_solicit_interval);
2888 		spin_unlock(&ifp->lock);
2889 
2890 		ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2891 	} else {
2892 		spin_unlock(&ifp->lock);
2893 		/*
2894 		 * Note: we do not support deprecated "all on-link"
2895 		 * assumption any longer.
2896 		 */
2897 		printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2898 		       idev->dev->name);
2899 	}
2900 
2901 out:
2902 	read_unlock(&idev->lock);
2903 	in6_ifa_put(ifp);
2904 }
2905 
2906 /*
2907  *	Duplicate Address Detection
2908  */
2909 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2910 {
2911 	unsigned long rand_num;
2912 	struct inet6_dev *idev = ifp->idev;
2913 
2914 	if (ifp->flags & IFA_F_OPTIMISTIC)
2915 		rand_num = 0;
2916 	else
2917 		rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2918 
2919 	ifp->probes = idev->cnf.dad_transmits;
2920 	addrconf_mod_timer(ifp, AC_DAD, rand_num);
2921 }
2922 
2923 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2924 {
2925 	struct inet6_dev *idev = ifp->idev;
2926 	struct net_device *dev = idev->dev;
2927 
2928 	addrconf_join_solict(dev, &ifp->addr);
2929 
2930 	net_srandom(ifp->addr.s6_addr32[3]);
2931 
2932 	read_lock_bh(&idev->lock);
2933 	spin_lock(&ifp->lock);
2934 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
2935 		goto out;
2936 
2937 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2938 	    idev->cnf.accept_dad < 1 ||
2939 	    !(ifp->flags&IFA_F_TENTATIVE) ||
2940 	    ifp->flags & IFA_F_NODAD) {
2941 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2942 		spin_unlock(&ifp->lock);
2943 		read_unlock_bh(&idev->lock);
2944 
2945 		addrconf_dad_completed(ifp);
2946 		return;
2947 	}
2948 
2949 	if (!(idev->if_flags & IF_READY)) {
2950 		spin_unlock(&ifp->lock);
2951 		read_unlock_bh(&idev->lock);
2952 		/*
2953 		 * If the device is not ready:
2954 		 * - keep it tentative if it is a permanent address.
2955 		 * - otherwise, kill it.
2956 		 */
2957 		in6_ifa_hold(ifp);
2958 		addrconf_dad_stop(ifp, 0);
2959 		return;
2960 	}
2961 
2962 	/*
2963 	 * Optimistic nodes can start receiving
2964 	 * Frames right away
2965 	 */
2966 	if (ifp->flags & IFA_F_OPTIMISTIC)
2967 		ip6_ins_rt(ifp->rt);
2968 
2969 	addrconf_dad_kick(ifp);
2970 out:
2971 	spin_unlock(&ifp->lock);
2972 	read_unlock_bh(&idev->lock);
2973 }
2974 
2975 static void addrconf_dad_timer(unsigned long data)
2976 {
2977 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2978 	struct inet6_dev *idev = ifp->idev;
2979 	struct in6_addr mcaddr;
2980 
2981 	if (!ifp->probes && addrconf_dad_end(ifp))
2982 		goto out;
2983 
2984 	read_lock(&idev->lock);
2985 	if (idev->dead || !(idev->if_flags & IF_READY)) {
2986 		read_unlock(&idev->lock);
2987 		goto out;
2988 	}
2989 
2990 	spin_lock(&ifp->lock);
2991 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2992 		spin_unlock(&ifp->lock);
2993 		read_unlock(&idev->lock);
2994 		goto out;
2995 	}
2996 
2997 	if (ifp->probes == 0) {
2998 		/*
2999 		 * DAD was successful
3000 		 */
3001 
3002 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3003 		spin_unlock(&ifp->lock);
3004 		read_unlock(&idev->lock);
3005 
3006 		addrconf_dad_completed(ifp);
3007 
3008 		goto out;
3009 	}
3010 
3011 	ifp->probes--;
3012 	addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3013 	spin_unlock(&ifp->lock);
3014 	read_unlock(&idev->lock);
3015 
3016 	/* send a neighbour solicitation for our addr */
3017 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3018 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3019 out:
3020 	in6_ifa_put(ifp);
3021 }
3022 
3023 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3024 {
3025 	struct net_device *dev = ifp->idev->dev;
3026 
3027 	/*
3028 	 *	Configure the address for reception. Now it is valid.
3029 	 */
3030 
3031 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
3032 
3033 	/* If added prefix is link local and we are prepared to process
3034 	   router advertisements, start sending router solicitations.
3035 	 */
3036 
3037 	if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3038 	     ifp->idev->cnf.accept_ra == 2) &&
3039 	    ifp->idev->cnf.rtr_solicits > 0 &&
3040 	    (dev->flags&IFF_LOOPBACK) == 0 &&
3041 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3042 		/*
3043 		 *	If a host as already performed a random delay
3044 		 *	[...] as part of DAD [...] there is no need
3045 		 *	to delay again before sending the first RS
3046 		 */
3047 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3048 
3049 		spin_lock_bh(&ifp->lock);
3050 		ifp->probes = 1;
3051 		ifp->idev->if_flags |= IF_RS_SENT;
3052 		addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3053 		spin_unlock_bh(&ifp->lock);
3054 	}
3055 }
3056 
3057 static void addrconf_dad_run(struct inet6_dev *idev)
3058 {
3059 	struct inet6_ifaddr *ifp;
3060 
3061 	read_lock_bh(&idev->lock);
3062 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3063 		spin_lock(&ifp->lock);
3064 		if (ifp->flags & IFA_F_TENTATIVE &&
3065 		    ifp->state == INET6_IFADDR_STATE_DAD)
3066 			addrconf_dad_kick(ifp);
3067 		spin_unlock(&ifp->lock);
3068 	}
3069 	read_unlock_bh(&idev->lock);
3070 }
3071 
3072 #ifdef CONFIG_PROC_FS
3073 struct if6_iter_state {
3074 	struct seq_net_private p;
3075 	int bucket;
3076 	int offset;
3077 };
3078 
3079 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3080 {
3081 	struct inet6_ifaddr *ifa = NULL;
3082 	struct if6_iter_state *state = seq->private;
3083 	struct net *net = seq_file_net(seq);
3084 	int p = 0;
3085 
3086 	/* initial bucket if pos is 0 */
3087 	if (pos == 0) {
3088 		state->bucket = 0;
3089 		state->offset = 0;
3090 	}
3091 
3092 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3093 		struct hlist_node *n;
3094 		hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3095 					 addr_lst) {
3096 			/* sync with offset */
3097 			if (p < state->offset) {
3098 				p++;
3099 				continue;
3100 			}
3101 			state->offset++;
3102 			if (net_eq(dev_net(ifa->idev->dev), net))
3103 				return ifa;
3104 		}
3105 
3106 		/* prepare for next bucket */
3107 		state->offset = 0;
3108 		p = 0;
3109 	}
3110 	return NULL;
3111 }
3112 
3113 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3114 					 struct inet6_ifaddr *ifa)
3115 {
3116 	struct if6_iter_state *state = seq->private;
3117 	struct net *net = seq_file_net(seq);
3118 	struct hlist_node *n = &ifa->addr_lst;
3119 
3120 	hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3121 		state->offset++;
3122 		if (net_eq(dev_net(ifa->idev->dev), net))
3123 			return ifa;
3124 	}
3125 
3126 	while (++state->bucket < IN6_ADDR_HSIZE) {
3127 		state->offset = 0;
3128 		hlist_for_each_entry_rcu_bh(ifa, n,
3129 				     &inet6_addr_lst[state->bucket], addr_lst) {
3130 			state->offset++;
3131 			if (net_eq(dev_net(ifa->idev->dev), net))
3132 				return ifa;
3133 		}
3134 	}
3135 
3136 	return NULL;
3137 }
3138 
3139 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3140 	__acquires(rcu_bh)
3141 {
3142 	rcu_read_lock_bh();
3143 	return if6_get_first(seq, *pos);
3144 }
3145 
3146 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3147 {
3148 	struct inet6_ifaddr *ifa;
3149 
3150 	ifa = if6_get_next(seq, v);
3151 	++*pos;
3152 	return ifa;
3153 }
3154 
3155 static void if6_seq_stop(struct seq_file *seq, void *v)
3156 	__releases(rcu_bh)
3157 {
3158 	rcu_read_unlock_bh();
3159 }
3160 
3161 static int if6_seq_show(struct seq_file *seq, void *v)
3162 {
3163 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3164 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3165 		   &ifp->addr,
3166 		   ifp->idev->dev->ifindex,
3167 		   ifp->prefix_len,
3168 		   ifp->scope,
3169 		   ifp->flags,
3170 		   ifp->idev->dev->name);
3171 	return 0;
3172 }
3173 
3174 static const struct seq_operations if6_seq_ops = {
3175 	.start	= if6_seq_start,
3176 	.next	= if6_seq_next,
3177 	.show	= if6_seq_show,
3178 	.stop	= if6_seq_stop,
3179 };
3180 
3181 static int if6_seq_open(struct inode *inode, struct file *file)
3182 {
3183 	return seq_open_net(inode, file, &if6_seq_ops,
3184 			    sizeof(struct if6_iter_state));
3185 }
3186 
3187 static const struct file_operations if6_fops = {
3188 	.owner		= THIS_MODULE,
3189 	.open		= if6_seq_open,
3190 	.read		= seq_read,
3191 	.llseek		= seq_lseek,
3192 	.release	= seq_release_net,
3193 };
3194 
3195 static int __net_init if6_proc_net_init(struct net *net)
3196 {
3197 	if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3198 		return -ENOMEM;
3199 	return 0;
3200 }
3201 
3202 static void __net_exit if6_proc_net_exit(struct net *net)
3203 {
3204        proc_net_remove(net, "if_inet6");
3205 }
3206 
3207 static struct pernet_operations if6_proc_net_ops = {
3208        .init = if6_proc_net_init,
3209        .exit = if6_proc_net_exit,
3210 };
3211 
3212 int __init if6_proc_init(void)
3213 {
3214 	return register_pernet_subsys(&if6_proc_net_ops);
3215 }
3216 
3217 void if6_proc_exit(void)
3218 {
3219 	unregister_pernet_subsys(&if6_proc_net_ops);
3220 }
3221 #endif	/* CONFIG_PROC_FS */
3222 
3223 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3224 /* Check if address is a home address configured on any interface. */
3225 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3226 {
3227 	int ret = 0;
3228 	struct inet6_ifaddr *ifp = NULL;
3229 	struct hlist_node *n;
3230 	unsigned int hash = ipv6_addr_hash(addr);
3231 
3232 	rcu_read_lock_bh();
3233 	hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3234 		if (!net_eq(dev_net(ifp->idev->dev), net))
3235 			continue;
3236 		if (ipv6_addr_equal(&ifp->addr, addr) &&
3237 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3238 			ret = 1;
3239 			break;
3240 		}
3241 	}
3242 	rcu_read_unlock_bh();
3243 	return ret;
3244 }
3245 #endif
3246 
3247 /*
3248  *	Periodic address status verification
3249  */
3250 
3251 static void addrconf_verify(unsigned long foo)
3252 {
3253 	unsigned long now, next, next_sec, next_sched;
3254 	struct inet6_ifaddr *ifp;
3255 	struct hlist_node *node;
3256 	int i;
3257 
3258 	rcu_read_lock_bh();
3259 	spin_lock(&addrconf_verify_lock);
3260 	now = jiffies;
3261 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3262 
3263 	del_timer(&addr_chk_timer);
3264 
3265 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3266 restart:
3267 		hlist_for_each_entry_rcu_bh(ifp, node,
3268 					 &inet6_addr_lst[i], addr_lst) {
3269 			unsigned long age;
3270 
3271 			if (ifp->flags & IFA_F_PERMANENT)
3272 				continue;
3273 
3274 			spin_lock(&ifp->lock);
3275 			/* We try to batch several events at once. */
3276 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3277 
3278 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3279 			    age >= ifp->valid_lft) {
3280 				spin_unlock(&ifp->lock);
3281 				in6_ifa_hold(ifp);
3282 				ipv6_del_addr(ifp);
3283 				goto restart;
3284 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3285 				spin_unlock(&ifp->lock);
3286 				continue;
3287 			} else if (age >= ifp->prefered_lft) {
3288 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3289 				int deprecate = 0;
3290 
3291 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3292 					deprecate = 1;
3293 					ifp->flags |= IFA_F_DEPRECATED;
3294 				}
3295 
3296 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3297 					next = ifp->tstamp + ifp->valid_lft * HZ;
3298 
3299 				spin_unlock(&ifp->lock);
3300 
3301 				if (deprecate) {
3302 					in6_ifa_hold(ifp);
3303 
3304 					ipv6_ifa_notify(0, ifp);
3305 					in6_ifa_put(ifp);
3306 					goto restart;
3307 				}
3308 #ifdef CONFIG_IPV6_PRIVACY
3309 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3310 				   !(ifp->flags&IFA_F_TENTATIVE)) {
3311 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3312 					ifp->idev->cnf.dad_transmits *
3313 					ifp->idev->nd_parms->retrans_time / HZ;
3314 
3315 				if (age >= ifp->prefered_lft - regen_advance) {
3316 					struct inet6_ifaddr *ifpub = ifp->ifpub;
3317 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3318 						next = ifp->tstamp + ifp->prefered_lft * HZ;
3319 					if (!ifp->regen_count && ifpub) {
3320 						ifp->regen_count++;
3321 						in6_ifa_hold(ifp);
3322 						in6_ifa_hold(ifpub);
3323 						spin_unlock(&ifp->lock);
3324 
3325 						spin_lock(&ifpub->lock);
3326 						ifpub->regen_count = 0;
3327 						spin_unlock(&ifpub->lock);
3328 						ipv6_create_tempaddr(ifpub, ifp);
3329 						in6_ifa_put(ifpub);
3330 						in6_ifa_put(ifp);
3331 						goto restart;
3332 					}
3333 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3334 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3335 				spin_unlock(&ifp->lock);
3336 #endif
3337 			} else {
3338 				/* ifp->prefered_lft <= ifp->valid_lft */
3339 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3340 					next = ifp->tstamp + ifp->prefered_lft * HZ;
3341 				spin_unlock(&ifp->lock);
3342 			}
3343 		}
3344 	}
3345 
3346 	next_sec = round_jiffies_up(next);
3347 	next_sched = next;
3348 
3349 	/* If rounded timeout is accurate enough, accept it. */
3350 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3351 		next_sched = next_sec;
3352 
3353 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3354 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3355 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3356 
3357 	ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3358 	      now, next, next_sec, next_sched));
3359 
3360 	addr_chk_timer.expires = next_sched;
3361 	add_timer(&addr_chk_timer);
3362 	spin_unlock(&addrconf_verify_lock);
3363 	rcu_read_unlock_bh();
3364 }
3365 
3366 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3367 {
3368 	struct in6_addr *pfx = NULL;
3369 
3370 	if (addr)
3371 		pfx = nla_data(addr);
3372 
3373 	if (local) {
3374 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3375 			pfx = NULL;
3376 		else
3377 			pfx = nla_data(local);
3378 	}
3379 
3380 	return pfx;
3381 }
3382 
3383 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3384 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
3385 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
3386 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
3387 };
3388 
3389 static int
3390 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3391 {
3392 	struct net *net = sock_net(skb->sk);
3393 	struct ifaddrmsg *ifm;
3394 	struct nlattr *tb[IFA_MAX+1];
3395 	struct in6_addr *pfx;
3396 	int err;
3397 
3398 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3399 	if (err < 0)
3400 		return err;
3401 
3402 	ifm = nlmsg_data(nlh);
3403 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3404 	if (pfx == NULL)
3405 		return -EINVAL;
3406 
3407 	return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3408 }
3409 
3410 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3411 			     u32 prefered_lft, u32 valid_lft)
3412 {
3413 	u32 flags;
3414 	clock_t expires;
3415 	unsigned long timeout;
3416 
3417 	if (!valid_lft || (prefered_lft > valid_lft))
3418 		return -EINVAL;
3419 
3420 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
3421 	if (addrconf_finite_timeout(timeout)) {
3422 		expires = jiffies_to_clock_t(timeout * HZ);
3423 		valid_lft = timeout;
3424 		flags = RTF_EXPIRES;
3425 	} else {
3426 		expires = 0;
3427 		flags = 0;
3428 		ifa_flags |= IFA_F_PERMANENT;
3429 	}
3430 
3431 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3432 	if (addrconf_finite_timeout(timeout)) {
3433 		if (timeout == 0)
3434 			ifa_flags |= IFA_F_DEPRECATED;
3435 		prefered_lft = timeout;
3436 	}
3437 
3438 	spin_lock_bh(&ifp->lock);
3439 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3440 	ifp->tstamp = jiffies;
3441 	ifp->valid_lft = valid_lft;
3442 	ifp->prefered_lft = prefered_lft;
3443 
3444 	spin_unlock_bh(&ifp->lock);
3445 	if (!(ifp->flags&IFA_F_TENTATIVE))
3446 		ipv6_ifa_notify(0, ifp);
3447 
3448 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3449 			      expires, flags);
3450 	addrconf_verify(0);
3451 
3452 	return 0;
3453 }
3454 
3455 static int
3456 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3457 {
3458 	struct net *net = sock_net(skb->sk);
3459 	struct ifaddrmsg *ifm;
3460 	struct nlattr *tb[IFA_MAX+1];
3461 	struct in6_addr *pfx;
3462 	struct inet6_ifaddr *ifa;
3463 	struct net_device *dev;
3464 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3465 	u8 ifa_flags;
3466 	int err;
3467 
3468 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3469 	if (err < 0)
3470 		return err;
3471 
3472 	ifm = nlmsg_data(nlh);
3473 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3474 	if (pfx == NULL)
3475 		return -EINVAL;
3476 
3477 	if (tb[IFA_CACHEINFO]) {
3478 		struct ifa_cacheinfo *ci;
3479 
3480 		ci = nla_data(tb[IFA_CACHEINFO]);
3481 		valid_lft = ci->ifa_valid;
3482 		preferred_lft = ci->ifa_prefered;
3483 	} else {
3484 		preferred_lft = INFINITY_LIFE_TIME;
3485 		valid_lft = INFINITY_LIFE_TIME;
3486 	}
3487 
3488 	dev =  __dev_get_by_index(net, ifm->ifa_index);
3489 	if (dev == NULL)
3490 		return -ENODEV;
3491 
3492 	/* We ignore other flags so far. */
3493 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3494 
3495 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3496 	if (ifa == NULL) {
3497 		/*
3498 		 * It would be best to check for !NLM_F_CREATE here but
3499 		 * userspace alreay relies on not having to provide this.
3500 		 */
3501 		return inet6_addr_add(net, ifm->ifa_index, pfx,
3502 				      ifm->ifa_prefixlen, ifa_flags,
3503 				      preferred_lft, valid_lft);
3504 	}
3505 
3506 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3507 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3508 		err = -EEXIST;
3509 	else
3510 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3511 
3512 	in6_ifa_put(ifa);
3513 
3514 	return err;
3515 }
3516 
3517 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3518 			  u8 scope, int ifindex)
3519 {
3520 	struct ifaddrmsg *ifm;
3521 
3522 	ifm = nlmsg_data(nlh);
3523 	ifm->ifa_family = AF_INET6;
3524 	ifm->ifa_prefixlen = prefixlen;
3525 	ifm->ifa_flags = flags;
3526 	ifm->ifa_scope = scope;
3527 	ifm->ifa_index = ifindex;
3528 }
3529 
3530 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3531 			 unsigned long tstamp, u32 preferred, u32 valid)
3532 {
3533 	struct ifa_cacheinfo ci;
3534 
3535 	ci.cstamp = cstamp_delta(cstamp);
3536 	ci.tstamp = cstamp_delta(tstamp);
3537 	ci.ifa_prefered = preferred;
3538 	ci.ifa_valid = valid;
3539 
3540 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3541 }
3542 
3543 static inline int rt_scope(int ifa_scope)
3544 {
3545 	if (ifa_scope & IFA_HOST)
3546 		return RT_SCOPE_HOST;
3547 	else if (ifa_scope & IFA_LINK)
3548 		return RT_SCOPE_LINK;
3549 	else if (ifa_scope & IFA_SITE)
3550 		return RT_SCOPE_SITE;
3551 	else
3552 		return RT_SCOPE_UNIVERSE;
3553 }
3554 
3555 static inline int inet6_ifaddr_msgsize(void)
3556 {
3557 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3558 	       + nla_total_size(16) /* IFA_ADDRESS */
3559 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3560 }
3561 
3562 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3563 			     u32 pid, u32 seq, int event, unsigned int flags)
3564 {
3565 	struct nlmsghdr  *nlh;
3566 	u32 preferred, valid;
3567 
3568 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3569 	if (nlh == NULL)
3570 		return -EMSGSIZE;
3571 
3572 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3573 		      ifa->idev->dev->ifindex);
3574 
3575 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3576 		preferred = ifa->prefered_lft;
3577 		valid = ifa->valid_lft;
3578 		if (preferred != INFINITY_LIFE_TIME) {
3579 			long tval = (jiffies - ifa->tstamp)/HZ;
3580 			if (preferred > tval)
3581 				preferred -= tval;
3582 			else
3583 				preferred = 0;
3584 			if (valid != INFINITY_LIFE_TIME) {
3585 				if (valid > tval)
3586 					valid -= tval;
3587 				else
3588 					valid = 0;
3589 			}
3590 		}
3591 	} else {
3592 		preferred = INFINITY_LIFE_TIME;
3593 		valid = INFINITY_LIFE_TIME;
3594 	}
3595 
3596 	if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3597 	    put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3598 		nlmsg_cancel(skb, nlh);
3599 		return -EMSGSIZE;
3600 	}
3601 
3602 	return nlmsg_end(skb, nlh);
3603 }
3604 
3605 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3606 				u32 pid, u32 seq, int event, u16 flags)
3607 {
3608 	struct nlmsghdr  *nlh;
3609 	u8 scope = RT_SCOPE_UNIVERSE;
3610 	int ifindex = ifmca->idev->dev->ifindex;
3611 
3612 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3613 		scope = RT_SCOPE_SITE;
3614 
3615 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3616 	if (nlh == NULL)
3617 		return -EMSGSIZE;
3618 
3619 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3620 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3621 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3622 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3623 		nlmsg_cancel(skb, nlh);
3624 		return -EMSGSIZE;
3625 	}
3626 
3627 	return nlmsg_end(skb, nlh);
3628 }
3629 
3630 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3631 				u32 pid, u32 seq, int event, unsigned int flags)
3632 {
3633 	struct nlmsghdr  *nlh;
3634 	u8 scope = RT_SCOPE_UNIVERSE;
3635 	int ifindex = ifaca->aca_idev->dev->ifindex;
3636 
3637 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3638 		scope = RT_SCOPE_SITE;
3639 
3640 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3641 	if (nlh == NULL)
3642 		return -EMSGSIZE;
3643 
3644 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3645 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3646 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3647 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3648 		nlmsg_cancel(skb, nlh);
3649 		return -EMSGSIZE;
3650 	}
3651 
3652 	return nlmsg_end(skb, nlh);
3653 }
3654 
3655 enum addr_type_t {
3656 	UNICAST_ADDR,
3657 	MULTICAST_ADDR,
3658 	ANYCAST_ADDR,
3659 };
3660 
3661 /* called with rcu_read_lock() */
3662 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3663 			  struct netlink_callback *cb, enum addr_type_t type,
3664 			  int s_ip_idx, int *p_ip_idx)
3665 {
3666 	struct ifmcaddr6 *ifmca;
3667 	struct ifacaddr6 *ifaca;
3668 	int err = 1;
3669 	int ip_idx = *p_ip_idx;
3670 
3671 	read_lock_bh(&idev->lock);
3672 	switch (type) {
3673 	case UNICAST_ADDR: {
3674 		struct inet6_ifaddr *ifa;
3675 
3676 		/* unicast address incl. temp addr */
3677 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
3678 			if (++ip_idx < s_ip_idx)
3679 				continue;
3680 			err = inet6_fill_ifaddr(skb, ifa,
3681 						NETLINK_CB(cb->skb).pid,
3682 						cb->nlh->nlmsg_seq,
3683 						RTM_NEWADDR,
3684 						NLM_F_MULTI);
3685 			if (err <= 0)
3686 				break;
3687 		}
3688 		break;
3689 	}
3690 	case MULTICAST_ADDR:
3691 		/* multicast address */
3692 		for (ifmca = idev->mc_list; ifmca;
3693 		     ifmca = ifmca->next, ip_idx++) {
3694 			if (ip_idx < s_ip_idx)
3695 				continue;
3696 			err = inet6_fill_ifmcaddr(skb, ifmca,
3697 						  NETLINK_CB(cb->skb).pid,
3698 						  cb->nlh->nlmsg_seq,
3699 						  RTM_GETMULTICAST,
3700 						  NLM_F_MULTI);
3701 			if (err <= 0)
3702 				break;
3703 		}
3704 		break;
3705 	case ANYCAST_ADDR:
3706 		/* anycast address */
3707 		for (ifaca = idev->ac_list; ifaca;
3708 		     ifaca = ifaca->aca_next, ip_idx++) {
3709 			if (ip_idx < s_ip_idx)
3710 				continue;
3711 			err = inet6_fill_ifacaddr(skb, ifaca,
3712 						  NETLINK_CB(cb->skb).pid,
3713 						  cb->nlh->nlmsg_seq,
3714 						  RTM_GETANYCAST,
3715 						  NLM_F_MULTI);
3716 			if (err <= 0)
3717 				break;
3718 		}
3719 		break;
3720 	default:
3721 		break;
3722 	}
3723 	read_unlock_bh(&idev->lock);
3724 	*p_ip_idx = ip_idx;
3725 	return err;
3726 }
3727 
3728 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3729 			   enum addr_type_t type)
3730 {
3731 	struct net *net = sock_net(skb->sk);
3732 	int h, s_h;
3733 	int idx, ip_idx;
3734 	int s_idx, s_ip_idx;
3735 	struct net_device *dev;
3736 	struct inet6_dev *idev;
3737 	struct hlist_head *head;
3738 	struct hlist_node *node;
3739 
3740 	s_h = cb->args[0];
3741 	s_idx = idx = cb->args[1];
3742 	s_ip_idx = ip_idx = cb->args[2];
3743 
3744 	rcu_read_lock();
3745 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3746 		idx = 0;
3747 		head = &net->dev_index_head[h];
3748 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3749 			if (idx < s_idx)
3750 				goto cont;
3751 			if (h > s_h || idx > s_idx)
3752 				s_ip_idx = 0;
3753 			ip_idx = 0;
3754 			idev = __in6_dev_get(dev);
3755 			if (!idev)
3756 				goto cont;
3757 
3758 			if (in6_dump_addrs(idev, skb, cb, type,
3759 					   s_ip_idx, &ip_idx) <= 0)
3760 				goto done;
3761 cont:
3762 			idx++;
3763 		}
3764 	}
3765 done:
3766 	rcu_read_unlock();
3767 	cb->args[0] = h;
3768 	cb->args[1] = idx;
3769 	cb->args[2] = ip_idx;
3770 
3771 	return skb->len;
3772 }
3773 
3774 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3775 {
3776 	enum addr_type_t type = UNICAST_ADDR;
3777 
3778 	return inet6_dump_addr(skb, cb, type);
3779 }
3780 
3781 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3782 {
3783 	enum addr_type_t type = MULTICAST_ADDR;
3784 
3785 	return inet6_dump_addr(skb, cb, type);
3786 }
3787 
3788 
3789 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3790 {
3791 	enum addr_type_t type = ANYCAST_ADDR;
3792 
3793 	return inet6_dump_addr(skb, cb, type);
3794 }
3795 
3796 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3797 			     void *arg)
3798 {
3799 	struct net *net = sock_net(in_skb->sk);
3800 	struct ifaddrmsg *ifm;
3801 	struct nlattr *tb[IFA_MAX+1];
3802 	struct in6_addr *addr = NULL;
3803 	struct net_device *dev = NULL;
3804 	struct inet6_ifaddr *ifa;
3805 	struct sk_buff *skb;
3806 	int err;
3807 
3808 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3809 	if (err < 0)
3810 		goto errout;
3811 
3812 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3813 	if (addr == NULL) {
3814 		err = -EINVAL;
3815 		goto errout;
3816 	}
3817 
3818 	ifm = nlmsg_data(nlh);
3819 	if (ifm->ifa_index)
3820 		dev = __dev_get_by_index(net, ifm->ifa_index);
3821 
3822 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3823 	if (!ifa) {
3824 		err = -EADDRNOTAVAIL;
3825 		goto errout;
3826 	}
3827 
3828 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3829 	if (!skb) {
3830 		err = -ENOBUFS;
3831 		goto errout_ifa;
3832 	}
3833 
3834 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3835 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
3836 	if (err < 0) {
3837 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3838 		WARN_ON(err == -EMSGSIZE);
3839 		kfree_skb(skb);
3840 		goto errout_ifa;
3841 	}
3842 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3843 errout_ifa:
3844 	in6_ifa_put(ifa);
3845 errout:
3846 	return err;
3847 }
3848 
3849 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3850 {
3851 	struct sk_buff *skb;
3852 	struct net *net = dev_net(ifa->idev->dev);
3853 	int err = -ENOBUFS;
3854 
3855 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3856 	if (skb == NULL)
3857 		goto errout;
3858 
3859 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3860 	if (err < 0) {
3861 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3862 		WARN_ON(err == -EMSGSIZE);
3863 		kfree_skb(skb);
3864 		goto errout;
3865 	}
3866 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3867 	return;
3868 errout:
3869 	if (err < 0)
3870 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3871 }
3872 
3873 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3874 				__s32 *array, int bytes)
3875 {
3876 	BUG_ON(bytes < (DEVCONF_MAX * 4));
3877 
3878 	memset(array, 0, bytes);
3879 	array[DEVCONF_FORWARDING] = cnf->forwarding;
3880 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3881 	array[DEVCONF_MTU6] = cnf->mtu6;
3882 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3883 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3884 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
3885 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3886 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3887 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3888 		jiffies_to_msecs(cnf->rtr_solicit_interval);
3889 	array[DEVCONF_RTR_SOLICIT_DELAY] =
3890 		jiffies_to_msecs(cnf->rtr_solicit_delay);
3891 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3892 #ifdef CONFIG_IPV6_PRIVACY
3893 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3894 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3895 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3896 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3897 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3898 #endif
3899 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3900 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3901 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3902 #ifdef CONFIG_IPV6_ROUTER_PREF
3903 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3904 	array[DEVCONF_RTR_PROBE_INTERVAL] =
3905 		jiffies_to_msecs(cnf->rtr_probe_interval);
3906 #ifdef CONFIG_IPV6_ROUTE_INFO
3907 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3908 #endif
3909 #endif
3910 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3911 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3912 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3913 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3914 #endif
3915 #ifdef CONFIG_IPV6_MROUTE
3916 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3917 #endif
3918 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3919 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3920 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3921 }
3922 
3923 static inline size_t inet6_ifla6_size(void)
3924 {
3925 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
3926 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
3927 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3928 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3929 	     + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3930 }
3931 
3932 static inline size_t inet6_if_nlmsg_size(void)
3933 {
3934 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3935 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3936 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3937 	       + nla_total_size(4) /* IFLA_MTU */
3938 	       + nla_total_size(4) /* IFLA_LINK */
3939 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3940 }
3941 
3942 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3943 				      int items, int bytes)
3944 {
3945 	int i;
3946 	int pad = bytes - sizeof(u64) * items;
3947 	BUG_ON(pad < 0);
3948 
3949 	/* Use put_unaligned() because stats may not be aligned for u64. */
3950 	put_unaligned(items, &stats[0]);
3951 	for (i = 1; i < items; i++)
3952 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3953 
3954 	memset(&stats[items], 0, pad);
3955 }
3956 
3957 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3958 				      int items, int bytes, size_t syncpoff)
3959 {
3960 	int i;
3961 	int pad = bytes - sizeof(u64) * items;
3962 	BUG_ON(pad < 0);
3963 
3964 	/* Use put_unaligned() because stats may not be aligned for u64. */
3965 	put_unaligned(items, &stats[0]);
3966 	for (i = 1; i < items; i++)
3967 		put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3968 
3969 	memset(&stats[items], 0, pad);
3970 }
3971 
3972 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3973 			     int bytes)
3974 {
3975 	switch (attrtype) {
3976 	case IFLA_INET6_STATS:
3977 		__snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3978 				     IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3979 		break;
3980 	case IFLA_INET6_ICMP6STATS:
3981 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3982 		break;
3983 	}
3984 }
3985 
3986 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3987 {
3988 	struct nlattr *nla;
3989 	struct ifla_cacheinfo ci;
3990 
3991 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3992 		goto nla_put_failure;
3993 	ci.max_reasm_len = IPV6_MAXPLEN;
3994 	ci.tstamp = cstamp_delta(idev->tstamp);
3995 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3996 	ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3997 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3998 		goto nla_put_failure;
3999 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4000 	if (nla == NULL)
4001 		goto nla_put_failure;
4002 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4003 
4004 	/* XXX - MC not implemented */
4005 
4006 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4007 	if (nla == NULL)
4008 		goto nla_put_failure;
4009 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4010 
4011 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4012 	if (nla == NULL)
4013 		goto nla_put_failure;
4014 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4015 
4016 	return 0;
4017 
4018 nla_put_failure:
4019 	return -EMSGSIZE;
4020 }
4021 
4022 static size_t inet6_get_link_af_size(const struct net_device *dev)
4023 {
4024 	if (!__in6_dev_get(dev))
4025 		return 0;
4026 
4027 	return inet6_ifla6_size();
4028 }
4029 
4030 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4031 {
4032 	struct inet6_dev *idev = __in6_dev_get(dev);
4033 
4034 	if (!idev)
4035 		return -ENODATA;
4036 
4037 	if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4038 		return -EMSGSIZE;
4039 
4040 	return 0;
4041 }
4042 
4043 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4044 			     u32 pid, u32 seq, int event, unsigned int flags)
4045 {
4046 	struct net_device *dev = idev->dev;
4047 	struct ifinfomsg *hdr;
4048 	struct nlmsghdr *nlh;
4049 	void *protoinfo;
4050 
4051 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4052 	if (nlh == NULL)
4053 		return -EMSGSIZE;
4054 
4055 	hdr = nlmsg_data(nlh);
4056 	hdr->ifi_family = AF_INET6;
4057 	hdr->__ifi_pad = 0;
4058 	hdr->ifi_type = dev->type;
4059 	hdr->ifi_index = dev->ifindex;
4060 	hdr->ifi_flags = dev_get_flags(dev);
4061 	hdr->ifi_change = 0;
4062 
4063 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4064 	    (dev->addr_len &&
4065 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4066 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4067 	    (dev->ifindex != dev->iflink &&
4068 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4069 		goto nla_put_failure;
4070 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4071 	if (protoinfo == NULL)
4072 		goto nla_put_failure;
4073 
4074 	if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4075 		goto nla_put_failure;
4076 
4077 	nla_nest_end(skb, protoinfo);
4078 	return nlmsg_end(skb, nlh);
4079 
4080 nla_put_failure:
4081 	nlmsg_cancel(skb, nlh);
4082 	return -EMSGSIZE;
4083 }
4084 
4085 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4086 {
4087 	struct net *net = sock_net(skb->sk);
4088 	int h, s_h;
4089 	int idx = 0, s_idx;
4090 	struct net_device *dev;
4091 	struct inet6_dev *idev;
4092 	struct hlist_head *head;
4093 	struct hlist_node *node;
4094 
4095 	s_h = cb->args[0];
4096 	s_idx = cb->args[1];
4097 
4098 	rcu_read_lock();
4099 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4100 		idx = 0;
4101 		head = &net->dev_index_head[h];
4102 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4103 			if (idx < s_idx)
4104 				goto cont;
4105 			idev = __in6_dev_get(dev);
4106 			if (!idev)
4107 				goto cont;
4108 			if (inet6_fill_ifinfo(skb, idev,
4109 					      NETLINK_CB(cb->skb).pid,
4110 					      cb->nlh->nlmsg_seq,
4111 					      RTM_NEWLINK, NLM_F_MULTI) <= 0)
4112 				goto out;
4113 cont:
4114 			idx++;
4115 		}
4116 	}
4117 out:
4118 	rcu_read_unlock();
4119 	cb->args[1] = idx;
4120 	cb->args[0] = h;
4121 
4122 	return skb->len;
4123 }
4124 
4125 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4126 {
4127 	struct sk_buff *skb;
4128 	struct net *net = dev_net(idev->dev);
4129 	int err = -ENOBUFS;
4130 
4131 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4132 	if (skb == NULL)
4133 		goto errout;
4134 
4135 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4136 	if (err < 0) {
4137 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4138 		WARN_ON(err == -EMSGSIZE);
4139 		kfree_skb(skb);
4140 		goto errout;
4141 	}
4142 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4143 	return;
4144 errout:
4145 	if (err < 0)
4146 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4147 }
4148 
4149 static inline size_t inet6_prefix_nlmsg_size(void)
4150 {
4151 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
4152 	       + nla_total_size(sizeof(struct in6_addr))
4153 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
4154 }
4155 
4156 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4157 			     struct prefix_info *pinfo, u32 pid, u32 seq,
4158 			     int event, unsigned int flags)
4159 {
4160 	struct prefixmsg *pmsg;
4161 	struct nlmsghdr *nlh;
4162 	struct prefix_cacheinfo	ci;
4163 
4164 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4165 	if (nlh == NULL)
4166 		return -EMSGSIZE;
4167 
4168 	pmsg = nlmsg_data(nlh);
4169 	pmsg->prefix_family = AF_INET6;
4170 	pmsg->prefix_pad1 = 0;
4171 	pmsg->prefix_pad2 = 0;
4172 	pmsg->prefix_ifindex = idev->dev->ifindex;
4173 	pmsg->prefix_len = pinfo->prefix_len;
4174 	pmsg->prefix_type = pinfo->type;
4175 	pmsg->prefix_pad3 = 0;
4176 	pmsg->prefix_flags = 0;
4177 	if (pinfo->onlink)
4178 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4179 	if (pinfo->autoconf)
4180 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4181 
4182 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4183 		goto nla_put_failure;
4184 	ci.preferred_time = ntohl(pinfo->prefered);
4185 	ci.valid_time = ntohl(pinfo->valid);
4186 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4187 		goto nla_put_failure;
4188 	return nlmsg_end(skb, nlh);
4189 
4190 nla_put_failure:
4191 	nlmsg_cancel(skb, nlh);
4192 	return -EMSGSIZE;
4193 }
4194 
4195 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4196 			 struct prefix_info *pinfo)
4197 {
4198 	struct sk_buff *skb;
4199 	struct net *net = dev_net(idev->dev);
4200 	int err = -ENOBUFS;
4201 
4202 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4203 	if (skb == NULL)
4204 		goto errout;
4205 
4206 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4207 	if (err < 0) {
4208 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4209 		WARN_ON(err == -EMSGSIZE);
4210 		kfree_skb(skb);
4211 		goto errout;
4212 	}
4213 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4214 	return;
4215 errout:
4216 	if (err < 0)
4217 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4218 }
4219 
4220 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4221 {
4222 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4223 
4224 	switch (event) {
4225 	case RTM_NEWADDR:
4226 		/*
4227 		 * If the address was optimistic
4228 		 * we inserted the route at the start of
4229 		 * our DAD process, so we don't need
4230 		 * to do it again
4231 		 */
4232 		if (!(ifp->rt->rt6i_node))
4233 			ip6_ins_rt(ifp->rt);
4234 		if (ifp->idev->cnf.forwarding)
4235 			addrconf_join_anycast(ifp);
4236 		break;
4237 	case RTM_DELADDR:
4238 		if (ifp->idev->cnf.forwarding)
4239 			addrconf_leave_anycast(ifp);
4240 		addrconf_leave_solict(ifp->idev, &ifp->addr);
4241 		dst_hold(&ifp->rt->dst);
4242 
4243 		if (ip6_del_rt(ifp->rt))
4244 			dst_free(&ifp->rt->dst);
4245 		break;
4246 	}
4247 }
4248 
4249 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4250 {
4251 	rcu_read_lock_bh();
4252 	if (likely(ifp->idev->dead == 0))
4253 		__ipv6_ifa_notify(event, ifp);
4254 	rcu_read_unlock_bh();
4255 }
4256 
4257 #ifdef CONFIG_SYSCTL
4258 
4259 static
4260 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4261 			   void __user *buffer, size_t *lenp, loff_t *ppos)
4262 {
4263 	int *valp = ctl->data;
4264 	int val = *valp;
4265 	loff_t pos = *ppos;
4266 	ctl_table lctl;
4267 	int ret;
4268 
4269 	/*
4270 	 * ctl->data points to idev->cnf.forwarding, we should
4271 	 * not modify it until we get the rtnl lock.
4272 	 */
4273 	lctl = *ctl;
4274 	lctl.data = &val;
4275 
4276 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4277 
4278 	if (write)
4279 		ret = addrconf_fixup_forwarding(ctl, valp, val);
4280 	if (ret)
4281 		*ppos = pos;
4282 	return ret;
4283 }
4284 
4285 static void dev_disable_change(struct inet6_dev *idev)
4286 {
4287 	if (!idev || !idev->dev)
4288 		return;
4289 
4290 	if (idev->cnf.disable_ipv6)
4291 		addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4292 	else
4293 		addrconf_notify(NULL, NETDEV_UP, idev->dev);
4294 }
4295 
4296 static void addrconf_disable_change(struct net *net, __s32 newf)
4297 {
4298 	struct net_device *dev;
4299 	struct inet6_dev *idev;
4300 
4301 	rcu_read_lock();
4302 	for_each_netdev_rcu(net, dev) {
4303 		idev = __in6_dev_get(dev);
4304 		if (idev) {
4305 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4306 			idev->cnf.disable_ipv6 = newf;
4307 			if (changed)
4308 				dev_disable_change(idev);
4309 		}
4310 	}
4311 	rcu_read_unlock();
4312 }
4313 
4314 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4315 {
4316 	struct net *net;
4317 	int old;
4318 
4319 	if (!rtnl_trylock())
4320 		return restart_syscall();
4321 
4322 	net = (struct net *)table->extra2;
4323 	old = *p;
4324 	*p = newf;
4325 
4326 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4327 		rtnl_unlock();
4328 		return 0;
4329 	}
4330 
4331 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
4332 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
4333 		addrconf_disable_change(net, newf);
4334 	} else if ((!newf) ^ (!old))
4335 		dev_disable_change((struct inet6_dev *)table->extra1);
4336 
4337 	rtnl_unlock();
4338 	return 0;
4339 }
4340 
4341 static
4342 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4343 			    void __user *buffer, size_t *lenp, loff_t *ppos)
4344 {
4345 	int *valp = ctl->data;
4346 	int val = *valp;
4347 	loff_t pos = *ppos;
4348 	ctl_table lctl;
4349 	int ret;
4350 
4351 	/*
4352 	 * ctl->data points to idev->cnf.disable_ipv6, we should
4353 	 * not modify it until we get the rtnl lock.
4354 	 */
4355 	lctl = *ctl;
4356 	lctl.data = &val;
4357 
4358 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4359 
4360 	if (write)
4361 		ret = addrconf_disable_ipv6(ctl, valp, val);
4362 	if (ret)
4363 		*ppos = pos;
4364 	return ret;
4365 }
4366 
4367 static struct addrconf_sysctl_table
4368 {
4369 	struct ctl_table_header *sysctl_header;
4370 	ctl_table addrconf_vars[DEVCONF_MAX+1];
4371 	char *dev_name;
4372 } addrconf_sysctl __read_mostly = {
4373 	.sysctl_header = NULL,
4374 	.addrconf_vars = {
4375 		{
4376 			.procname	= "forwarding",
4377 			.data		= &ipv6_devconf.forwarding,
4378 			.maxlen		= sizeof(int),
4379 			.mode		= 0644,
4380 			.proc_handler	= addrconf_sysctl_forward,
4381 		},
4382 		{
4383 			.procname	= "hop_limit",
4384 			.data		= &ipv6_devconf.hop_limit,
4385 			.maxlen		= sizeof(int),
4386 			.mode		= 0644,
4387 			.proc_handler	= proc_dointvec,
4388 		},
4389 		{
4390 			.procname	= "mtu",
4391 			.data		= &ipv6_devconf.mtu6,
4392 			.maxlen		= sizeof(int),
4393 			.mode		= 0644,
4394 			.proc_handler	= proc_dointvec,
4395 		},
4396 		{
4397 			.procname	= "accept_ra",
4398 			.data		= &ipv6_devconf.accept_ra,
4399 			.maxlen		= sizeof(int),
4400 			.mode		= 0644,
4401 			.proc_handler	= proc_dointvec,
4402 		},
4403 		{
4404 			.procname	= "accept_redirects",
4405 			.data		= &ipv6_devconf.accept_redirects,
4406 			.maxlen		= sizeof(int),
4407 			.mode		= 0644,
4408 			.proc_handler	= proc_dointvec,
4409 		},
4410 		{
4411 			.procname	= "autoconf",
4412 			.data		= &ipv6_devconf.autoconf,
4413 			.maxlen		= sizeof(int),
4414 			.mode		= 0644,
4415 			.proc_handler	= proc_dointvec,
4416 		},
4417 		{
4418 			.procname	= "dad_transmits",
4419 			.data		= &ipv6_devconf.dad_transmits,
4420 			.maxlen		= sizeof(int),
4421 			.mode		= 0644,
4422 			.proc_handler	= proc_dointvec,
4423 		},
4424 		{
4425 			.procname	= "router_solicitations",
4426 			.data		= &ipv6_devconf.rtr_solicits,
4427 			.maxlen		= sizeof(int),
4428 			.mode		= 0644,
4429 			.proc_handler	= proc_dointvec,
4430 		},
4431 		{
4432 			.procname	= "router_solicitation_interval",
4433 			.data		= &ipv6_devconf.rtr_solicit_interval,
4434 			.maxlen		= sizeof(int),
4435 			.mode		= 0644,
4436 			.proc_handler	= proc_dointvec_jiffies,
4437 		},
4438 		{
4439 			.procname	= "router_solicitation_delay",
4440 			.data		= &ipv6_devconf.rtr_solicit_delay,
4441 			.maxlen		= sizeof(int),
4442 			.mode		= 0644,
4443 			.proc_handler	= proc_dointvec_jiffies,
4444 		},
4445 		{
4446 			.procname	= "force_mld_version",
4447 			.data		= &ipv6_devconf.force_mld_version,
4448 			.maxlen		= sizeof(int),
4449 			.mode		= 0644,
4450 			.proc_handler	= proc_dointvec,
4451 		},
4452 #ifdef CONFIG_IPV6_PRIVACY
4453 		{
4454 			.procname	= "use_tempaddr",
4455 			.data		= &ipv6_devconf.use_tempaddr,
4456 			.maxlen		= sizeof(int),
4457 			.mode		= 0644,
4458 			.proc_handler	= proc_dointvec,
4459 		},
4460 		{
4461 			.procname	= "temp_valid_lft",
4462 			.data		= &ipv6_devconf.temp_valid_lft,
4463 			.maxlen		= sizeof(int),
4464 			.mode		= 0644,
4465 			.proc_handler	= proc_dointvec,
4466 		},
4467 		{
4468 			.procname	= "temp_prefered_lft",
4469 			.data		= &ipv6_devconf.temp_prefered_lft,
4470 			.maxlen		= sizeof(int),
4471 			.mode		= 0644,
4472 			.proc_handler	= proc_dointvec,
4473 		},
4474 		{
4475 			.procname	= "regen_max_retry",
4476 			.data		= &ipv6_devconf.regen_max_retry,
4477 			.maxlen		= sizeof(int),
4478 			.mode		= 0644,
4479 			.proc_handler	= proc_dointvec,
4480 		},
4481 		{
4482 			.procname	= "max_desync_factor",
4483 			.data		= &ipv6_devconf.max_desync_factor,
4484 			.maxlen		= sizeof(int),
4485 			.mode		= 0644,
4486 			.proc_handler	= proc_dointvec,
4487 		},
4488 #endif
4489 		{
4490 			.procname	= "max_addresses",
4491 			.data		= &ipv6_devconf.max_addresses,
4492 			.maxlen		= sizeof(int),
4493 			.mode		= 0644,
4494 			.proc_handler	= proc_dointvec,
4495 		},
4496 		{
4497 			.procname	= "accept_ra_defrtr",
4498 			.data		= &ipv6_devconf.accept_ra_defrtr,
4499 			.maxlen		= sizeof(int),
4500 			.mode		= 0644,
4501 			.proc_handler	= proc_dointvec,
4502 		},
4503 		{
4504 			.procname	= "accept_ra_pinfo",
4505 			.data		= &ipv6_devconf.accept_ra_pinfo,
4506 			.maxlen		= sizeof(int),
4507 			.mode		= 0644,
4508 			.proc_handler	= proc_dointvec,
4509 		},
4510 #ifdef CONFIG_IPV6_ROUTER_PREF
4511 		{
4512 			.procname	= "accept_ra_rtr_pref",
4513 			.data		= &ipv6_devconf.accept_ra_rtr_pref,
4514 			.maxlen		= sizeof(int),
4515 			.mode		= 0644,
4516 			.proc_handler	= proc_dointvec,
4517 		},
4518 		{
4519 			.procname	= "router_probe_interval",
4520 			.data		= &ipv6_devconf.rtr_probe_interval,
4521 			.maxlen		= sizeof(int),
4522 			.mode		= 0644,
4523 			.proc_handler	= proc_dointvec_jiffies,
4524 		},
4525 #ifdef CONFIG_IPV6_ROUTE_INFO
4526 		{
4527 			.procname	= "accept_ra_rt_info_max_plen",
4528 			.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
4529 			.maxlen		= sizeof(int),
4530 			.mode		= 0644,
4531 			.proc_handler	= proc_dointvec,
4532 		},
4533 #endif
4534 #endif
4535 		{
4536 			.procname	= "proxy_ndp",
4537 			.data		= &ipv6_devconf.proxy_ndp,
4538 			.maxlen		= sizeof(int),
4539 			.mode		= 0644,
4540 			.proc_handler	= proc_dointvec,
4541 		},
4542 		{
4543 			.procname	= "accept_source_route",
4544 			.data		= &ipv6_devconf.accept_source_route,
4545 			.maxlen		= sizeof(int),
4546 			.mode		= 0644,
4547 			.proc_handler	= proc_dointvec,
4548 		},
4549 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4550 		{
4551 			.procname       = "optimistic_dad",
4552 			.data           = &ipv6_devconf.optimistic_dad,
4553 			.maxlen         = sizeof(int),
4554 			.mode           = 0644,
4555 			.proc_handler   = proc_dointvec,
4556 
4557 		},
4558 #endif
4559 #ifdef CONFIG_IPV6_MROUTE
4560 		{
4561 			.procname	= "mc_forwarding",
4562 			.data		= &ipv6_devconf.mc_forwarding,
4563 			.maxlen		= sizeof(int),
4564 			.mode		= 0444,
4565 			.proc_handler	= proc_dointvec,
4566 		},
4567 #endif
4568 		{
4569 			.procname	= "disable_ipv6",
4570 			.data		= &ipv6_devconf.disable_ipv6,
4571 			.maxlen		= sizeof(int),
4572 			.mode		= 0644,
4573 			.proc_handler	= addrconf_sysctl_disable,
4574 		},
4575 		{
4576 			.procname	= "accept_dad",
4577 			.data		= &ipv6_devconf.accept_dad,
4578 			.maxlen		= sizeof(int),
4579 			.mode		= 0644,
4580 			.proc_handler	= proc_dointvec,
4581 		},
4582 		{
4583 			.procname       = "force_tllao",
4584 			.data           = &ipv6_devconf.force_tllao,
4585 			.maxlen         = sizeof(int),
4586 			.mode           = 0644,
4587 			.proc_handler   = proc_dointvec
4588 		},
4589 		{
4590 			/* sentinel */
4591 		}
4592 	},
4593 };
4594 
4595 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4596 		struct inet6_dev *idev, struct ipv6_devconf *p)
4597 {
4598 	int i;
4599 	struct addrconf_sysctl_table *t;
4600 
4601 #define ADDRCONF_CTL_PATH_DEV	3
4602 
4603 	struct ctl_path addrconf_ctl_path[] = {
4604 		{ .procname = "net", },
4605 		{ .procname = "ipv6", },
4606 		{ .procname = "conf", },
4607 		{ /* to be set */ },
4608 		{ },
4609 	};
4610 
4611 
4612 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4613 	if (t == NULL)
4614 		goto out;
4615 
4616 	for (i = 0; t->addrconf_vars[i].data; i++) {
4617 		t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4618 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4619 		t->addrconf_vars[i].extra2 = net;
4620 	}
4621 
4622 	/*
4623 	 * Make a copy of dev_name, because '.procname' is regarded as const
4624 	 * by sysctl and we wouldn't want anyone to change it under our feet
4625 	 * (see SIOCSIFNAME).
4626 	 */
4627 	t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4628 	if (!t->dev_name)
4629 		goto free;
4630 
4631 	addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4632 
4633 	t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4634 			t->addrconf_vars);
4635 	if (t->sysctl_header == NULL)
4636 		goto free_procname;
4637 
4638 	p->sysctl = t;
4639 	return 0;
4640 
4641 free_procname:
4642 	kfree(t->dev_name);
4643 free:
4644 	kfree(t);
4645 out:
4646 	return -ENOBUFS;
4647 }
4648 
4649 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4650 {
4651 	struct addrconf_sysctl_table *t;
4652 
4653 	if (p->sysctl == NULL)
4654 		return;
4655 
4656 	t = p->sysctl;
4657 	p->sysctl = NULL;
4658 	unregister_net_sysctl_table(t->sysctl_header);
4659 	kfree(t->dev_name);
4660 	kfree(t);
4661 }
4662 
4663 static void addrconf_sysctl_register(struct inet6_dev *idev)
4664 {
4665 	neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4666 			      &ndisc_ifinfo_sysctl_change);
4667 	__addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4668 					idev, &idev->cnf);
4669 }
4670 
4671 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4672 {
4673 	__addrconf_sysctl_unregister(&idev->cnf);
4674 	neigh_sysctl_unregister(idev->nd_parms);
4675 }
4676 
4677 
4678 #endif
4679 
4680 static int __net_init addrconf_init_net(struct net *net)
4681 {
4682 	int err;
4683 	struct ipv6_devconf *all, *dflt;
4684 
4685 	err = -ENOMEM;
4686 	all = &ipv6_devconf;
4687 	dflt = &ipv6_devconf_dflt;
4688 
4689 	if (!net_eq(net, &init_net)) {
4690 		all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4691 		if (all == NULL)
4692 			goto err_alloc_all;
4693 
4694 		dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4695 		if (dflt == NULL)
4696 			goto err_alloc_dflt;
4697 	} else {
4698 		/* these will be inherited by all namespaces */
4699 		dflt->autoconf = ipv6_defaults.autoconf;
4700 		dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4701 	}
4702 
4703 	net->ipv6.devconf_all = all;
4704 	net->ipv6.devconf_dflt = dflt;
4705 
4706 #ifdef CONFIG_SYSCTL
4707 	err = __addrconf_sysctl_register(net, "all", NULL, all);
4708 	if (err < 0)
4709 		goto err_reg_all;
4710 
4711 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4712 	if (err < 0)
4713 		goto err_reg_dflt;
4714 #endif
4715 	return 0;
4716 
4717 #ifdef CONFIG_SYSCTL
4718 err_reg_dflt:
4719 	__addrconf_sysctl_unregister(all);
4720 err_reg_all:
4721 	kfree(dflt);
4722 #endif
4723 err_alloc_dflt:
4724 	kfree(all);
4725 err_alloc_all:
4726 	return err;
4727 }
4728 
4729 static void __net_exit addrconf_exit_net(struct net *net)
4730 {
4731 #ifdef CONFIG_SYSCTL
4732 	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4733 	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
4734 #endif
4735 	if (!net_eq(net, &init_net)) {
4736 		kfree(net->ipv6.devconf_dflt);
4737 		kfree(net->ipv6.devconf_all);
4738 	}
4739 }
4740 
4741 static struct pernet_operations addrconf_ops = {
4742 	.init = addrconf_init_net,
4743 	.exit = addrconf_exit_net,
4744 };
4745 
4746 /*
4747  *      Device notifier
4748  */
4749 
4750 int register_inet6addr_notifier(struct notifier_block *nb)
4751 {
4752 	return atomic_notifier_chain_register(&inet6addr_chain, nb);
4753 }
4754 EXPORT_SYMBOL(register_inet6addr_notifier);
4755 
4756 int unregister_inet6addr_notifier(struct notifier_block *nb)
4757 {
4758 	return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4759 }
4760 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4761 
4762 static struct rtnl_af_ops inet6_ops = {
4763 	.family		  = AF_INET6,
4764 	.fill_link_af	  = inet6_fill_link_af,
4765 	.get_link_af_size = inet6_get_link_af_size,
4766 };
4767 
4768 /*
4769  *	Init / cleanup code
4770  */
4771 
4772 int __init addrconf_init(void)
4773 {
4774 	int i, err;
4775 
4776 	err = ipv6_addr_label_init();
4777 	if (err < 0) {
4778 		printk(KERN_CRIT "IPv6 Addrconf:"
4779 		       " cannot initialize default policy table: %d.\n", err);
4780 		goto out;
4781 	}
4782 
4783 	err = register_pernet_subsys(&addrconf_ops);
4784 	if (err < 0)
4785 		goto out_addrlabel;
4786 
4787 	/* The addrconf netdev notifier requires that loopback_dev
4788 	 * has it's ipv6 private information allocated and setup
4789 	 * before it can bring up and give link-local addresses
4790 	 * to other devices which are up.
4791 	 *
4792 	 * Unfortunately, loopback_dev is not necessarily the first
4793 	 * entry in the global dev_base list of net devices.  In fact,
4794 	 * it is likely to be the very last entry on that list.
4795 	 * So this causes the notifier registry below to try and
4796 	 * give link-local addresses to all devices besides loopback_dev
4797 	 * first, then loopback_dev, which cases all the non-loopback_dev
4798 	 * devices to fail to get a link-local address.
4799 	 *
4800 	 * So, as a temporary fix, allocate the ipv6 structure for
4801 	 * loopback_dev first by hand.
4802 	 * Longer term, all of the dependencies ipv6 has upon the loopback
4803 	 * device and it being up should be removed.
4804 	 */
4805 	rtnl_lock();
4806 	if (!ipv6_add_dev(init_net.loopback_dev))
4807 		err = -ENOMEM;
4808 	rtnl_unlock();
4809 	if (err)
4810 		goto errlo;
4811 
4812 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4813 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4814 
4815 	register_netdevice_notifier(&ipv6_dev_notf);
4816 
4817 	addrconf_verify(0);
4818 
4819 	err = rtnl_af_register(&inet6_ops);
4820 	if (err < 0)
4821 		goto errout_af;
4822 
4823 	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4824 			      NULL);
4825 	if (err < 0)
4826 		goto errout;
4827 
4828 	/* Only the first call to __rtnl_register can fail */
4829 	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4830 	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4831 	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4832 			inet6_dump_ifaddr, NULL);
4833 	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4834 			inet6_dump_ifmcaddr, NULL);
4835 	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4836 			inet6_dump_ifacaddr, NULL);
4837 
4838 	ipv6_addr_label_rtnl_register();
4839 
4840 	return 0;
4841 errout:
4842 	rtnl_af_unregister(&inet6_ops);
4843 errout_af:
4844 	unregister_netdevice_notifier(&ipv6_dev_notf);
4845 errlo:
4846 	unregister_pernet_subsys(&addrconf_ops);
4847 out_addrlabel:
4848 	ipv6_addr_label_cleanup();
4849 out:
4850 	return err;
4851 }
4852 
4853 void addrconf_cleanup(void)
4854 {
4855 	struct net_device *dev;
4856 	int i;
4857 
4858 	unregister_netdevice_notifier(&ipv6_dev_notf);
4859 	unregister_pernet_subsys(&addrconf_ops);
4860 	ipv6_addr_label_cleanup();
4861 
4862 	rtnl_lock();
4863 
4864 	__rtnl_af_unregister(&inet6_ops);
4865 
4866 	/* clean dev list */
4867 	for_each_netdev(&init_net, dev) {
4868 		if (__in6_dev_get(dev) == NULL)
4869 			continue;
4870 		addrconf_ifdown(dev, 1);
4871 	}
4872 	addrconf_ifdown(init_net.loopback_dev, 2);
4873 
4874 	/*
4875 	 *	Check hash table.
4876 	 */
4877 	spin_lock_bh(&addrconf_hash_lock);
4878 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4879 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4880 	spin_unlock_bh(&addrconf_hash_lock);
4881 
4882 	del_timer(&addr_chk_timer);
4883 	rtnl_unlock();
4884 }
4885