xref: /linux/net/ipv6/addrconf.c (revision d39d0ed196aa1685bb24771e92f78633c66ac9cb)
1 /*
2  *	IPv6 Address [auto]configuration
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14 
15 /*
16  *	Changes:
17  *
18  *	Janos Farkas			:	delete timer on ifdown
19  *	<chexum@bankinf.banki.hu>
20  *	Andi Kleen			:	kill double kfree on module
21  *						unload.
22  *	Maciej W. Rozycki		:	FDDI support
23  *	sekiya@USAGI			:	Don't send too many RS
24  *						packets.
25  *	yoshfuji@USAGI			:       Fixed interval between DAD
26  *						packets.
27  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
28  *						address validation timer.
29  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
30  *						support.
31  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
32  *						address on a same interface.
33  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
34  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
35  *						seq_file.
36  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
37  *						selection; consider scope,
38  *						status etc.
39  */
40 
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
68 
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
80 
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
84 
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
87 
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90 
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
93 
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
99 
100 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
101 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
102 
103 #define ADDRCONF_TIMER_FUZZ_MINUS	(HZ > 50 ? HZ/50 : 1)
104 #define ADDRCONF_TIMER_FUZZ		(HZ / 4)
105 #define ADDRCONF_TIMER_FUZZ_MAX		(HZ)
106 
107 #ifdef CONFIG_SYSCTL
108 static void addrconf_sysctl_register(struct inet6_dev *idev);
109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
110 #else
111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
112 {
113 }
114 
115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
116 {
117 }
118 #endif
119 
120 #ifdef CONFIG_IPV6_PRIVACY
121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
123 static void ipv6_regen_rndid(unsigned long data);
124 #endif
125 
126 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
127 static int ipv6_count_addresses(struct inet6_dev *idev);
128 
129 /*
130  *	Configured unicast address hash table
131  */
132 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
133 static DEFINE_SPINLOCK(addrconf_hash_lock);
134 
135 static void addrconf_verify(unsigned long);
136 
137 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
138 static DEFINE_SPINLOCK(addrconf_verify_lock);
139 
140 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
141 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
142 
143 static void addrconf_type_change(struct net_device *dev,
144 				 unsigned long event);
145 static int addrconf_ifdown(struct net_device *dev, int how);
146 
147 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
148 static void addrconf_dad_timer(unsigned long data);
149 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
150 static void addrconf_dad_run(struct inet6_dev *idev);
151 static void addrconf_rs_timer(unsigned long data);
152 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
153 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
154 
155 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
156 				struct prefix_info *pinfo);
157 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
158 			       struct net_device *dev);
159 
160 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
161 
162 static struct ipv6_devconf ipv6_devconf __read_mostly = {
163 	.forwarding		= 0,
164 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
165 	.mtu6			= IPV6_MIN_MTU,
166 	.accept_ra		= 1,
167 	.accept_redirects	= 1,
168 	.autoconf		= 1,
169 	.force_mld_version	= 0,
170 	.dad_transmits		= 1,
171 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
172 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
173 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
174 #ifdef CONFIG_IPV6_PRIVACY
175 	.use_tempaddr 		= 0,
176 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
177 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
178 	.regen_max_retry	= REGEN_MAX_RETRY,
179 	.max_desync_factor	= MAX_DESYNC_FACTOR,
180 #endif
181 	.max_addresses		= IPV6_MAX_ADDRESSES,
182 	.accept_ra_defrtr	= 1,
183 	.accept_ra_pinfo	= 1,
184 #ifdef CONFIG_IPV6_ROUTER_PREF
185 	.accept_ra_rtr_pref	= 1,
186 	.rtr_probe_interval	= 60 * HZ,
187 #ifdef CONFIG_IPV6_ROUTE_INFO
188 	.accept_ra_rt_info_max_plen = 0,
189 #endif
190 #endif
191 	.proxy_ndp		= 0,
192 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
193 	.disable_ipv6		= 0,
194 	.accept_dad		= 1,
195 };
196 
197 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
198 	.forwarding		= 0,
199 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
200 	.mtu6			= IPV6_MIN_MTU,
201 	.accept_ra		= 1,
202 	.accept_redirects	= 1,
203 	.autoconf		= 1,
204 	.dad_transmits		= 1,
205 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
206 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
207 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
208 #ifdef CONFIG_IPV6_PRIVACY
209 	.use_tempaddr		= 0,
210 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
211 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
212 	.regen_max_retry	= REGEN_MAX_RETRY,
213 	.max_desync_factor	= MAX_DESYNC_FACTOR,
214 #endif
215 	.max_addresses		= IPV6_MAX_ADDRESSES,
216 	.accept_ra_defrtr	= 1,
217 	.accept_ra_pinfo	= 1,
218 #ifdef CONFIG_IPV6_ROUTER_PREF
219 	.accept_ra_rtr_pref	= 1,
220 	.rtr_probe_interval	= 60 * HZ,
221 #ifdef CONFIG_IPV6_ROUTE_INFO
222 	.accept_ra_rt_info_max_plen = 0,
223 #endif
224 #endif
225 	.proxy_ndp		= 0,
226 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
227 	.disable_ipv6		= 0,
228 	.accept_dad		= 1,
229 };
230 
231 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
232 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
233 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
234 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
235 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
236 
237 /* Check if a valid qdisc is available */
238 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
239 {
240 	return !qdisc_tx_is_noop(dev);
241 }
242 
243 /* Check if a route is valid prefix route */
244 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
245 {
246 	return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
247 }
248 
249 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
250 {
251 	if (del_timer(&ifp->timer))
252 		__in6_ifa_put(ifp);
253 }
254 
255 enum addrconf_timer_t {
256 	AC_NONE,
257 	AC_DAD,
258 	AC_RS,
259 };
260 
261 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
262 			       enum addrconf_timer_t what,
263 			       unsigned long when)
264 {
265 	if (!del_timer(&ifp->timer))
266 		in6_ifa_hold(ifp);
267 
268 	switch (what) {
269 	case AC_DAD:
270 		ifp->timer.function = addrconf_dad_timer;
271 		break;
272 	case AC_RS:
273 		ifp->timer.function = addrconf_rs_timer;
274 		break;
275 	default:
276 		break;
277 	}
278 	ifp->timer.expires = jiffies + when;
279 	add_timer(&ifp->timer);
280 }
281 
282 static int snmp6_alloc_dev(struct inet6_dev *idev)
283 {
284 	if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
285 			  sizeof(struct ipstats_mib),
286 			  __alignof__(struct ipstats_mib)) < 0)
287 		goto err_ip;
288 	if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
289 			  sizeof(struct icmpv6_mib),
290 			  __alignof__(struct icmpv6_mib)) < 0)
291 		goto err_icmp;
292 	if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
293 			  sizeof(struct icmpv6msg_mib),
294 			  __alignof__(struct icmpv6msg_mib)) < 0)
295 		goto err_icmpmsg;
296 
297 	return 0;
298 
299 err_icmpmsg:
300 	snmp_mib_free((void __percpu **)idev->stats.icmpv6);
301 err_icmp:
302 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
303 err_ip:
304 	return -ENOMEM;
305 }
306 
307 static void snmp6_free_dev(struct inet6_dev *idev)
308 {
309 	snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
310 	snmp_mib_free((void __percpu **)idev->stats.icmpv6);
311 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
312 }
313 
314 /* Nobody refers to this device, we may destroy it. */
315 
316 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
317 {
318 	struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
319 	kfree(idev);
320 }
321 
322 void in6_dev_finish_destroy(struct inet6_dev *idev)
323 {
324 	struct net_device *dev = idev->dev;
325 
326 	WARN_ON(!list_empty(&idev->addr_list));
327 	WARN_ON(idev->mc_list != NULL);
328 
329 #ifdef NET_REFCNT_DEBUG
330 	printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
331 #endif
332 	dev_put(dev);
333 	if (!idev->dead) {
334 		pr_warning("Freeing alive inet6 device %p\n", idev);
335 		return;
336 	}
337 	snmp6_free_dev(idev);
338 	call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
339 }
340 
341 EXPORT_SYMBOL(in6_dev_finish_destroy);
342 
343 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 {
345 	struct inet6_dev *ndev;
346 
347 	ASSERT_RTNL();
348 
349 	if (dev->mtu < IPV6_MIN_MTU)
350 		return NULL;
351 
352 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
353 
354 	if (ndev == NULL)
355 		return NULL;
356 
357 	rwlock_init(&ndev->lock);
358 	ndev->dev = dev;
359 	INIT_LIST_HEAD(&ndev->addr_list);
360 
361 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
362 	ndev->cnf.mtu6 = dev->mtu;
363 	ndev->cnf.sysctl = NULL;
364 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
365 	if (ndev->nd_parms == NULL) {
366 		kfree(ndev);
367 		return NULL;
368 	}
369 	if (ndev->cnf.forwarding)
370 		dev_disable_lro(dev);
371 	/* We refer to the device */
372 	dev_hold(dev);
373 
374 	if (snmp6_alloc_dev(ndev) < 0) {
375 		ADBG((KERN_WARNING
376 			"%s(): cannot allocate memory for statistics; dev=%s.\n",
377 			__func__, dev->name));
378 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
379 		ndev->dead = 1;
380 		in6_dev_finish_destroy(ndev);
381 		return NULL;
382 	}
383 
384 	if (snmp6_register_dev(ndev) < 0) {
385 		ADBG((KERN_WARNING
386 			"%s(): cannot create /proc/net/dev_snmp6/%s\n",
387 			__func__, dev->name));
388 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
389 		ndev->dead = 1;
390 		in6_dev_finish_destroy(ndev);
391 		return NULL;
392 	}
393 
394 	/* One reference from device.  We must do this before
395 	 * we invoke __ipv6_regen_rndid().
396 	 */
397 	in6_dev_hold(ndev);
398 
399 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
400 		ndev->cnf.accept_dad = -1;
401 
402 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
403 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
404 		printk(KERN_INFO
405 		       "%s: Disabled Multicast RS\n",
406 		       dev->name);
407 		ndev->cnf.rtr_solicits = 0;
408 	}
409 #endif
410 
411 #ifdef CONFIG_IPV6_PRIVACY
412 	INIT_LIST_HEAD(&ndev->tempaddr_list);
413 	setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
414 	if ((dev->flags&IFF_LOOPBACK) ||
415 	    dev->type == ARPHRD_TUNNEL ||
416 	    dev->type == ARPHRD_TUNNEL6 ||
417 	    dev->type == ARPHRD_SIT ||
418 	    dev->type == ARPHRD_NONE) {
419 		printk(KERN_INFO
420 		       "%s: Disabled Privacy Extensions\n",
421 		       dev->name);
422 		ndev->cnf.use_tempaddr = -1;
423 	} else {
424 		in6_dev_hold(ndev);
425 		ipv6_regen_rndid((unsigned long) ndev);
426 	}
427 #endif
428 
429 	if (netif_running(dev) && addrconf_qdisc_ok(dev))
430 		ndev->if_flags |= IF_READY;
431 
432 	ipv6_mc_init_dev(ndev);
433 	ndev->tstamp = jiffies;
434 	addrconf_sysctl_register(ndev);
435 	/* protected by rtnl_lock */
436 	rcu_assign_pointer(dev->ip6_ptr, ndev);
437 
438 	/* Join all-node multicast group */
439 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
440 
441 	return ndev;
442 }
443 
444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
445 {
446 	struct inet6_dev *idev;
447 
448 	ASSERT_RTNL();
449 
450 	idev = __in6_dev_get(dev);
451 	if (!idev) {
452 		idev = ipv6_add_dev(dev);
453 		if (!idev)
454 			return NULL;
455 	}
456 
457 	if (dev->flags&IFF_UP)
458 		ipv6_mc_up(idev);
459 	return idev;
460 }
461 
462 #ifdef CONFIG_SYSCTL
463 static void dev_forward_change(struct inet6_dev *idev)
464 {
465 	struct net_device *dev;
466 	struct inet6_ifaddr *ifa;
467 
468 	if (!idev)
469 		return;
470 	dev = idev->dev;
471 	if (idev->cnf.forwarding)
472 		dev_disable_lro(dev);
473 	if (dev && (dev->flags & IFF_MULTICAST)) {
474 		if (idev->cnf.forwarding)
475 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
476 		else
477 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
478 	}
479 
480 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
481 		if (ifa->flags&IFA_F_TENTATIVE)
482 			continue;
483 		if (idev->cnf.forwarding)
484 			addrconf_join_anycast(ifa);
485 		else
486 			addrconf_leave_anycast(ifa);
487 	}
488 }
489 
490 
491 static void addrconf_forward_change(struct net *net, __s32 newf)
492 {
493 	struct net_device *dev;
494 	struct inet6_dev *idev;
495 
496 	rcu_read_lock();
497 	for_each_netdev_rcu(net, dev) {
498 		idev = __in6_dev_get(dev);
499 		if (idev) {
500 			int changed = (!idev->cnf.forwarding) ^ (!newf);
501 			idev->cnf.forwarding = newf;
502 			if (changed)
503 				dev_forward_change(idev);
504 		}
505 	}
506 	rcu_read_unlock();
507 }
508 
509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
510 {
511 	struct net *net;
512 
513 	net = (struct net *)table->extra2;
514 	if (p == &net->ipv6.devconf_dflt->forwarding)
515 		return 0;
516 
517 	if (!rtnl_trylock()) {
518 		/* Restore the original values before restarting */
519 		*p = old;
520 		return restart_syscall();
521 	}
522 
523 	if (p == &net->ipv6.devconf_all->forwarding) {
524 		__s32 newf = net->ipv6.devconf_all->forwarding;
525 		net->ipv6.devconf_dflt->forwarding = newf;
526 		addrconf_forward_change(net, newf);
527 	} else if ((!*p) ^ (!old))
528 		dev_forward_change((struct inet6_dev *)table->extra1);
529 	rtnl_unlock();
530 
531 	if (*p)
532 		rt6_purge_dflt_routers(net);
533 	return 1;
534 }
535 #endif
536 
537 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
538 {
539 	struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
540 	kfree(ifp);
541 }
542 
543 /* Nobody refers to this ifaddr, destroy it */
544 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
545 {
546 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
547 
548 #ifdef NET_REFCNT_DEBUG
549 	printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
550 #endif
551 
552 	in6_dev_put(ifp->idev);
553 
554 	if (del_timer(&ifp->timer))
555 		pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
556 
557 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
558 		pr_warning("Freeing alive inet6 address %p\n", ifp);
559 		return;
560 	}
561 	dst_release(&ifp->rt->dst);
562 
563 	call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
564 }
565 
566 static void
567 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
568 {
569 	struct list_head *p;
570 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
571 
572 	/*
573 	 * Each device address list is sorted in order of scope -
574 	 * global before linklocal.
575 	 */
576 	list_for_each(p, &idev->addr_list) {
577 		struct inet6_ifaddr *ifa
578 			= list_entry(p, struct inet6_ifaddr, if_list);
579 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
580 			break;
581 	}
582 
583 	list_add_tail(&ifp->if_list, p);
584 }
585 
586 static u32 ipv6_addr_hash(const struct in6_addr *addr)
587 {
588 	/*
589 	 * We perform the hash function over the last 64 bits of the address
590 	 * This will include the IEEE address token on links that support it.
591 	 */
592 	return jhash_2words((__force u32)addr->s6_addr32[2],
593 			    (__force u32)addr->s6_addr32[3], 0)
594 		& (IN6_ADDR_HSIZE - 1);
595 }
596 
597 /* On success it returns ifp with increased reference count */
598 
599 static struct inet6_ifaddr *
600 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
601 	      int scope, u32 flags)
602 {
603 	struct inet6_ifaddr *ifa = NULL;
604 	struct rt6_info *rt;
605 	unsigned int hash;
606 	int err = 0;
607 	int addr_type = ipv6_addr_type(addr);
608 
609 	if (addr_type == IPV6_ADDR_ANY ||
610 	    addr_type & IPV6_ADDR_MULTICAST ||
611 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
612 	     addr_type & IPV6_ADDR_LOOPBACK))
613 		return ERR_PTR(-EADDRNOTAVAIL);
614 
615 	rcu_read_lock_bh();
616 	if (idev->dead) {
617 		err = -ENODEV;			/*XXX*/
618 		goto out2;
619 	}
620 
621 	if (idev->cnf.disable_ipv6) {
622 		err = -EACCES;
623 		goto out2;
624 	}
625 
626 	spin_lock(&addrconf_hash_lock);
627 
628 	/* Ignore adding duplicate addresses on an interface */
629 	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
630 		ADBG(("ipv6_add_addr: already assigned\n"));
631 		err = -EEXIST;
632 		goto out;
633 	}
634 
635 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
636 
637 	if (ifa == NULL) {
638 		ADBG(("ipv6_add_addr: malloc failed\n"));
639 		err = -ENOBUFS;
640 		goto out;
641 	}
642 
643 	rt = addrconf_dst_alloc(idev, addr, 0);
644 	if (IS_ERR(rt)) {
645 		err = PTR_ERR(rt);
646 		goto out;
647 	}
648 
649 	ipv6_addr_copy(&ifa->addr, addr);
650 
651 	spin_lock_init(&ifa->lock);
652 	spin_lock_init(&ifa->state_lock);
653 	init_timer(&ifa->timer);
654 	INIT_HLIST_NODE(&ifa->addr_lst);
655 	ifa->timer.data = (unsigned long) ifa;
656 	ifa->scope = scope;
657 	ifa->prefix_len = pfxlen;
658 	ifa->flags = flags | IFA_F_TENTATIVE;
659 	ifa->cstamp = ifa->tstamp = jiffies;
660 
661 	ifa->rt = rt;
662 
663 	/*
664 	 * part one of RFC 4429, section 3.3
665 	 * We should not configure an address as
666 	 * optimistic if we do not yet know the link
667 	 * layer address of our nexhop router
668 	 */
669 
670 	if (rt->rt6i_nexthop == NULL)
671 		ifa->flags &= ~IFA_F_OPTIMISTIC;
672 
673 	ifa->idev = idev;
674 	in6_dev_hold(idev);
675 	/* For caller */
676 	in6_ifa_hold(ifa);
677 
678 	/* Add to big hash table */
679 	hash = ipv6_addr_hash(addr);
680 
681 	hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
682 	spin_unlock(&addrconf_hash_lock);
683 
684 	write_lock(&idev->lock);
685 	/* Add to inet6_dev unicast addr list. */
686 	ipv6_link_dev_addr(idev, ifa);
687 
688 #ifdef CONFIG_IPV6_PRIVACY
689 	if (ifa->flags&IFA_F_TEMPORARY) {
690 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
691 		in6_ifa_hold(ifa);
692 	}
693 #endif
694 
695 	in6_ifa_hold(ifa);
696 	write_unlock(&idev->lock);
697 out2:
698 	rcu_read_unlock_bh();
699 
700 	if (likely(err == 0))
701 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
702 	else {
703 		kfree(ifa);
704 		ifa = ERR_PTR(err);
705 	}
706 
707 	return ifa;
708 out:
709 	spin_unlock(&addrconf_hash_lock);
710 	goto out2;
711 }
712 
713 /* This function wants to get referenced ifp and releases it before return */
714 
715 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
716 {
717 	struct inet6_ifaddr *ifa, *ifn;
718 	struct inet6_dev *idev = ifp->idev;
719 	int state;
720 	int hash;
721 	int deleted = 0, onlink = 0;
722 	unsigned long expires = jiffies;
723 
724 	hash = ipv6_addr_hash(&ifp->addr);
725 
726 	spin_lock_bh(&ifp->state_lock);
727 	state = ifp->state;
728 	ifp->state = INET6_IFADDR_STATE_DEAD;
729 	spin_unlock_bh(&ifp->state_lock);
730 
731 	if (state == INET6_IFADDR_STATE_DEAD)
732 		goto out;
733 
734 	spin_lock_bh(&addrconf_hash_lock);
735 	hlist_del_init_rcu(&ifp->addr_lst);
736 	spin_unlock_bh(&addrconf_hash_lock);
737 
738 	write_lock_bh(&idev->lock);
739 #ifdef CONFIG_IPV6_PRIVACY
740 	if (ifp->flags&IFA_F_TEMPORARY) {
741 		list_del(&ifp->tmp_list);
742 		if (ifp->ifpub) {
743 			in6_ifa_put(ifp->ifpub);
744 			ifp->ifpub = NULL;
745 		}
746 		__in6_ifa_put(ifp);
747 	}
748 #endif
749 
750 	list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
751 		if (ifa == ifp) {
752 			list_del_init(&ifp->if_list);
753 			__in6_ifa_put(ifp);
754 
755 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
756 				break;
757 			deleted = 1;
758 			continue;
759 		} else if (ifp->flags & IFA_F_PERMANENT) {
760 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
761 					      ifp->prefix_len)) {
762 				if (ifa->flags & IFA_F_PERMANENT) {
763 					onlink = 1;
764 					if (deleted)
765 						break;
766 				} else {
767 					unsigned long lifetime;
768 
769 					if (!onlink)
770 						onlink = -1;
771 
772 					spin_lock(&ifa->lock);
773 
774 					lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
775 					/*
776 					 * Note: Because this address is
777 					 * not permanent, lifetime <
778 					 * LONG_MAX / HZ here.
779 					 */
780 					if (time_before(expires,
781 							ifa->tstamp + lifetime * HZ))
782 						expires = ifa->tstamp + lifetime * HZ;
783 					spin_unlock(&ifa->lock);
784 				}
785 			}
786 		}
787 	}
788 	write_unlock_bh(&idev->lock);
789 
790 	addrconf_del_timer(ifp);
791 
792 	ipv6_ifa_notify(RTM_DELADDR, ifp);
793 
794 	atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
795 
796 	/*
797 	 * Purge or update corresponding prefix
798 	 *
799 	 * 1) we don't purge prefix here if address was not permanent.
800 	 *    prefix is managed by its own lifetime.
801 	 * 2) if there're no addresses, delete prefix.
802 	 * 3) if there're still other permanent address(es),
803 	 *    corresponding prefix is still permanent.
804 	 * 4) otherwise, update prefix lifetime to the
805 	 *    longest valid lifetime among the corresponding
806 	 *    addresses on the device.
807 	 *    Note: subsequent RA will update lifetime.
808 	 *
809 	 * --yoshfuji
810 	 */
811 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
812 		struct in6_addr prefix;
813 		struct rt6_info *rt;
814 		struct net *net = dev_net(ifp->idev->dev);
815 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
816 		rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
817 
818 		if (rt && addrconf_is_prefix_route(rt)) {
819 			if (onlink == 0) {
820 				ip6_del_rt(rt);
821 				rt = NULL;
822 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
823 				rt->rt6i_expires = expires;
824 				rt->rt6i_flags |= RTF_EXPIRES;
825 			}
826 		}
827 		dst_release(&rt->dst);
828 	}
829 
830 out:
831 	in6_ifa_put(ifp);
832 }
833 
834 #ifdef CONFIG_IPV6_PRIVACY
835 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
836 {
837 	struct inet6_dev *idev = ifp->idev;
838 	struct in6_addr addr, *tmpaddr;
839 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
840 	unsigned long regen_advance;
841 	int tmp_plen;
842 	int ret = 0;
843 	int max_addresses;
844 	u32 addr_flags;
845 
846 	write_lock(&idev->lock);
847 	if (ift) {
848 		spin_lock_bh(&ift->lock);
849 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
850 		spin_unlock_bh(&ift->lock);
851 		tmpaddr = &addr;
852 	} else {
853 		tmpaddr = NULL;
854 	}
855 retry:
856 	in6_dev_hold(idev);
857 	if (idev->cnf.use_tempaddr <= 0) {
858 		write_unlock(&idev->lock);
859 		printk(KERN_INFO
860 			"ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
861 		in6_dev_put(idev);
862 		ret = -1;
863 		goto out;
864 	}
865 	spin_lock_bh(&ifp->lock);
866 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
867 		idev->cnf.use_tempaddr = -1;	/*XXX*/
868 		spin_unlock_bh(&ifp->lock);
869 		write_unlock(&idev->lock);
870 		printk(KERN_WARNING
871 			"ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
872 		in6_dev_put(idev);
873 		ret = -1;
874 		goto out;
875 	}
876 	in6_ifa_hold(ifp);
877 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
878 	if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
879 		spin_unlock_bh(&ifp->lock);
880 		write_unlock(&idev->lock);
881 		printk(KERN_WARNING
882 			"ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
883 		in6_ifa_put(ifp);
884 		in6_dev_put(idev);
885 		ret = -1;
886 		goto out;
887 	}
888 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
889 	tmp_valid_lft = min_t(__u32,
890 			      ifp->valid_lft,
891 			      idev->cnf.temp_valid_lft);
892 	tmp_prefered_lft = min_t(__u32,
893 				 ifp->prefered_lft,
894 				 idev->cnf.temp_prefered_lft -
895 				 idev->cnf.max_desync_factor);
896 	tmp_plen = ifp->prefix_len;
897 	max_addresses = idev->cnf.max_addresses;
898 	tmp_cstamp = ifp->cstamp;
899 	tmp_tstamp = ifp->tstamp;
900 	spin_unlock_bh(&ifp->lock);
901 
902 	regen_advance = idev->cnf.regen_max_retry *
903 	                idev->cnf.dad_transmits *
904 	                idev->nd_parms->retrans_time / HZ;
905 	write_unlock(&idev->lock);
906 
907 	/* A temporary address is created only if this calculated Preferred
908 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
909 	 * an implementation must not create a temporary address with a zero
910 	 * Preferred Lifetime.
911 	 */
912 	if (tmp_prefered_lft <= regen_advance) {
913 		in6_ifa_put(ifp);
914 		in6_dev_put(idev);
915 		ret = -1;
916 		goto out;
917 	}
918 
919 	addr_flags = IFA_F_TEMPORARY;
920 	/* set in addrconf_prefix_rcv() */
921 	if (ifp->flags & IFA_F_OPTIMISTIC)
922 		addr_flags |= IFA_F_OPTIMISTIC;
923 
924 	ift = !max_addresses ||
925 	      ipv6_count_addresses(idev) < max_addresses ?
926 		ipv6_add_addr(idev, &addr, tmp_plen,
927 			      ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
928 			      addr_flags) : NULL;
929 	if (!ift || IS_ERR(ift)) {
930 		in6_ifa_put(ifp);
931 		in6_dev_put(idev);
932 		printk(KERN_INFO
933 			"ipv6_create_tempaddr(): retry temporary address regeneration.\n");
934 		tmpaddr = &addr;
935 		write_lock(&idev->lock);
936 		goto retry;
937 	}
938 
939 	spin_lock_bh(&ift->lock);
940 	ift->ifpub = ifp;
941 	ift->valid_lft = tmp_valid_lft;
942 	ift->prefered_lft = tmp_prefered_lft;
943 	ift->cstamp = tmp_cstamp;
944 	ift->tstamp = tmp_tstamp;
945 	spin_unlock_bh(&ift->lock);
946 
947 	addrconf_dad_start(ift, 0);
948 	in6_ifa_put(ift);
949 	in6_dev_put(idev);
950 out:
951 	return ret;
952 }
953 #endif
954 
955 /*
956  *	Choose an appropriate source address (RFC3484)
957  */
958 enum {
959 	IPV6_SADDR_RULE_INIT = 0,
960 	IPV6_SADDR_RULE_LOCAL,
961 	IPV6_SADDR_RULE_SCOPE,
962 	IPV6_SADDR_RULE_PREFERRED,
963 #ifdef CONFIG_IPV6_MIP6
964 	IPV6_SADDR_RULE_HOA,
965 #endif
966 	IPV6_SADDR_RULE_OIF,
967 	IPV6_SADDR_RULE_LABEL,
968 #ifdef CONFIG_IPV6_PRIVACY
969 	IPV6_SADDR_RULE_PRIVACY,
970 #endif
971 	IPV6_SADDR_RULE_ORCHID,
972 	IPV6_SADDR_RULE_PREFIX,
973 	IPV6_SADDR_RULE_MAX
974 };
975 
976 struct ipv6_saddr_score {
977 	int			rule;
978 	int			addr_type;
979 	struct inet6_ifaddr	*ifa;
980 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
981 	int			scopedist;
982 	int			matchlen;
983 };
984 
985 struct ipv6_saddr_dst {
986 	const struct in6_addr *addr;
987 	int ifindex;
988 	int scope;
989 	int label;
990 	unsigned int prefs;
991 };
992 
993 static inline int ipv6_saddr_preferred(int type)
994 {
995 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
996 		return 1;
997 	return 0;
998 }
999 
1000 static int ipv6_get_saddr_eval(struct net *net,
1001 			       struct ipv6_saddr_score *score,
1002 			       struct ipv6_saddr_dst *dst,
1003 			       int i)
1004 {
1005 	int ret;
1006 
1007 	if (i <= score->rule) {
1008 		switch (i) {
1009 		case IPV6_SADDR_RULE_SCOPE:
1010 			ret = score->scopedist;
1011 			break;
1012 		case IPV6_SADDR_RULE_PREFIX:
1013 			ret = score->matchlen;
1014 			break;
1015 		default:
1016 			ret = !!test_bit(i, score->scorebits);
1017 		}
1018 		goto out;
1019 	}
1020 
1021 	switch (i) {
1022 	case IPV6_SADDR_RULE_INIT:
1023 		/* Rule 0: remember if hiscore is not ready yet */
1024 		ret = !!score->ifa;
1025 		break;
1026 	case IPV6_SADDR_RULE_LOCAL:
1027 		/* Rule 1: Prefer same address */
1028 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1029 		break;
1030 	case IPV6_SADDR_RULE_SCOPE:
1031 		/* Rule 2: Prefer appropriate scope
1032 		 *
1033 		 *      ret
1034 		 *       ^
1035 		 *    -1 |  d 15
1036 		 *    ---+--+-+---> scope
1037 		 *       |
1038 		 *       |             d is scope of the destination.
1039 		 *  B-d  |  \
1040 		 *       |   \      <- smaller scope is better if
1041 		 *  B-15 |    \        if scope is enough for destinaion.
1042 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1043 		 * d-C-1 | /
1044 		 *       |/         <- greater is better
1045 		 *   -C  /             if scope is not enough for destination.
1046 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1047 		 *
1048 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1049 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1050 		 * Assume B = 0 and we get C > 29.
1051 		 */
1052 		ret = __ipv6_addr_src_scope(score->addr_type);
1053 		if (ret >= dst->scope)
1054 			ret = -ret;
1055 		else
1056 			ret -= 128;	/* 30 is enough */
1057 		score->scopedist = ret;
1058 		break;
1059 	case IPV6_SADDR_RULE_PREFERRED:
1060 		/* Rule 3: Avoid deprecated and optimistic addresses */
1061 		ret = ipv6_saddr_preferred(score->addr_type) ||
1062 		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1063 		break;
1064 #ifdef CONFIG_IPV6_MIP6
1065 	case IPV6_SADDR_RULE_HOA:
1066 	    {
1067 		/* Rule 4: Prefer home address */
1068 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1069 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1070 		break;
1071 	    }
1072 #endif
1073 	case IPV6_SADDR_RULE_OIF:
1074 		/* Rule 5: Prefer outgoing interface */
1075 		ret = (!dst->ifindex ||
1076 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1077 		break;
1078 	case IPV6_SADDR_RULE_LABEL:
1079 		/* Rule 6: Prefer matching label */
1080 		ret = ipv6_addr_label(net,
1081 				      &score->ifa->addr, score->addr_type,
1082 				      score->ifa->idev->dev->ifindex) == dst->label;
1083 		break;
1084 #ifdef CONFIG_IPV6_PRIVACY
1085 	case IPV6_SADDR_RULE_PRIVACY:
1086 	    {
1087 		/* Rule 7: Prefer public address
1088 		 * Note: prefer temprary address if use_tempaddr >= 2
1089 		 */
1090 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1091 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1092 				score->ifa->idev->cnf.use_tempaddr >= 2;
1093 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1094 		break;
1095 	    }
1096 #endif
1097 	case IPV6_SADDR_RULE_ORCHID:
1098 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1099 		 *	    non-ORCHID vs non-ORCHID
1100 		 */
1101 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1102 			ipv6_addr_orchid(dst->addr));
1103 		break;
1104 	case IPV6_SADDR_RULE_PREFIX:
1105 		/* Rule 8: Use longest matching prefix */
1106 		score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1107 						       dst->addr);
1108 		break;
1109 	default:
1110 		ret = 0;
1111 	}
1112 
1113 	if (ret)
1114 		__set_bit(i, score->scorebits);
1115 	score->rule = i;
1116 out:
1117 	return ret;
1118 }
1119 
1120 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1121 		       const struct in6_addr *daddr, unsigned int prefs,
1122 		       struct in6_addr *saddr)
1123 {
1124 	struct ipv6_saddr_score scores[2],
1125 				*score = &scores[0], *hiscore = &scores[1];
1126 	struct ipv6_saddr_dst dst;
1127 	struct net_device *dev;
1128 	int dst_type;
1129 
1130 	dst_type = __ipv6_addr_type(daddr);
1131 	dst.addr = daddr;
1132 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1133 	dst.scope = __ipv6_addr_src_scope(dst_type);
1134 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1135 	dst.prefs = prefs;
1136 
1137 	hiscore->rule = -1;
1138 	hiscore->ifa = NULL;
1139 
1140 	rcu_read_lock();
1141 
1142 	for_each_netdev_rcu(net, dev) {
1143 		struct inet6_dev *idev;
1144 
1145 		/* Candidate Source Address (section 4)
1146 		 *  - multicast and link-local destination address,
1147 		 *    the set of candidate source address MUST only
1148 		 *    include addresses assigned to interfaces
1149 		 *    belonging to the same link as the outgoing
1150 		 *    interface.
1151 		 * (- For site-local destination addresses, the
1152 		 *    set of candidate source addresses MUST only
1153 		 *    include addresses assigned to interfaces
1154 		 *    belonging to the same site as the outgoing
1155 		 *    interface.)
1156 		 */
1157 		if (((dst_type & IPV6_ADDR_MULTICAST) ||
1158 		     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1159 		    dst.ifindex && dev->ifindex != dst.ifindex)
1160 			continue;
1161 
1162 		idev = __in6_dev_get(dev);
1163 		if (!idev)
1164 			continue;
1165 
1166 		read_lock_bh(&idev->lock);
1167 		list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1168 			int i;
1169 
1170 			/*
1171 			 * - Tentative Address (RFC2462 section 5.4)
1172 			 *  - A tentative address is not considered
1173 			 *    "assigned to an interface" in the traditional
1174 			 *    sense, unless it is also flagged as optimistic.
1175 			 * - Candidate Source Address (section 4)
1176 			 *  - In any case, anycast addresses, multicast
1177 			 *    addresses, and the unspecified address MUST
1178 			 *    NOT be included in a candidate set.
1179 			 */
1180 			if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1181 			    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1182 				continue;
1183 
1184 			score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1185 
1186 			if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1187 				     score->addr_type & IPV6_ADDR_MULTICAST)) {
1188 				LIMIT_NETDEBUG(KERN_DEBUG
1189 					       "ADDRCONF: unspecified / multicast address "
1190 					       "assigned as unicast address on %s",
1191 					       dev->name);
1192 				continue;
1193 			}
1194 
1195 			score->rule = -1;
1196 			bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1197 
1198 			for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1199 				int minihiscore, miniscore;
1200 
1201 				minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1202 				miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1203 
1204 				if (minihiscore > miniscore) {
1205 					if (i == IPV6_SADDR_RULE_SCOPE &&
1206 					    score->scopedist > 0) {
1207 						/*
1208 						 * special case:
1209 						 * each remaining entry
1210 						 * has too small (not enough)
1211 						 * scope, because ifa entries
1212 						 * are sorted by their scope
1213 						 * values.
1214 						 */
1215 						goto try_nextdev;
1216 					}
1217 					break;
1218 				} else if (minihiscore < miniscore) {
1219 					if (hiscore->ifa)
1220 						in6_ifa_put(hiscore->ifa);
1221 
1222 					in6_ifa_hold(score->ifa);
1223 
1224 					swap(hiscore, score);
1225 
1226 					/* restore our iterator */
1227 					score->ifa = hiscore->ifa;
1228 
1229 					break;
1230 				}
1231 			}
1232 		}
1233 try_nextdev:
1234 		read_unlock_bh(&idev->lock);
1235 	}
1236 	rcu_read_unlock();
1237 
1238 	if (!hiscore->ifa)
1239 		return -EADDRNOTAVAIL;
1240 
1241 	ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1242 	in6_ifa_put(hiscore->ifa);
1243 	return 0;
1244 }
1245 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1246 
1247 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1248 		    unsigned char banned_flags)
1249 {
1250 	struct inet6_dev *idev;
1251 	int err = -EADDRNOTAVAIL;
1252 
1253 	rcu_read_lock();
1254 	idev = __in6_dev_get(dev);
1255 	if (idev) {
1256 		struct inet6_ifaddr *ifp;
1257 
1258 		read_lock_bh(&idev->lock);
1259 		list_for_each_entry(ifp, &idev->addr_list, if_list) {
1260 			if (ifp->scope == IFA_LINK &&
1261 			    !(ifp->flags & banned_flags)) {
1262 				ipv6_addr_copy(addr, &ifp->addr);
1263 				err = 0;
1264 				break;
1265 			}
1266 		}
1267 		read_unlock_bh(&idev->lock);
1268 	}
1269 	rcu_read_unlock();
1270 	return err;
1271 }
1272 
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275 	int cnt = 0;
1276 	struct inet6_ifaddr *ifp;
1277 
1278 	read_lock_bh(&idev->lock);
1279 	list_for_each_entry(ifp, &idev->addr_list, if_list)
1280 		cnt++;
1281 	read_unlock_bh(&idev->lock);
1282 	return cnt;
1283 }
1284 
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286 		  struct net_device *dev, int strict)
1287 {
1288 	struct inet6_ifaddr *ifp;
1289 	struct hlist_node *node;
1290 	unsigned int hash = ipv6_addr_hash(addr);
1291 
1292 	rcu_read_lock_bh();
1293 	hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1294 		if (!net_eq(dev_net(ifp->idev->dev), net))
1295 			continue;
1296 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1297 		    !(ifp->flags&IFA_F_TENTATIVE) &&
1298 		    (dev == NULL || ifp->idev->dev == dev ||
1299 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1300 			rcu_read_unlock_bh();
1301 			return 1;
1302 		}
1303 	}
1304 
1305 	rcu_read_unlock_bh();
1306 	return 0;
1307 }
1308 EXPORT_SYMBOL(ipv6_chk_addr);
1309 
1310 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1311 			       struct net_device *dev)
1312 {
1313 	unsigned int hash = ipv6_addr_hash(addr);
1314 	struct inet6_ifaddr *ifp;
1315 	struct hlist_node *node;
1316 
1317 	hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1318 		if (!net_eq(dev_net(ifp->idev->dev), net))
1319 			continue;
1320 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1321 			if (dev == NULL || ifp->idev->dev == dev)
1322 				return true;
1323 		}
1324 	}
1325 	return false;
1326 }
1327 
1328 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1329 {
1330 	struct inet6_dev *idev;
1331 	struct inet6_ifaddr *ifa;
1332 	int	onlink;
1333 
1334 	onlink = 0;
1335 	rcu_read_lock();
1336 	idev = __in6_dev_get(dev);
1337 	if (idev) {
1338 		read_lock_bh(&idev->lock);
1339 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1340 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1341 						   ifa->prefix_len);
1342 			if (onlink)
1343 				break;
1344 		}
1345 		read_unlock_bh(&idev->lock);
1346 	}
1347 	rcu_read_unlock();
1348 	return onlink;
1349 }
1350 
1351 EXPORT_SYMBOL(ipv6_chk_prefix);
1352 
1353 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1354 				     struct net_device *dev, int strict)
1355 {
1356 	struct inet6_ifaddr *ifp, *result = NULL;
1357 	unsigned int hash = ipv6_addr_hash(addr);
1358 	struct hlist_node *node;
1359 
1360 	rcu_read_lock_bh();
1361 	hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1362 		if (!net_eq(dev_net(ifp->idev->dev), net))
1363 			continue;
1364 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1365 			if (dev == NULL || ifp->idev->dev == dev ||
1366 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1367 				result = ifp;
1368 				in6_ifa_hold(ifp);
1369 				break;
1370 			}
1371 		}
1372 	}
1373 	rcu_read_unlock_bh();
1374 
1375 	return result;
1376 }
1377 
1378 /* Gets referenced address, destroys ifaddr */
1379 
1380 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1381 {
1382 	if (ifp->flags&IFA_F_PERMANENT) {
1383 		spin_lock_bh(&ifp->lock);
1384 		addrconf_del_timer(ifp);
1385 		ifp->flags |= IFA_F_TENTATIVE;
1386 		if (dad_failed)
1387 			ifp->flags |= IFA_F_DADFAILED;
1388 		spin_unlock_bh(&ifp->lock);
1389 		if (dad_failed)
1390 			ipv6_ifa_notify(0, ifp);
1391 		in6_ifa_put(ifp);
1392 #ifdef CONFIG_IPV6_PRIVACY
1393 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1394 		struct inet6_ifaddr *ifpub;
1395 		spin_lock_bh(&ifp->lock);
1396 		ifpub = ifp->ifpub;
1397 		if (ifpub) {
1398 			in6_ifa_hold(ifpub);
1399 			spin_unlock_bh(&ifp->lock);
1400 			ipv6_create_tempaddr(ifpub, ifp);
1401 			in6_ifa_put(ifpub);
1402 		} else {
1403 			spin_unlock_bh(&ifp->lock);
1404 		}
1405 		ipv6_del_addr(ifp);
1406 #endif
1407 	} else
1408 		ipv6_del_addr(ifp);
1409 }
1410 
1411 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1412 {
1413 	int err = -ENOENT;
1414 
1415 	spin_lock(&ifp->state_lock);
1416 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
1417 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
1418 		err = 0;
1419 	}
1420 	spin_unlock(&ifp->state_lock);
1421 
1422 	return err;
1423 }
1424 
1425 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1426 {
1427 	struct inet6_dev *idev = ifp->idev;
1428 
1429 	if (addrconf_dad_end(ifp))
1430 		return;
1431 
1432 	if (net_ratelimit())
1433 		printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1434 			ifp->idev->dev->name, &ifp->addr);
1435 
1436 	if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1437 		struct in6_addr addr;
1438 
1439 		addr.s6_addr32[0] = htonl(0xfe800000);
1440 		addr.s6_addr32[1] = 0;
1441 
1442 		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1443 		    ipv6_addr_equal(&ifp->addr, &addr)) {
1444 			/* DAD failed for link-local based on MAC address */
1445 			idev->cnf.disable_ipv6 = 1;
1446 
1447 			printk(KERN_INFO "%s: IPv6 being disabled!\n",
1448 				ifp->idev->dev->name);
1449 		}
1450 	}
1451 
1452 	addrconf_dad_stop(ifp, 1);
1453 }
1454 
1455 /* Join to solicited addr multicast group. */
1456 
1457 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1458 {
1459 	struct in6_addr maddr;
1460 
1461 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1462 		return;
1463 
1464 	addrconf_addr_solict_mult(addr, &maddr);
1465 	ipv6_dev_mc_inc(dev, &maddr);
1466 }
1467 
1468 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1469 {
1470 	struct in6_addr maddr;
1471 
1472 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1473 		return;
1474 
1475 	addrconf_addr_solict_mult(addr, &maddr);
1476 	__ipv6_dev_mc_dec(idev, &maddr);
1477 }
1478 
1479 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1480 {
1481 	struct in6_addr addr;
1482 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1483 	if (ipv6_addr_any(&addr))
1484 		return;
1485 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1486 }
1487 
1488 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1489 {
1490 	struct in6_addr addr;
1491 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1492 	if (ipv6_addr_any(&addr))
1493 		return;
1494 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1495 }
1496 
1497 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1498 {
1499 	if (dev->addr_len != ETH_ALEN)
1500 		return -1;
1501 	memcpy(eui, dev->dev_addr, 3);
1502 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1503 
1504 	/*
1505 	 * The zSeries OSA network cards can be shared among various
1506 	 * OS instances, but the OSA cards have only one MAC address.
1507 	 * This leads to duplicate address conflicts in conjunction
1508 	 * with IPv6 if more than one instance uses the same card.
1509 	 *
1510 	 * The driver for these cards can deliver a unique 16-bit
1511 	 * identifier for each instance sharing the same card.  It is
1512 	 * placed instead of 0xFFFE in the interface identifier.  The
1513 	 * "u" bit of the interface identifier is not inverted in this
1514 	 * case.  Hence the resulting interface identifier has local
1515 	 * scope according to RFC2373.
1516 	 */
1517 	if (dev->dev_id) {
1518 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1519 		eui[4] = dev->dev_id & 0xFF;
1520 	} else {
1521 		eui[3] = 0xFF;
1522 		eui[4] = 0xFE;
1523 		eui[0] ^= 2;
1524 	}
1525 	return 0;
1526 }
1527 
1528 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1529 {
1530 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1531 	if (dev->addr_len != ARCNET_ALEN)
1532 		return -1;
1533 	memset(eui, 0, 7);
1534 	eui[7] = *(u8*)dev->dev_addr;
1535 	return 0;
1536 }
1537 
1538 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1539 {
1540 	if (dev->addr_len != INFINIBAND_ALEN)
1541 		return -1;
1542 	memcpy(eui, dev->dev_addr + 12, 8);
1543 	eui[0] |= 2;
1544 	return 0;
1545 }
1546 
1547 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1548 {
1549 	if (addr == 0)
1550 		return -1;
1551 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1552 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1553 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1554 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1555 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1556 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1557 	eui[1] = 0;
1558 	eui[2] = 0x5E;
1559 	eui[3] = 0xFE;
1560 	memcpy(eui + 4, &addr, 4);
1561 	return 0;
1562 }
1563 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1564 
1565 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1566 {
1567 	if (dev->priv_flags & IFF_ISATAP)
1568 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569 	return -1;
1570 }
1571 
1572 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1573 {
1574 	switch (dev->type) {
1575 	case ARPHRD_ETHER:
1576 	case ARPHRD_FDDI:
1577 	case ARPHRD_IEEE802_TR:
1578 		return addrconf_ifid_eui48(eui, dev);
1579 	case ARPHRD_ARCNET:
1580 		return addrconf_ifid_arcnet(eui, dev);
1581 	case ARPHRD_INFINIBAND:
1582 		return addrconf_ifid_infiniband(eui, dev);
1583 	case ARPHRD_SIT:
1584 		return addrconf_ifid_sit(eui, dev);
1585 	}
1586 	return -1;
1587 }
1588 
1589 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1590 {
1591 	int err = -1;
1592 	struct inet6_ifaddr *ifp;
1593 
1594 	read_lock_bh(&idev->lock);
1595 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
1596 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1597 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1598 			err = 0;
1599 			break;
1600 		}
1601 	}
1602 	read_unlock_bh(&idev->lock);
1603 	return err;
1604 }
1605 
1606 #ifdef CONFIG_IPV6_PRIVACY
1607 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1608 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1609 {
1610 regen:
1611 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1612 	idev->rndid[0] &= ~0x02;
1613 
1614 	/*
1615 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1616 	 * check if generated address is not inappropriate
1617 	 *
1618 	 *  - Reserved subnet anycast (RFC 2526)
1619 	 *	11111101 11....11 1xxxxxxx
1620 	 *  - ISATAP (RFC4214) 6.1
1621 	 *	00-00-5E-FE-xx-xx-xx-xx
1622 	 *  - value 0
1623 	 *  - XXX: already assigned to an address on the device
1624 	 */
1625 	if (idev->rndid[0] == 0xfd &&
1626 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1627 	    (idev->rndid[7]&0x80))
1628 		goto regen;
1629 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1630 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1631 			goto regen;
1632 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1633 			goto regen;
1634 	}
1635 
1636 	return 0;
1637 }
1638 
1639 static void ipv6_regen_rndid(unsigned long data)
1640 {
1641 	struct inet6_dev *idev = (struct inet6_dev *) data;
1642 	unsigned long expires;
1643 
1644 	rcu_read_lock_bh();
1645 	write_lock_bh(&idev->lock);
1646 
1647 	if (idev->dead)
1648 		goto out;
1649 
1650 	if (__ipv6_regen_rndid(idev) < 0)
1651 		goto out;
1652 
1653 	expires = jiffies +
1654 		idev->cnf.temp_prefered_lft * HZ -
1655 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1656 		idev->cnf.max_desync_factor * HZ;
1657 	if (time_before(expires, jiffies)) {
1658 		printk(KERN_WARNING
1659 			"ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1660 			idev->dev->name);
1661 		goto out;
1662 	}
1663 
1664 	if (!mod_timer(&idev->regen_timer, expires))
1665 		in6_dev_hold(idev);
1666 
1667 out:
1668 	write_unlock_bh(&idev->lock);
1669 	rcu_read_unlock_bh();
1670 	in6_dev_put(idev);
1671 }
1672 
1673 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1674 	int ret = 0;
1675 
1676 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1677 		ret = __ipv6_regen_rndid(idev);
1678 	return ret;
1679 }
1680 #endif
1681 
1682 /*
1683  *	Add prefix route.
1684  */
1685 
1686 static void
1687 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1688 		      unsigned long expires, u32 flags)
1689 {
1690 	struct fib6_config cfg = {
1691 		.fc_table = RT6_TABLE_PREFIX,
1692 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1693 		.fc_ifindex = dev->ifindex,
1694 		.fc_expires = expires,
1695 		.fc_dst_len = plen,
1696 		.fc_flags = RTF_UP | flags,
1697 		.fc_nlinfo.nl_net = dev_net(dev),
1698 		.fc_protocol = RTPROT_KERNEL,
1699 	};
1700 
1701 	ipv6_addr_copy(&cfg.fc_dst, pfx);
1702 
1703 	/* Prevent useless cloning on PtP SIT.
1704 	   This thing is done here expecting that the whole
1705 	   class of non-broadcast devices need not cloning.
1706 	 */
1707 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1708 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1709 		cfg.fc_flags |= RTF_NONEXTHOP;
1710 #endif
1711 
1712 	ip6_route_add(&cfg);
1713 }
1714 
1715 /* Create "default" multicast route to the interface */
1716 
1717 static void addrconf_add_mroute(struct net_device *dev)
1718 {
1719 	struct fib6_config cfg = {
1720 		.fc_table = RT6_TABLE_LOCAL,
1721 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1722 		.fc_ifindex = dev->ifindex,
1723 		.fc_dst_len = 8,
1724 		.fc_flags = RTF_UP,
1725 		.fc_nlinfo.nl_net = dev_net(dev),
1726 	};
1727 
1728 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1729 
1730 	ip6_route_add(&cfg);
1731 }
1732 
1733 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1734 static void sit_route_add(struct net_device *dev)
1735 {
1736 	struct fib6_config cfg = {
1737 		.fc_table = RT6_TABLE_MAIN,
1738 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1739 		.fc_ifindex = dev->ifindex,
1740 		.fc_dst_len = 96,
1741 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
1742 		.fc_nlinfo.nl_net = dev_net(dev),
1743 	};
1744 
1745 	/* prefix length - 96 bits "::d.d.d.d" */
1746 	ip6_route_add(&cfg);
1747 }
1748 #endif
1749 
1750 static void addrconf_add_lroute(struct net_device *dev)
1751 {
1752 	struct in6_addr addr;
1753 
1754 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1755 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
1756 }
1757 
1758 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1759 {
1760 	struct inet6_dev *idev;
1761 
1762 	ASSERT_RTNL();
1763 
1764 	idev = ipv6_find_idev(dev);
1765 	if (!idev)
1766 		return ERR_PTR(-ENOBUFS);
1767 
1768 	if (idev->cnf.disable_ipv6)
1769 		return ERR_PTR(-EACCES);
1770 
1771 	/* Add default multicast route */
1772 	addrconf_add_mroute(dev);
1773 
1774 	/* Add link local route */
1775 	addrconf_add_lroute(dev);
1776 	return idev;
1777 }
1778 
1779 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1780 {
1781 	struct prefix_info *pinfo;
1782 	__u32 valid_lft;
1783 	__u32 prefered_lft;
1784 	int addr_type;
1785 	struct inet6_dev *in6_dev;
1786 	struct net *net = dev_net(dev);
1787 
1788 	pinfo = (struct prefix_info *) opt;
1789 
1790 	if (len < sizeof(struct prefix_info)) {
1791 		ADBG(("addrconf: prefix option too short\n"));
1792 		return;
1793 	}
1794 
1795 	/*
1796 	 *	Validation checks ([ADDRCONF], page 19)
1797 	 */
1798 
1799 	addr_type = ipv6_addr_type(&pinfo->prefix);
1800 
1801 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1802 		return;
1803 
1804 	valid_lft = ntohl(pinfo->valid);
1805 	prefered_lft = ntohl(pinfo->prefered);
1806 
1807 	if (prefered_lft > valid_lft) {
1808 		if (net_ratelimit())
1809 			printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1810 		return;
1811 	}
1812 
1813 	in6_dev = in6_dev_get(dev);
1814 
1815 	if (in6_dev == NULL) {
1816 		if (net_ratelimit())
1817 			printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1818 		return;
1819 	}
1820 
1821 	/*
1822 	 *	Two things going on here:
1823 	 *	1) Add routes for on-link prefixes
1824 	 *	2) Configure prefixes with the auto flag set
1825 	 */
1826 
1827 	if (pinfo->onlink) {
1828 		struct rt6_info *rt;
1829 		unsigned long rt_expires;
1830 
1831 		/* Avoid arithmetic overflow. Really, we could
1832 		 * save rt_expires in seconds, likely valid_lft,
1833 		 * but it would require division in fib gc, that it
1834 		 * not good.
1835 		 */
1836 		if (HZ > USER_HZ)
1837 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1838 		else
1839 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1840 
1841 		if (addrconf_finite_timeout(rt_expires))
1842 			rt_expires *= HZ;
1843 
1844 		rt = rt6_lookup(net, &pinfo->prefix, NULL,
1845 				dev->ifindex, 1);
1846 
1847 		if (rt && addrconf_is_prefix_route(rt)) {
1848 			/* Autoconf prefix route */
1849 			if (valid_lft == 0) {
1850 				ip6_del_rt(rt);
1851 				rt = NULL;
1852 			} else if (addrconf_finite_timeout(rt_expires)) {
1853 				/* not infinity */
1854 				rt->rt6i_expires = jiffies + rt_expires;
1855 				rt->rt6i_flags |= RTF_EXPIRES;
1856 			} else {
1857 				rt->rt6i_flags &= ~RTF_EXPIRES;
1858 				rt->rt6i_expires = 0;
1859 			}
1860 		} else if (valid_lft) {
1861 			clock_t expires = 0;
1862 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1863 			if (addrconf_finite_timeout(rt_expires)) {
1864 				/* not infinity */
1865 				flags |= RTF_EXPIRES;
1866 				expires = jiffies_to_clock_t(rt_expires);
1867 			}
1868 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1869 					      dev, expires, flags);
1870 		}
1871 		if (rt)
1872 			dst_release(&rt->dst);
1873 	}
1874 
1875 	/* Try to figure out our local address for this prefix */
1876 
1877 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1878 		struct inet6_ifaddr * ifp;
1879 		struct in6_addr addr;
1880 		int create = 0, update_lft = 0;
1881 
1882 		if (pinfo->prefix_len == 64) {
1883 			memcpy(&addr, &pinfo->prefix, 8);
1884 			if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1885 			    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1886 				in6_dev_put(in6_dev);
1887 				return;
1888 			}
1889 			goto ok;
1890 		}
1891 		if (net_ratelimit())
1892 			printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1893 			       pinfo->prefix_len);
1894 		in6_dev_put(in6_dev);
1895 		return;
1896 
1897 ok:
1898 
1899 		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1900 
1901 		if (ifp == NULL && valid_lft) {
1902 			int max_addresses = in6_dev->cnf.max_addresses;
1903 			u32 addr_flags = 0;
1904 
1905 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1906 			if (in6_dev->cnf.optimistic_dad &&
1907 			    !net->ipv6.devconf_all->forwarding)
1908 				addr_flags = IFA_F_OPTIMISTIC;
1909 #endif
1910 
1911 			/* Do not allow to create too much of autoconfigured
1912 			 * addresses; this would be too easy way to crash kernel.
1913 			 */
1914 			if (!max_addresses ||
1915 			    ipv6_count_addresses(in6_dev) < max_addresses)
1916 				ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1917 						    addr_type&IPV6_ADDR_SCOPE_MASK,
1918 						    addr_flags);
1919 
1920 			if (!ifp || IS_ERR(ifp)) {
1921 				in6_dev_put(in6_dev);
1922 				return;
1923 			}
1924 
1925 			update_lft = create = 1;
1926 			ifp->cstamp = jiffies;
1927 			addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1928 		}
1929 
1930 		if (ifp) {
1931 			int flags;
1932 			unsigned long now;
1933 #ifdef CONFIG_IPV6_PRIVACY
1934 			struct inet6_ifaddr *ift;
1935 #endif
1936 			u32 stored_lft;
1937 
1938 			/* update lifetime (RFC2462 5.5.3 e) */
1939 			spin_lock(&ifp->lock);
1940 			now = jiffies;
1941 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1942 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1943 			else
1944 				stored_lft = 0;
1945 			if (!update_lft && stored_lft) {
1946 				if (valid_lft > MIN_VALID_LIFETIME ||
1947 				    valid_lft > stored_lft)
1948 					update_lft = 1;
1949 				else if (stored_lft <= MIN_VALID_LIFETIME) {
1950 					/* valid_lft <= stored_lft is always true */
1951 					/*
1952 					 * RFC 4862 Section 5.5.3e:
1953 					 * "Note that the preferred lifetime of
1954 					 *  the corresponding address is always
1955 					 *  reset to the Preferred Lifetime in
1956 					 *  the received Prefix Information
1957 					 *  option, regardless of whether the
1958 					 *  valid lifetime is also reset or
1959 					 *  ignored."
1960 					 *
1961 					 *  So if the preferred lifetime in
1962 					 *  this advertisement is different
1963 					 *  than what we have stored, but the
1964 					 *  valid lifetime is invalid, just
1965 					 *  reset prefered_lft.
1966 					 *
1967 					 *  We must set the valid lifetime
1968 					 *  to the stored lifetime since we'll
1969 					 *  be updating the timestamp below,
1970 					 *  else we'll set it back to the
1971 					 *  minumum.
1972 					 */
1973 					if (prefered_lft != ifp->prefered_lft) {
1974 						valid_lft = stored_lft;
1975 						update_lft = 1;
1976 					}
1977 				} else {
1978 					valid_lft = MIN_VALID_LIFETIME;
1979 					if (valid_lft < prefered_lft)
1980 						prefered_lft = valid_lft;
1981 					update_lft = 1;
1982 				}
1983 			}
1984 
1985 			if (update_lft) {
1986 				ifp->valid_lft = valid_lft;
1987 				ifp->prefered_lft = prefered_lft;
1988 				ifp->tstamp = now;
1989 				flags = ifp->flags;
1990 				ifp->flags &= ~IFA_F_DEPRECATED;
1991 				spin_unlock(&ifp->lock);
1992 
1993 				if (!(flags&IFA_F_TENTATIVE))
1994 					ipv6_ifa_notify(0, ifp);
1995 			} else
1996 				spin_unlock(&ifp->lock);
1997 
1998 #ifdef CONFIG_IPV6_PRIVACY
1999 			read_lock_bh(&in6_dev->lock);
2000 			/* update all temporary addresses in the list */
2001 			list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2002 				/*
2003 				 * When adjusting the lifetimes of an existing
2004 				 * temporary address, only lower the lifetimes.
2005 				 * Implementations must not increase the
2006 				 * lifetimes of an existing temporary address
2007 				 * when processing a Prefix Information Option.
2008 				 */
2009 				if (ifp != ift->ifpub)
2010 					continue;
2011 
2012 				spin_lock(&ift->lock);
2013 				flags = ift->flags;
2014 				if (ift->valid_lft > valid_lft &&
2015 				    ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2016 					ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2017 				if (ift->prefered_lft > prefered_lft &&
2018 				    ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2019 					ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2020 				spin_unlock(&ift->lock);
2021 				if (!(flags&IFA_F_TENTATIVE))
2022 					ipv6_ifa_notify(0, ift);
2023 			}
2024 
2025 			if (create && in6_dev->cnf.use_tempaddr > 0) {
2026 				/*
2027 				 * When a new public address is created as described in [ADDRCONF],
2028 				 * also create a new temporary address.
2029 				 */
2030 				read_unlock_bh(&in6_dev->lock);
2031 				ipv6_create_tempaddr(ifp, NULL);
2032 			} else {
2033 				read_unlock_bh(&in6_dev->lock);
2034 			}
2035 #endif
2036 			in6_ifa_put(ifp);
2037 			addrconf_verify(0);
2038 		}
2039 	}
2040 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2041 	in6_dev_put(in6_dev);
2042 }
2043 
2044 /*
2045  *	Set destination address.
2046  *	Special case for SIT interfaces where we create a new "virtual"
2047  *	device.
2048  */
2049 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2050 {
2051 	struct in6_ifreq ireq;
2052 	struct net_device *dev;
2053 	int err = -EINVAL;
2054 
2055 	rtnl_lock();
2056 
2057 	err = -EFAULT;
2058 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2059 		goto err_exit;
2060 
2061 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2062 
2063 	err = -ENODEV;
2064 	if (dev == NULL)
2065 		goto err_exit;
2066 
2067 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2068 	if (dev->type == ARPHRD_SIT) {
2069 		const struct net_device_ops *ops = dev->netdev_ops;
2070 		struct ifreq ifr;
2071 		struct ip_tunnel_parm p;
2072 
2073 		err = -EADDRNOTAVAIL;
2074 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2075 			goto err_exit;
2076 
2077 		memset(&p, 0, sizeof(p));
2078 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2079 		p.iph.saddr = 0;
2080 		p.iph.version = 4;
2081 		p.iph.ihl = 5;
2082 		p.iph.protocol = IPPROTO_IPV6;
2083 		p.iph.ttl = 64;
2084 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2085 
2086 		if (ops->ndo_do_ioctl) {
2087 			mm_segment_t oldfs = get_fs();
2088 
2089 			set_fs(KERNEL_DS);
2090 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2091 			set_fs(oldfs);
2092 		} else
2093 			err = -EOPNOTSUPP;
2094 
2095 		if (err == 0) {
2096 			err = -ENOBUFS;
2097 			dev = __dev_get_by_name(net, p.name);
2098 			if (!dev)
2099 				goto err_exit;
2100 			err = dev_open(dev);
2101 		}
2102 	}
2103 #endif
2104 
2105 err_exit:
2106 	rtnl_unlock();
2107 	return err;
2108 }
2109 
2110 /*
2111  *	Manual configuration of address on an interface
2112  */
2113 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2114 			  unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2115 			  __u32 valid_lft)
2116 {
2117 	struct inet6_ifaddr *ifp;
2118 	struct inet6_dev *idev;
2119 	struct net_device *dev;
2120 	int scope;
2121 	u32 flags;
2122 	clock_t expires;
2123 	unsigned long timeout;
2124 
2125 	ASSERT_RTNL();
2126 
2127 	if (plen > 128)
2128 		return -EINVAL;
2129 
2130 	/* check the lifetime */
2131 	if (!valid_lft || prefered_lft > valid_lft)
2132 		return -EINVAL;
2133 
2134 	dev = __dev_get_by_index(net, ifindex);
2135 	if (!dev)
2136 		return -ENODEV;
2137 
2138 	idev = addrconf_add_dev(dev);
2139 	if (IS_ERR(idev))
2140 		return PTR_ERR(idev);
2141 
2142 	scope = ipv6_addr_scope(pfx);
2143 
2144 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2145 	if (addrconf_finite_timeout(timeout)) {
2146 		expires = jiffies_to_clock_t(timeout * HZ);
2147 		valid_lft = timeout;
2148 		flags = RTF_EXPIRES;
2149 	} else {
2150 		expires = 0;
2151 		flags = 0;
2152 		ifa_flags |= IFA_F_PERMANENT;
2153 	}
2154 
2155 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2156 	if (addrconf_finite_timeout(timeout)) {
2157 		if (timeout == 0)
2158 			ifa_flags |= IFA_F_DEPRECATED;
2159 		prefered_lft = timeout;
2160 	}
2161 
2162 	ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2163 
2164 	if (!IS_ERR(ifp)) {
2165 		spin_lock_bh(&ifp->lock);
2166 		ifp->valid_lft = valid_lft;
2167 		ifp->prefered_lft = prefered_lft;
2168 		ifp->tstamp = jiffies;
2169 		spin_unlock_bh(&ifp->lock);
2170 
2171 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2172 				      expires, flags);
2173 		/*
2174 		 * Note that section 3.1 of RFC 4429 indicates
2175 		 * that the Optimistic flag should not be set for
2176 		 * manually configured addresses
2177 		 */
2178 		addrconf_dad_start(ifp, 0);
2179 		in6_ifa_put(ifp);
2180 		addrconf_verify(0);
2181 		return 0;
2182 	}
2183 
2184 	return PTR_ERR(ifp);
2185 }
2186 
2187 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2188 			  unsigned int plen)
2189 {
2190 	struct inet6_ifaddr *ifp;
2191 	struct inet6_dev *idev;
2192 	struct net_device *dev;
2193 
2194 	if (plen > 128)
2195 		return -EINVAL;
2196 
2197 	dev = __dev_get_by_index(net, ifindex);
2198 	if (!dev)
2199 		return -ENODEV;
2200 
2201 	if ((idev = __in6_dev_get(dev)) == NULL)
2202 		return -ENXIO;
2203 
2204 	read_lock_bh(&idev->lock);
2205 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2206 		if (ifp->prefix_len == plen &&
2207 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2208 			in6_ifa_hold(ifp);
2209 			read_unlock_bh(&idev->lock);
2210 
2211 			ipv6_del_addr(ifp);
2212 
2213 			/* If the last address is deleted administratively,
2214 			   disable IPv6 on this interface.
2215 			 */
2216 			if (list_empty(&idev->addr_list))
2217 				addrconf_ifdown(idev->dev, 1);
2218 			return 0;
2219 		}
2220 	}
2221 	read_unlock_bh(&idev->lock);
2222 	return -EADDRNOTAVAIL;
2223 }
2224 
2225 
2226 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2227 {
2228 	struct in6_ifreq ireq;
2229 	int err;
2230 
2231 	if (!capable(CAP_NET_ADMIN))
2232 		return -EPERM;
2233 
2234 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2235 		return -EFAULT;
2236 
2237 	rtnl_lock();
2238 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2239 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2240 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2241 	rtnl_unlock();
2242 	return err;
2243 }
2244 
2245 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2246 {
2247 	struct in6_ifreq ireq;
2248 	int err;
2249 
2250 	if (!capable(CAP_NET_ADMIN))
2251 		return -EPERM;
2252 
2253 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2254 		return -EFAULT;
2255 
2256 	rtnl_lock();
2257 	err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2258 			     ireq.ifr6_prefixlen);
2259 	rtnl_unlock();
2260 	return err;
2261 }
2262 
2263 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2264 		     int plen, int scope)
2265 {
2266 	struct inet6_ifaddr *ifp;
2267 
2268 	ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2269 	if (!IS_ERR(ifp)) {
2270 		spin_lock_bh(&ifp->lock);
2271 		ifp->flags &= ~IFA_F_TENTATIVE;
2272 		spin_unlock_bh(&ifp->lock);
2273 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2274 		in6_ifa_put(ifp);
2275 	}
2276 }
2277 
2278 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2279 static void sit_add_v4_addrs(struct inet6_dev *idev)
2280 {
2281 	struct in6_addr addr;
2282 	struct net_device *dev;
2283 	struct net *net = dev_net(idev->dev);
2284 	int scope;
2285 
2286 	ASSERT_RTNL();
2287 
2288 	memset(&addr, 0, sizeof(struct in6_addr));
2289 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2290 
2291 	if (idev->dev->flags&IFF_POINTOPOINT) {
2292 		addr.s6_addr32[0] = htonl(0xfe800000);
2293 		scope = IFA_LINK;
2294 	} else {
2295 		scope = IPV6_ADDR_COMPATv4;
2296 	}
2297 
2298 	if (addr.s6_addr32[3]) {
2299 		add_addr(idev, &addr, 128, scope);
2300 		return;
2301 	}
2302 
2303 	for_each_netdev(net, dev) {
2304 		struct in_device * in_dev = __in_dev_get_rtnl(dev);
2305 		if (in_dev && (dev->flags & IFF_UP)) {
2306 			struct in_ifaddr * ifa;
2307 
2308 			int flag = scope;
2309 
2310 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2311 				int plen;
2312 
2313 				addr.s6_addr32[3] = ifa->ifa_local;
2314 
2315 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2316 					continue;
2317 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2318 					if (idev->dev->flags&IFF_POINTOPOINT)
2319 						continue;
2320 					flag |= IFA_HOST;
2321 				}
2322 				if (idev->dev->flags&IFF_POINTOPOINT)
2323 					plen = 64;
2324 				else
2325 					plen = 96;
2326 
2327 				add_addr(idev, &addr, plen, flag);
2328 			}
2329 		}
2330 	}
2331 }
2332 #endif
2333 
2334 static void init_loopback(struct net_device *dev)
2335 {
2336 	struct inet6_dev  *idev;
2337 
2338 	/* ::1 */
2339 
2340 	ASSERT_RTNL();
2341 
2342 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2343 		printk(KERN_DEBUG "init loopback: add_dev failed\n");
2344 		return;
2345 	}
2346 
2347 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2348 }
2349 
2350 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2351 {
2352 	struct inet6_ifaddr * ifp;
2353 	u32 addr_flags = IFA_F_PERMANENT;
2354 
2355 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2356 	if (idev->cnf.optimistic_dad &&
2357 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2358 		addr_flags |= IFA_F_OPTIMISTIC;
2359 #endif
2360 
2361 
2362 	ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2363 	if (!IS_ERR(ifp)) {
2364 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2365 		addrconf_dad_start(ifp, 0);
2366 		in6_ifa_put(ifp);
2367 	}
2368 }
2369 
2370 static void addrconf_dev_config(struct net_device *dev)
2371 {
2372 	struct in6_addr addr;
2373 	struct inet6_dev    * idev;
2374 
2375 	ASSERT_RTNL();
2376 
2377 	if ((dev->type != ARPHRD_ETHER) &&
2378 	    (dev->type != ARPHRD_FDDI) &&
2379 	    (dev->type != ARPHRD_IEEE802_TR) &&
2380 	    (dev->type != ARPHRD_ARCNET) &&
2381 	    (dev->type != ARPHRD_INFINIBAND)) {
2382 		/* Alas, we support only Ethernet autoconfiguration. */
2383 		return;
2384 	}
2385 
2386 	idev = addrconf_add_dev(dev);
2387 	if (IS_ERR(idev))
2388 		return;
2389 
2390 	memset(&addr, 0, sizeof(struct in6_addr));
2391 	addr.s6_addr32[0] = htonl(0xFE800000);
2392 
2393 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2394 		addrconf_add_linklocal(idev, &addr);
2395 }
2396 
2397 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2398 static void addrconf_sit_config(struct net_device *dev)
2399 {
2400 	struct inet6_dev *idev;
2401 
2402 	ASSERT_RTNL();
2403 
2404 	/*
2405 	 * Configure the tunnel with one of our IPv4
2406 	 * addresses... we should configure all of
2407 	 * our v4 addrs in the tunnel
2408 	 */
2409 
2410 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2411 		printk(KERN_DEBUG "init sit: add_dev failed\n");
2412 		return;
2413 	}
2414 
2415 	if (dev->priv_flags & IFF_ISATAP) {
2416 		struct in6_addr addr;
2417 
2418 		ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2419 		addrconf_prefix_route(&addr, 64, dev, 0, 0);
2420 		if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2421 			addrconf_add_linklocal(idev, &addr);
2422 		return;
2423 	}
2424 
2425 	sit_add_v4_addrs(idev);
2426 
2427 	if (dev->flags&IFF_POINTOPOINT) {
2428 		addrconf_add_mroute(dev);
2429 		addrconf_add_lroute(dev);
2430 	} else
2431 		sit_route_add(dev);
2432 }
2433 #endif
2434 
2435 static inline int
2436 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2437 {
2438 	struct in6_addr lladdr;
2439 
2440 	if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2441 		addrconf_add_linklocal(idev, &lladdr);
2442 		return 0;
2443 	}
2444 	return -1;
2445 }
2446 
2447 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2448 {
2449 	struct net_device *link_dev;
2450 	struct net *net = dev_net(idev->dev);
2451 
2452 	/* first try to inherit the link-local address from the link device */
2453 	if (idev->dev->iflink &&
2454 	    (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2455 		if (!ipv6_inherit_linklocal(idev, link_dev))
2456 			return;
2457 	}
2458 	/* then try to inherit it from any device */
2459 	for_each_netdev(net, link_dev) {
2460 		if (!ipv6_inherit_linklocal(idev, link_dev))
2461 			return;
2462 	}
2463 	printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2464 }
2465 
2466 /*
2467  * Autoconfigure tunnel with a link-local address so routing protocols,
2468  * DHCPv6, MLD etc. can be run over the virtual link
2469  */
2470 
2471 static void addrconf_ip6_tnl_config(struct net_device *dev)
2472 {
2473 	struct inet6_dev *idev;
2474 
2475 	ASSERT_RTNL();
2476 
2477 	idev = addrconf_add_dev(dev);
2478 	if (IS_ERR(idev)) {
2479 		printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2480 		return;
2481 	}
2482 	ip6_tnl_add_linklocal(idev);
2483 }
2484 
2485 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2486 			   void * data)
2487 {
2488 	struct net_device *dev = (struct net_device *) data;
2489 	struct inet6_dev *idev = __in6_dev_get(dev);
2490 	int run_pending = 0;
2491 	int err;
2492 
2493 	switch (event) {
2494 	case NETDEV_REGISTER:
2495 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2496 			idev = ipv6_add_dev(dev);
2497 			if (!idev)
2498 				return notifier_from_errno(-ENOMEM);
2499 		}
2500 		break;
2501 
2502 	case NETDEV_UP:
2503 	case NETDEV_CHANGE:
2504 		if (dev->flags & IFF_SLAVE)
2505 			break;
2506 
2507 		if (event == NETDEV_UP) {
2508 			if (!addrconf_qdisc_ok(dev)) {
2509 				/* device is not ready yet. */
2510 				printk(KERN_INFO
2511 					"ADDRCONF(NETDEV_UP): %s: "
2512 					"link is not ready\n",
2513 					dev->name);
2514 				break;
2515 			}
2516 
2517 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
2518 				idev = ipv6_add_dev(dev);
2519 
2520 			if (idev) {
2521 				idev->if_flags |= IF_READY;
2522 				run_pending = 1;
2523 			}
2524 		} else {
2525 			if (!addrconf_qdisc_ok(dev)) {
2526 				/* device is still not ready. */
2527 				break;
2528 			}
2529 
2530 			if (idev) {
2531 				if (idev->if_flags & IF_READY)
2532 					/* device is already configured. */
2533 					break;
2534 				idev->if_flags |= IF_READY;
2535 			}
2536 
2537 			printk(KERN_INFO
2538 					"ADDRCONF(NETDEV_CHANGE): %s: "
2539 					"link becomes ready\n",
2540 					dev->name);
2541 
2542 			run_pending = 1;
2543 		}
2544 
2545 		switch (dev->type) {
2546 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2547 		case ARPHRD_SIT:
2548 			addrconf_sit_config(dev);
2549 			break;
2550 #endif
2551 		case ARPHRD_TUNNEL6:
2552 			addrconf_ip6_tnl_config(dev);
2553 			break;
2554 		case ARPHRD_LOOPBACK:
2555 			init_loopback(dev);
2556 			break;
2557 
2558 		default:
2559 			addrconf_dev_config(dev);
2560 			break;
2561 		}
2562 
2563 		if (idev) {
2564 			if (run_pending)
2565 				addrconf_dad_run(idev);
2566 
2567 			/*
2568 			 * If the MTU changed during the interface down,
2569 			 * when the interface up, the changed MTU must be
2570 			 * reflected in the idev as well as routers.
2571 			 */
2572 			if (idev->cnf.mtu6 != dev->mtu &&
2573 			    dev->mtu >= IPV6_MIN_MTU) {
2574 				rt6_mtu_change(dev, dev->mtu);
2575 				idev->cnf.mtu6 = dev->mtu;
2576 			}
2577 			idev->tstamp = jiffies;
2578 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2579 
2580 			/*
2581 			 * If the changed mtu during down is lower than
2582 			 * IPV6_MIN_MTU stop IPv6 on this interface.
2583 			 */
2584 			if (dev->mtu < IPV6_MIN_MTU)
2585 				addrconf_ifdown(dev, 1);
2586 		}
2587 		break;
2588 
2589 	case NETDEV_CHANGEMTU:
2590 		if (idev && dev->mtu >= IPV6_MIN_MTU) {
2591 			rt6_mtu_change(dev, dev->mtu);
2592 			idev->cnf.mtu6 = dev->mtu;
2593 			break;
2594 		}
2595 
2596 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2597 			idev = ipv6_add_dev(dev);
2598 			if (idev)
2599 				break;
2600 		}
2601 
2602 		/*
2603 		 * MTU falled under IPV6_MIN_MTU.
2604 		 * Stop IPv6 on this interface.
2605 		 */
2606 
2607 	case NETDEV_DOWN:
2608 	case NETDEV_UNREGISTER:
2609 		/*
2610 		 *	Remove all addresses from this interface.
2611 		 */
2612 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2613 		break;
2614 
2615 	case NETDEV_CHANGENAME:
2616 		if (idev) {
2617 			snmp6_unregister_dev(idev);
2618 			addrconf_sysctl_unregister(idev);
2619 			addrconf_sysctl_register(idev);
2620 			err = snmp6_register_dev(idev);
2621 			if (err)
2622 				return notifier_from_errno(err);
2623 		}
2624 		break;
2625 
2626 	case NETDEV_PRE_TYPE_CHANGE:
2627 	case NETDEV_POST_TYPE_CHANGE:
2628 		addrconf_type_change(dev, event);
2629 		break;
2630 	}
2631 
2632 	return NOTIFY_OK;
2633 }
2634 
2635 /*
2636  *	addrconf module should be notified of a device going up
2637  */
2638 static struct notifier_block ipv6_dev_notf = {
2639 	.notifier_call = addrconf_notify,
2640 };
2641 
2642 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2643 {
2644 	struct inet6_dev *idev;
2645 	ASSERT_RTNL();
2646 
2647 	idev = __in6_dev_get(dev);
2648 
2649 	if (event == NETDEV_POST_TYPE_CHANGE)
2650 		ipv6_mc_remap(idev);
2651 	else if (event == NETDEV_PRE_TYPE_CHANGE)
2652 		ipv6_mc_unmap(idev);
2653 }
2654 
2655 static int addrconf_ifdown(struct net_device *dev, int how)
2656 {
2657 	struct net *net = dev_net(dev);
2658 	struct inet6_dev *idev;
2659 	struct inet6_ifaddr *ifa;
2660 	LIST_HEAD(keep_list);
2661 	int state;
2662 
2663 	ASSERT_RTNL();
2664 
2665 	rt6_ifdown(net, dev);
2666 	neigh_ifdown(&nd_tbl, dev);
2667 
2668 	idev = __in6_dev_get(dev);
2669 	if (idev == NULL)
2670 		return -ENODEV;
2671 
2672 	/*
2673 	 * Step 1: remove reference to ipv6 device from parent device.
2674 	 *	   Do not dev_put!
2675 	 */
2676 	if (how) {
2677 		idev->dead = 1;
2678 
2679 		/* protected by rtnl_lock */
2680 		rcu_assign_pointer(dev->ip6_ptr, NULL);
2681 
2682 		/* Step 1.5: remove snmp6 entry */
2683 		snmp6_unregister_dev(idev);
2684 
2685 	}
2686 
2687 	write_lock_bh(&idev->lock);
2688 
2689 	/* Step 2: clear flags for stateless addrconf */
2690 	if (!how)
2691 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2692 
2693 #ifdef CONFIG_IPV6_PRIVACY
2694 	if (how && del_timer(&idev->regen_timer))
2695 		in6_dev_put(idev);
2696 
2697 	/* Step 3: clear tempaddr list */
2698 	while (!list_empty(&idev->tempaddr_list)) {
2699 		ifa = list_first_entry(&idev->tempaddr_list,
2700 				       struct inet6_ifaddr, tmp_list);
2701 		list_del(&ifa->tmp_list);
2702 		write_unlock_bh(&idev->lock);
2703 		spin_lock_bh(&ifa->lock);
2704 
2705 		if (ifa->ifpub) {
2706 			in6_ifa_put(ifa->ifpub);
2707 			ifa->ifpub = NULL;
2708 		}
2709 		spin_unlock_bh(&ifa->lock);
2710 		in6_ifa_put(ifa);
2711 		write_lock_bh(&idev->lock);
2712 	}
2713 #endif
2714 
2715 	while (!list_empty(&idev->addr_list)) {
2716 		ifa = list_first_entry(&idev->addr_list,
2717 				       struct inet6_ifaddr, if_list);
2718 		addrconf_del_timer(ifa);
2719 
2720 		/* If just doing link down, and address is permanent
2721 		   and not link-local, then retain it. */
2722 		if (!how &&
2723 		    (ifa->flags&IFA_F_PERMANENT) &&
2724 		    !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2725 			list_move_tail(&ifa->if_list, &keep_list);
2726 
2727 			/* If not doing DAD on this address, just keep it. */
2728 			if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2729 			    idev->cnf.accept_dad <= 0 ||
2730 			    (ifa->flags & IFA_F_NODAD))
2731 				continue;
2732 
2733 			/* If it was tentative already, no need to notify */
2734 			if (ifa->flags & IFA_F_TENTATIVE)
2735 				continue;
2736 
2737 			/* Flag it for later restoration when link comes up */
2738 			ifa->flags |= IFA_F_TENTATIVE;
2739 			ifa->state = INET6_IFADDR_STATE_DAD;
2740 
2741 			write_unlock_bh(&idev->lock);
2742 
2743 			in6_ifa_hold(ifa);
2744 		} else {
2745 			list_del(&ifa->if_list);
2746 
2747 			/* clear hash table */
2748 			spin_lock_bh(&addrconf_hash_lock);
2749 			hlist_del_init_rcu(&ifa->addr_lst);
2750 			spin_unlock_bh(&addrconf_hash_lock);
2751 
2752 			write_unlock_bh(&idev->lock);
2753 			spin_lock_bh(&ifa->state_lock);
2754 			state = ifa->state;
2755 			ifa->state = INET6_IFADDR_STATE_DEAD;
2756 			spin_unlock_bh(&ifa->state_lock);
2757 
2758 			if (state == INET6_IFADDR_STATE_DEAD)
2759 				goto put_ifa;
2760 		}
2761 
2762 		__ipv6_ifa_notify(RTM_DELADDR, ifa);
2763 		if (ifa->state == INET6_IFADDR_STATE_DEAD)
2764 			atomic_notifier_call_chain(&inet6addr_chain,
2765 						   NETDEV_DOWN, ifa);
2766 
2767 put_ifa:
2768 		in6_ifa_put(ifa);
2769 
2770 		write_lock_bh(&idev->lock);
2771 	}
2772 
2773 	list_splice(&keep_list, &idev->addr_list);
2774 
2775 	write_unlock_bh(&idev->lock);
2776 
2777 	/* Step 5: Discard multicast list */
2778 	if (how)
2779 		ipv6_mc_destroy_dev(idev);
2780 	else
2781 		ipv6_mc_down(idev);
2782 
2783 	idev->tstamp = jiffies;
2784 
2785 	/* Last: Shot the device (if unregistered) */
2786 	if (how) {
2787 		addrconf_sysctl_unregister(idev);
2788 		neigh_parms_release(&nd_tbl, idev->nd_parms);
2789 		neigh_ifdown(&nd_tbl, dev);
2790 		in6_dev_put(idev);
2791 	}
2792 	return 0;
2793 }
2794 
2795 static void addrconf_rs_timer(unsigned long data)
2796 {
2797 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2798 	struct inet6_dev *idev = ifp->idev;
2799 
2800 	read_lock(&idev->lock);
2801 	if (idev->dead || !(idev->if_flags & IF_READY))
2802 		goto out;
2803 
2804 	if (idev->cnf.forwarding)
2805 		goto out;
2806 
2807 	/* Announcement received after solicitation was sent */
2808 	if (idev->if_flags & IF_RA_RCVD)
2809 		goto out;
2810 
2811 	spin_lock(&ifp->lock);
2812 	if (ifp->probes++ < idev->cnf.rtr_solicits) {
2813 		/* The wait after the last probe can be shorter */
2814 		addrconf_mod_timer(ifp, AC_RS,
2815 				   (ifp->probes == idev->cnf.rtr_solicits) ?
2816 				   idev->cnf.rtr_solicit_delay :
2817 				   idev->cnf.rtr_solicit_interval);
2818 		spin_unlock(&ifp->lock);
2819 
2820 		ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2821 	} else {
2822 		spin_unlock(&ifp->lock);
2823 		/*
2824 		 * Note: we do not support deprecated "all on-link"
2825 		 * assumption any longer.
2826 		 */
2827 		printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2828 		       idev->dev->name);
2829 	}
2830 
2831 out:
2832 	read_unlock(&idev->lock);
2833 	in6_ifa_put(ifp);
2834 }
2835 
2836 /*
2837  *	Duplicate Address Detection
2838  */
2839 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2840 {
2841 	unsigned long rand_num;
2842 	struct inet6_dev *idev = ifp->idev;
2843 
2844 	if (ifp->flags & IFA_F_OPTIMISTIC)
2845 		rand_num = 0;
2846 	else
2847 		rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2848 
2849 	ifp->probes = idev->cnf.dad_transmits;
2850 	addrconf_mod_timer(ifp, AC_DAD, rand_num);
2851 }
2852 
2853 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2854 {
2855 	struct inet6_dev *idev = ifp->idev;
2856 	struct net_device *dev = idev->dev;
2857 
2858 	addrconf_join_solict(dev, &ifp->addr);
2859 
2860 	net_srandom(ifp->addr.s6_addr32[3]);
2861 
2862 	read_lock_bh(&idev->lock);
2863 	spin_lock(&ifp->lock);
2864 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
2865 		goto out;
2866 
2867 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2868 	    idev->cnf.accept_dad < 1 ||
2869 	    !(ifp->flags&IFA_F_TENTATIVE) ||
2870 	    ifp->flags & IFA_F_NODAD) {
2871 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2872 		spin_unlock(&ifp->lock);
2873 		read_unlock_bh(&idev->lock);
2874 
2875 		addrconf_dad_completed(ifp);
2876 		return;
2877 	}
2878 
2879 	if (!(idev->if_flags & IF_READY)) {
2880 		spin_unlock(&ifp->lock);
2881 		read_unlock_bh(&idev->lock);
2882 		/*
2883 		 * If the device is not ready:
2884 		 * - keep it tentative if it is a permanent address.
2885 		 * - otherwise, kill it.
2886 		 */
2887 		in6_ifa_hold(ifp);
2888 		addrconf_dad_stop(ifp, 0);
2889 		return;
2890 	}
2891 
2892 	/*
2893 	 * Optimistic nodes can start receiving
2894 	 * Frames right away
2895 	 */
2896 	if (ifp->flags & IFA_F_OPTIMISTIC)
2897 		ip6_ins_rt(ifp->rt);
2898 
2899 	addrconf_dad_kick(ifp);
2900 out:
2901 	spin_unlock(&ifp->lock);
2902 	read_unlock_bh(&idev->lock);
2903 }
2904 
2905 static void addrconf_dad_timer(unsigned long data)
2906 {
2907 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2908 	struct inet6_dev *idev = ifp->idev;
2909 	struct in6_addr mcaddr;
2910 
2911 	if (!ifp->probes && addrconf_dad_end(ifp))
2912 		goto out;
2913 
2914 	read_lock(&idev->lock);
2915 	if (idev->dead || !(idev->if_flags & IF_READY)) {
2916 		read_unlock(&idev->lock);
2917 		goto out;
2918 	}
2919 
2920 	spin_lock(&ifp->lock);
2921 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2922 		spin_unlock(&ifp->lock);
2923 		read_unlock(&idev->lock);
2924 		goto out;
2925 	}
2926 
2927 	if (ifp->probes == 0) {
2928 		/*
2929 		 * DAD was successful
2930 		 */
2931 
2932 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2933 		spin_unlock(&ifp->lock);
2934 		read_unlock(&idev->lock);
2935 
2936 		addrconf_dad_completed(ifp);
2937 
2938 		goto out;
2939 	}
2940 
2941 	ifp->probes--;
2942 	addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2943 	spin_unlock(&ifp->lock);
2944 	read_unlock(&idev->lock);
2945 
2946 	/* send a neighbour solicitation for our addr */
2947 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2948 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2949 out:
2950 	in6_ifa_put(ifp);
2951 }
2952 
2953 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2954 {
2955 	struct net_device *dev = ifp->idev->dev;
2956 
2957 	/*
2958 	 *	Configure the address for reception. Now it is valid.
2959 	 */
2960 
2961 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
2962 
2963 	/* If added prefix is link local and forwarding is off,
2964 	   start sending router solicitations.
2965 	 */
2966 
2967 	if (ifp->idev->cnf.forwarding == 0 &&
2968 	    ifp->idev->cnf.rtr_solicits > 0 &&
2969 	    (dev->flags&IFF_LOOPBACK) == 0 &&
2970 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2971 		/*
2972 		 *	If a host as already performed a random delay
2973 		 *	[...] as part of DAD [...] there is no need
2974 		 *	to delay again before sending the first RS
2975 		 */
2976 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2977 
2978 		spin_lock_bh(&ifp->lock);
2979 		ifp->probes = 1;
2980 		ifp->idev->if_flags |= IF_RS_SENT;
2981 		addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2982 		spin_unlock_bh(&ifp->lock);
2983 	}
2984 }
2985 
2986 static void addrconf_dad_run(struct inet6_dev *idev)
2987 {
2988 	struct inet6_ifaddr *ifp;
2989 
2990 	read_lock_bh(&idev->lock);
2991 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2992 		spin_lock(&ifp->lock);
2993 		if (ifp->flags & IFA_F_TENTATIVE &&
2994 		    ifp->state == INET6_IFADDR_STATE_DAD)
2995 			addrconf_dad_kick(ifp);
2996 		spin_unlock(&ifp->lock);
2997 	}
2998 	read_unlock_bh(&idev->lock);
2999 }
3000 
3001 #ifdef CONFIG_PROC_FS
3002 struct if6_iter_state {
3003 	struct seq_net_private p;
3004 	int bucket;
3005 };
3006 
3007 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3008 {
3009 	struct inet6_ifaddr *ifa = NULL;
3010 	struct if6_iter_state *state = seq->private;
3011 	struct net *net = seq_file_net(seq);
3012 
3013 	for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3014 		struct hlist_node *n;
3015 		hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3016 					 addr_lst)
3017 			if (net_eq(dev_net(ifa->idev->dev), net))
3018 				return ifa;
3019 	}
3020 	return NULL;
3021 }
3022 
3023 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3024 					 struct inet6_ifaddr *ifa)
3025 {
3026 	struct if6_iter_state *state = seq->private;
3027 	struct net *net = seq_file_net(seq);
3028 	struct hlist_node *n = &ifa->addr_lst;
3029 
3030 	hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3031 		if (net_eq(dev_net(ifa->idev->dev), net))
3032 			return ifa;
3033 
3034 	while (++state->bucket < IN6_ADDR_HSIZE) {
3035 		hlist_for_each_entry_rcu_bh(ifa, n,
3036 				     &inet6_addr_lst[state->bucket], addr_lst) {
3037 			if (net_eq(dev_net(ifa->idev->dev), net))
3038 				return ifa;
3039 		}
3040 	}
3041 
3042 	return NULL;
3043 }
3044 
3045 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3046 {
3047 	struct inet6_ifaddr *ifa = if6_get_first(seq);
3048 
3049 	if (ifa)
3050 		while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3051 			--pos;
3052 	return pos ? NULL : ifa;
3053 }
3054 
3055 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3056 	__acquires(rcu_bh)
3057 {
3058 	rcu_read_lock_bh();
3059 	return if6_get_idx(seq, *pos);
3060 }
3061 
3062 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3063 {
3064 	struct inet6_ifaddr *ifa;
3065 
3066 	ifa = if6_get_next(seq, v);
3067 	++*pos;
3068 	return ifa;
3069 }
3070 
3071 static void if6_seq_stop(struct seq_file *seq, void *v)
3072 	__releases(rcu_bh)
3073 {
3074 	rcu_read_unlock_bh();
3075 }
3076 
3077 static int if6_seq_show(struct seq_file *seq, void *v)
3078 {
3079 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3080 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3081 		   &ifp->addr,
3082 		   ifp->idev->dev->ifindex,
3083 		   ifp->prefix_len,
3084 		   ifp->scope,
3085 		   ifp->flags,
3086 		   ifp->idev->dev->name);
3087 	return 0;
3088 }
3089 
3090 static const struct seq_operations if6_seq_ops = {
3091 	.start	= if6_seq_start,
3092 	.next	= if6_seq_next,
3093 	.show	= if6_seq_show,
3094 	.stop	= if6_seq_stop,
3095 };
3096 
3097 static int if6_seq_open(struct inode *inode, struct file *file)
3098 {
3099 	return seq_open_net(inode, file, &if6_seq_ops,
3100 			    sizeof(struct if6_iter_state));
3101 }
3102 
3103 static const struct file_operations if6_fops = {
3104 	.owner		= THIS_MODULE,
3105 	.open		= if6_seq_open,
3106 	.read		= seq_read,
3107 	.llseek		= seq_lseek,
3108 	.release	= seq_release_net,
3109 };
3110 
3111 static int __net_init if6_proc_net_init(struct net *net)
3112 {
3113 	if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3114 		return -ENOMEM;
3115 	return 0;
3116 }
3117 
3118 static void __net_exit if6_proc_net_exit(struct net *net)
3119 {
3120        proc_net_remove(net, "if_inet6");
3121 }
3122 
3123 static struct pernet_operations if6_proc_net_ops = {
3124        .init = if6_proc_net_init,
3125        .exit = if6_proc_net_exit,
3126 };
3127 
3128 int __init if6_proc_init(void)
3129 {
3130 	return register_pernet_subsys(&if6_proc_net_ops);
3131 }
3132 
3133 void if6_proc_exit(void)
3134 {
3135 	unregister_pernet_subsys(&if6_proc_net_ops);
3136 }
3137 #endif	/* CONFIG_PROC_FS */
3138 
3139 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3140 /* Check if address is a home address configured on any interface. */
3141 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3142 {
3143 	int ret = 0;
3144 	struct inet6_ifaddr *ifp = NULL;
3145 	struct hlist_node *n;
3146 	unsigned int hash = ipv6_addr_hash(addr);
3147 
3148 	rcu_read_lock_bh();
3149 	hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3150 		if (!net_eq(dev_net(ifp->idev->dev), net))
3151 			continue;
3152 		if (ipv6_addr_equal(&ifp->addr, addr) &&
3153 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3154 			ret = 1;
3155 			break;
3156 		}
3157 	}
3158 	rcu_read_unlock_bh();
3159 	return ret;
3160 }
3161 #endif
3162 
3163 /*
3164  *	Periodic address status verification
3165  */
3166 
3167 static void addrconf_verify(unsigned long foo)
3168 {
3169 	unsigned long now, next, next_sec, next_sched;
3170 	struct inet6_ifaddr *ifp;
3171 	struct hlist_node *node;
3172 	int i;
3173 
3174 	rcu_read_lock_bh();
3175 	spin_lock(&addrconf_verify_lock);
3176 	now = jiffies;
3177 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3178 
3179 	del_timer(&addr_chk_timer);
3180 
3181 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3182 restart:
3183 		hlist_for_each_entry_rcu_bh(ifp, node,
3184 					 &inet6_addr_lst[i], addr_lst) {
3185 			unsigned long age;
3186 
3187 			if (ifp->flags & IFA_F_PERMANENT)
3188 				continue;
3189 
3190 			spin_lock(&ifp->lock);
3191 			/* We try to batch several events at once. */
3192 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3193 
3194 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3195 			    age >= ifp->valid_lft) {
3196 				spin_unlock(&ifp->lock);
3197 				in6_ifa_hold(ifp);
3198 				ipv6_del_addr(ifp);
3199 				goto restart;
3200 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3201 				spin_unlock(&ifp->lock);
3202 				continue;
3203 			} else if (age >= ifp->prefered_lft) {
3204 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3205 				int deprecate = 0;
3206 
3207 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3208 					deprecate = 1;
3209 					ifp->flags |= IFA_F_DEPRECATED;
3210 				}
3211 
3212 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3213 					next = ifp->tstamp + ifp->valid_lft * HZ;
3214 
3215 				spin_unlock(&ifp->lock);
3216 
3217 				if (deprecate) {
3218 					in6_ifa_hold(ifp);
3219 
3220 					ipv6_ifa_notify(0, ifp);
3221 					in6_ifa_put(ifp);
3222 					goto restart;
3223 				}
3224 #ifdef CONFIG_IPV6_PRIVACY
3225 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3226 				   !(ifp->flags&IFA_F_TENTATIVE)) {
3227 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3228 					ifp->idev->cnf.dad_transmits *
3229 					ifp->idev->nd_parms->retrans_time / HZ;
3230 
3231 				if (age >= ifp->prefered_lft - regen_advance) {
3232 					struct inet6_ifaddr *ifpub = ifp->ifpub;
3233 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3234 						next = ifp->tstamp + ifp->prefered_lft * HZ;
3235 					if (!ifp->regen_count && ifpub) {
3236 						ifp->regen_count++;
3237 						in6_ifa_hold(ifp);
3238 						in6_ifa_hold(ifpub);
3239 						spin_unlock(&ifp->lock);
3240 
3241 						spin_lock(&ifpub->lock);
3242 						ifpub->regen_count = 0;
3243 						spin_unlock(&ifpub->lock);
3244 						ipv6_create_tempaddr(ifpub, ifp);
3245 						in6_ifa_put(ifpub);
3246 						in6_ifa_put(ifp);
3247 						goto restart;
3248 					}
3249 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3250 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3251 				spin_unlock(&ifp->lock);
3252 #endif
3253 			} else {
3254 				/* ifp->prefered_lft <= ifp->valid_lft */
3255 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3256 					next = ifp->tstamp + ifp->prefered_lft * HZ;
3257 				spin_unlock(&ifp->lock);
3258 			}
3259 		}
3260 	}
3261 
3262 	next_sec = round_jiffies_up(next);
3263 	next_sched = next;
3264 
3265 	/* If rounded timeout is accurate enough, accept it. */
3266 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3267 		next_sched = next_sec;
3268 
3269 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3270 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3271 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3272 
3273 	ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3274 	      now, next, next_sec, next_sched));
3275 
3276 	addr_chk_timer.expires = next_sched;
3277 	add_timer(&addr_chk_timer);
3278 	spin_unlock(&addrconf_verify_lock);
3279 	rcu_read_unlock_bh();
3280 }
3281 
3282 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3283 {
3284 	struct in6_addr *pfx = NULL;
3285 
3286 	if (addr)
3287 		pfx = nla_data(addr);
3288 
3289 	if (local) {
3290 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3291 			pfx = NULL;
3292 		else
3293 			pfx = nla_data(local);
3294 	}
3295 
3296 	return pfx;
3297 }
3298 
3299 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3300 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
3301 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
3302 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
3303 };
3304 
3305 static int
3306 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3307 {
3308 	struct net *net = sock_net(skb->sk);
3309 	struct ifaddrmsg *ifm;
3310 	struct nlattr *tb[IFA_MAX+1];
3311 	struct in6_addr *pfx;
3312 	int err;
3313 
3314 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3315 	if (err < 0)
3316 		return err;
3317 
3318 	ifm = nlmsg_data(nlh);
3319 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3320 	if (pfx == NULL)
3321 		return -EINVAL;
3322 
3323 	return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3324 }
3325 
3326 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3327 			     u32 prefered_lft, u32 valid_lft)
3328 {
3329 	u32 flags;
3330 	clock_t expires;
3331 	unsigned long timeout;
3332 
3333 	if (!valid_lft || (prefered_lft > valid_lft))
3334 		return -EINVAL;
3335 
3336 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
3337 	if (addrconf_finite_timeout(timeout)) {
3338 		expires = jiffies_to_clock_t(timeout * HZ);
3339 		valid_lft = timeout;
3340 		flags = RTF_EXPIRES;
3341 	} else {
3342 		expires = 0;
3343 		flags = 0;
3344 		ifa_flags |= IFA_F_PERMANENT;
3345 	}
3346 
3347 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3348 	if (addrconf_finite_timeout(timeout)) {
3349 		if (timeout == 0)
3350 			ifa_flags |= IFA_F_DEPRECATED;
3351 		prefered_lft = timeout;
3352 	}
3353 
3354 	spin_lock_bh(&ifp->lock);
3355 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3356 	ifp->tstamp = jiffies;
3357 	ifp->valid_lft = valid_lft;
3358 	ifp->prefered_lft = prefered_lft;
3359 
3360 	spin_unlock_bh(&ifp->lock);
3361 	if (!(ifp->flags&IFA_F_TENTATIVE))
3362 		ipv6_ifa_notify(0, ifp);
3363 
3364 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3365 			      expires, flags);
3366 	addrconf_verify(0);
3367 
3368 	return 0;
3369 }
3370 
3371 static int
3372 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3373 {
3374 	struct net *net = sock_net(skb->sk);
3375 	struct ifaddrmsg *ifm;
3376 	struct nlattr *tb[IFA_MAX+1];
3377 	struct in6_addr *pfx;
3378 	struct inet6_ifaddr *ifa;
3379 	struct net_device *dev;
3380 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3381 	u8 ifa_flags;
3382 	int err;
3383 
3384 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3385 	if (err < 0)
3386 		return err;
3387 
3388 	ifm = nlmsg_data(nlh);
3389 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3390 	if (pfx == NULL)
3391 		return -EINVAL;
3392 
3393 	if (tb[IFA_CACHEINFO]) {
3394 		struct ifa_cacheinfo *ci;
3395 
3396 		ci = nla_data(tb[IFA_CACHEINFO]);
3397 		valid_lft = ci->ifa_valid;
3398 		preferred_lft = ci->ifa_prefered;
3399 	} else {
3400 		preferred_lft = INFINITY_LIFE_TIME;
3401 		valid_lft = INFINITY_LIFE_TIME;
3402 	}
3403 
3404 	dev =  __dev_get_by_index(net, ifm->ifa_index);
3405 	if (dev == NULL)
3406 		return -ENODEV;
3407 
3408 	/* We ignore other flags so far. */
3409 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3410 
3411 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3412 	if (ifa == NULL) {
3413 		/*
3414 		 * It would be best to check for !NLM_F_CREATE here but
3415 		 * userspace alreay relies on not having to provide this.
3416 		 */
3417 		return inet6_addr_add(net, ifm->ifa_index, pfx,
3418 				      ifm->ifa_prefixlen, ifa_flags,
3419 				      preferred_lft, valid_lft);
3420 	}
3421 
3422 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3423 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3424 		err = -EEXIST;
3425 	else
3426 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3427 
3428 	in6_ifa_put(ifa);
3429 
3430 	return err;
3431 }
3432 
3433 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3434 			  u8 scope, int ifindex)
3435 {
3436 	struct ifaddrmsg *ifm;
3437 
3438 	ifm = nlmsg_data(nlh);
3439 	ifm->ifa_family = AF_INET6;
3440 	ifm->ifa_prefixlen = prefixlen;
3441 	ifm->ifa_flags = flags;
3442 	ifm->ifa_scope = scope;
3443 	ifm->ifa_index = ifindex;
3444 }
3445 
3446 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3447 			 unsigned long tstamp, u32 preferred, u32 valid)
3448 {
3449 	struct ifa_cacheinfo ci;
3450 
3451 	ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3452 			+ TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3453 	ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3454 			+ TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3455 	ci.ifa_prefered = preferred;
3456 	ci.ifa_valid = valid;
3457 
3458 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3459 }
3460 
3461 static inline int rt_scope(int ifa_scope)
3462 {
3463 	if (ifa_scope & IFA_HOST)
3464 		return RT_SCOPE_HOST;
3465 	else if (ifa_scope & IFA_LINK)
3466 		return RT_SCOPE_LINK;
3467 	else if (ifa_scope & IFA_SITE)
3468 		return RT_SCOPE_SITE;
3469 	else
3470 		return RT_SCOPE_UNIVERSE;
3471 }
3472 
3473 static inline int inet6_ifaddr_msgsize(void)
3474 {
3475 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3476 	       + nla_total_size(16) /* IFA_ADDRESS */
3477 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3478 }
3479 
3480 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3481 			     u32 pid, u32 seq, int event, unsigned int flags)
3482 {
3483 	struct nlmsghdr  *nlh;
3484 	u32 preferred, valid;
3485 
3486 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3487 	if (nlh == NULL)
3488 		return -EMSGSIZE;
3489 
3490 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3491 		      ifa->idev->dev->ifindex);
3492 
3493 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3494 		preferred = ifa->prefered_lft;
3495 		valid = ifa->valid_lft;
3496 		if (preferred != INFINITY_LIFE_TIME) {
3497 			long tval = (jiffies - ifa->tstamp)/HZ;
3498 			if (preferred > tval)
3499 				preferred -= tval;
3500 			else
3501 				preferred = 0;
3502 			if (valid != INFINITY_LIFE_TIME) {
3503 				if (valid > tval)
3504 					valid -= tval;
3505 				else
3506 					valid = 0;
3507 			}
3508 		}
3509 	} else {
3510 		preferred = INFINITY_LIFE_TIME;
3511 		valid = INFINITY_LIFE_TIME;
3512 	}
3513 
3514 	if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3515 	    put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3516 		nlmsg_cancel(skb, nlh);
3517 		return -EMSGSIZE;
3518 	}
3519 
3520 	return nlmsg_end(skb, nlh);
3521 }
3522 
3523 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3524 				u32 pid, u32 seq, int event, u16 flags)
3525 {
3526 	struct nlmsghdr  *nlh;
3527 	u8 scope = RT_SCOPE_UNIVERSE;
3528 	int ifindex = ifmca->idev->dev->ifindex;
3529 
3530 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3531 		scope = RT_SCOPE_SITE;
3532 
3533 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3534 	if (nlh == NULL)
3535 		return -EMSGSIZE;
3536 
3537 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3538 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3539 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3540 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3541 		nlmsg_cancel(skb, nlh);
3542 		return -EMSGSIZE;
3543 	}
3544 
3545 	return nlmsg_end(skb, nlh);
3546 }
3547 
3548 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3549 				u32 pid, u32 seq, int event, unsigned int flags)
3550 {
3551 	struct nlmsghdr  *nlh;
3552 	u8 scope = RT_SCOPE_UNIVERSE;
3553 	int ifindex = ifaca->aca_idev->dev->ifindex;
3554 
3555 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3556 		scope = RT_SCOPE_SITE;
3557 
3558 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3559 	if (nlh == NULL)
3560 		return -EMSGSIZE;
3561 
3562 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3563 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3564 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3565 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3566 		nlmsg_cancel(skb, nlh);
3567 		return -EMSGSIZE;
3568 	}
3569 
3570 	return nlmsg_end(skb, nlh);
3571 }
3572 
3573 enum addr_type_t {
3574 	UNICAST_ADDR,
3575 	MULTICAST_ADDR,
3576 	ANYCAST_ADDR,
3577 };
3578 
3579 /* called with rcu_read_lock() */
3580 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3581 			  struct netlink_callback *cb, enum addr_type_t type,
3582 			  int s_ip_idx, int *p_ip_idx)
3583 {
3584 	struct ifmcaddr6 *ifmca;
3585 	struct ifacaddr6 *ifaca;
3586 	int err = 1;
3587 	int ip_idx = *p_ip_idx;
3588 
3589 	read_lock_bh(&idev->lock);
3590 	switch (type) {
3591 	case UNICAST_ADDR: {
3592 		struct inet6_ifaddr *ifa;
3593 
3594 		/* unicast address incl. temp addr */
3595 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
3596 			if (++ip_idx < s_ip_idx)
3597 				continue;
3598 			err = inet6_fill_ifaddr(skb, ifa,
3599 						NETLINK_CB(cb->skb).pid,
3600 						cb->nlh->nlmsg_seq,
3601 						RTM_NEWADDR,
3602 						NLM_F_MULTI);
3603 			if (err <= 0)
3604 				break;
3605 		}
3606 		break;
3607 	}
3608 	case MULTICAST_ADDR:
3609 		/* multicast address */
3610 		for (ifmca = idev->mc_list; ifmca;
3611 		     ifmca = ifmca->next, ip_idx++) {
3612 			if (ip_idx < s_ip_idx)
3613 				continue;
3614 			err = inet6_fill_ifmcaddr(skb, ifmca,
3615 						  NETLINK_CB(cb->skb).pid,
3616 						  cb->nlh->nlmsg_seq,
3617 						  RTM_GETMULTICAST,
3618 						  NLM_F_MULTI);
3619 			if (err <= 0)
3620 				break;
3621 		}
3622 		break;
3623 	case ANYCAST_ADDR:
3624 		/* anycast address */
3625 		for (ifaca = idev->ac_list; ifaca;
3626 		     ifaca = ifaca->aca_next, ip_idx++) {
3627 			if (ip_idx < s_ip_idx)
3628 				continue;
3629 			err = inet6_fill_ifacaddr(skb, ifaca,
3630 						  NETLINK_CB(cb->skb).pid,
3631 						  cb->nlh->nlmsg_seq,
3632 						  RTM_GETANYCAST,
3633 						  NLM_F_MULTI);
3634 			if (err <= 0)
3635 				break;
3636 		}
3637 		break;
3638 	default:
3639 		break;
3640 	}
3641 	read_unlock_bh(&idev->lock);
3642 	*p_ip_idx = ip_idx;
3643 	return err;
3644 }
3645 
3646 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3647 			   enum addr_type_t type)
3648 {
3649 	struct net *net = sock_net(skb->sk);
3650 	int h, s_h;
3651 	int idx, ip_idx;
3652 	int s_idx, s_ip_idx;
3653 	struct net_device *dev;
3654 	struct inet6_dev *idev;
3655 	struct hlist_head *head;
3656 	struct hlist_node *node;
3657 
3658 	s_h = cb->args[0];
3659 	s_idx = idx = cb->args[1];
3660 	s_ip_idx = ip_idx = cb->args[2];
3661 
3662 	rcu_read_lock();
3663 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3664 		idx = 0;
3665 		head = &net->dev_index_head[h];
3666 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3667 			if (idx < s_idx)
3668 				goto cont;
3669 			if (h > s_h || idx > s_idx)
3670 				s_ip_idx = 0;
3671 			ip_idx = 0;
3672 			idev = __in6_dev_get(dev);
3673 			if (!idev)
3674 				goto cont;
3675 
3676 			if (in6_dump_addrs(idev, skb, cb, type,
3677 					   s_ip_idx, &ip_idx) <= 0)
3678 				goto done;
3679 cont:
3680 			idx++;
3681 		}
3682 	}
3683 done:
3684 	rcu_read_unlock();
3685 	cb->args[0] = h;
3686 	cb->args[1] = idx;
3687 	cb->args[2] = ip_idx;
3688 
3689 	return skb->len;
3690 }
3691 
3692 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3693 {
3694 	enum addr_type_t type = UNICAST_ADDR;
3695 
3696 	return inet6_dump_addr(skb, cb, type);
3697 }
3698 
3699 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3700 {
3701 	enum addr_type_t type = MULTICAST_ADDR;
3702 
3703 	return inet6_dump_addr(skb, cb, type);
3704 }
3705 
3706 
3707 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3708 {
3709 	enum addr_type_t type = ANYCAST_ADDR;
3710 
3711 	return inet6_dump_addr(skb, cb, type);
3712 }
3713 
3714 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3715 			     void *arg)
3716 {
3717 	struct net *net = sock_net(in_skb->sk);
3718 	struct ifaddrmsg *ifm;
3719 	struct nlattr *tb[IFA_MAX+1];
3720 	struct in6_addr *addr = NULL;
3721 	struct net_device *dev = NULL;
3722 	struct inet6_ifaddr *ifa;
3723 	struct sk_buff *skb;
3724 	int err;
3725 
3726 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3727 	if (err < 0)
3728 		goto errout;
3729 
3730 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3731 	if (addr == NULL) {
3732 		err = -EINVAL;
3733 		goto errout;
3734 	}
3735 
3736 	ifm = nlmsg_data(nlh);
3737 	if (ifm->ifa_index)
3738 		dev = __dev_get_by_index(net, ifm->ifa_index);
3739 
3740 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3741 	if (!ifa) {
3742 		err = -EADDRNOTAVAIL;
3743 		goto errout;
3744 	}
3745 
3746 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3747 	if (!skb) {
3748 		err = -ENOBUFS;
3749 		goto errout_ifa;
3750 	}
3751 
3752 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3753 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
3754 	if (err < 0) {
3755 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3756 		WARN_ON(err == -EMSGSIZE);
3757 		kfree_skb(skb);
3758 		goto errout_ifa;
3759 	}
3760 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3761 errout_ifa:
3762 	in6_ifa_put(ifa);
3763 errout:
3764 	return err;
3765 }
3766 
3767 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3768 {
3769 	struct sk_buff *skb;
3770 	struct net *net = dev_net(ifa->idev->dev);
3771 	int err = -ENOBUFS;
3772 
3773 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3774 	if (skb == NULL)
3775 		goto errout;
3776 
3777 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3778 	if (err < 0) {
3779 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3780 		WARN_ON(err == -EMSGSIZE);
3781 		kfree_skb(skb);
3782 		goto errout;
3783 	}
3784 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3785 	return;
3786 errout:
3787 	if (err < 0)
3788 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3789 }
3790 
3791 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3792 				__s32 *array, int bytes)
3793 {
3794 	BUG_ON(bytes < (DEVCONF_MAX * 4));
3795 
3796 	memset(array, 0, bytes);
3797 	array[DEVCONF_FORWARDING] = cnf->forwarding;
3798 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3799 	array[DEVCONF_MTU6] = cnf->mtu6;
3800 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3801 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3802 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
3803 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3804 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3805 	array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3806 	array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3807 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3808 #ifdef CONFIG_IPV6_PRIVACY
3809 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3810 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3811 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3812 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3813 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3814 #endif
3815 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3816 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3817 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3818 #ifdef CONFIG_IPV6_ROUTER_PREF
3819 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3820 	array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3821 #ifdef CONFIG_IPV6_ROUTE_INFO
3822 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3823 #endif
3824 #endif
3825 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3826 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3827 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3828 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3829 #endif
3830 #ifdef CONFIG_IPV6_MROUTE
3831 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3832 #endif
3833 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3834 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3835 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3836 }
3837 
3838 static inline size_t inet6_if_nlmsg_size(void)
3839 {
3840 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3841 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3842 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3843 	       + nla_total_size(4) /* IFLA_MTU */
3844 	       + nla_total_size(4) /* IFLA_LINK */
3845 	       + nla_total_size( /* IFLA_PROTINFO */
3846 			nla_total_size(4) /* IFLA_INET6_FLAGS */
3847 			+ nla_total_size(sizeof(struct ifla_cacheinfo))
3848 			+ nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3849 			+ nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3850 			+ nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3851 		 );
3852 }
3853 
3854 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3855 				      int items, int bytes)
3856 {
3857 	int i;
3858 	int pad = bytes - sizeof(u64) * items;
3859 	BUG_ON(pad < 0);
3860 
3861 	/* Use put_unaligned() because stats may not be aligned for u64. */
3862 	put_unaligned(items, &stats[0]);
3863 	for (i = 1; i < items; i++)
3864 		put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3865 
3866 	memset(&stats[items], 0, pad);
3867 }
3868 
3869 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3870 				      int items, int bytes, size_t syncpoff)
3871 {
3872 	int i;
3873 	int pad = bytes - sizeof(u64) * items;
3874 	BUG_ON(pad < 0);
3875 
3876 	/* Use put_unaligned() because stats may not be aligned for u64. */
3877 	put_unaligned(items, &stats[0]);
3878 	for (i = 1; i < items; i++)
3879 		put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3880 
3881 	memset(&stats[items], 0, pad);
3882 }
3883 
3884 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3885 			     int bytes)
3886 {
3887 	switch (attrtype) {
3888 	case IFLA_INET6_STATS:
3889 		__snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3890 				     IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3891 		break;
3892 	case IFLA_INET6_ICMP6STATS:
3893 		__snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3894 		break;
3895 	}
3896 }
3897 
3898 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3899 			     u32 pid, u32 seq, int event, unsigned int flags)
3900 {
3901 	struct net_device *dev = idev->dev;
3902 	struct nlattr *nla;
3903 	struct ifinfomsg *hdr;
3904 	struct nlmsghdr *nlh;
3905 	void *protoinfo;
3906 	struct ifla_cacheinfo ci;
3907 
3908 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3909 	if (nlh == NULL)
3910 		return -EMSGSIZE;
3911 
3912 	hdr = nlmsg_data(nlh);
3913 	hdr->ifi_family = AF_INET6;
3914 	hdr->__ifi_pad = 0;
3915 	hdr->ifi_type = dev->type;
3916 	hdr->ifi_index = dev->ifindex;
3917 	hdr->ifi_flags = dev_get_flags(dev);
3918 	hdr->ifi_change = 0;
3919 
3920 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3921 
3922 	if (dev->addr_len)
3923 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3924 
3925 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3926 	if (dev->ifindex != dev->iflink)
3927 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3928 
3929 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3930 	if (protoinfo == NULL)
3931 		goto nla_put_failure;
3932 
3933 	NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3934 
3935 	ci.max_reasm_len = IPV6_MAXPLEN;
3936 	ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3937 		    + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3938 	ci.reachable_time = idev->nd_parms->reachable_time;
3939 	ci.retrans_time = idev->nd_parms->retrans_time;
3940 	NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3941 
3942 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3943 	if (nla == NULL)
3944 		goto nla_put_failure;
3945 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3946 
3947 	/* XXX - MC not implemented */
3948 
3949 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3950 	if (nla == NULL)
3951 		goto nla_put_failure;
3952 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3953 
3954 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3955 	if (nla == NULL)
3956 		goto nla_put_failure;
3957 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3958 
3959 	nla_nest_end(skb, protoinfo);
3960 	return nlmsg_end(skb, nlh);
3961 
3962 nla_put_failure:
3963 	nlmsg_cancel(skb, nlh);
3964 	return -EMSGSIZE;
3965 }
3966 
3967 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3968 {
3969 	struct net *net = sock_net(skb->sk);
3970 	int h, s_h;
3971 	int idx = 0, s_idx;
3972 	struct net_device *dev;
3973 	struct inet6_dev *idev;
3974 	struct hlist_head *head;
3975 	struct hlist_node *node;
3976 
3977 	s_h = cb->args[0];
3978 	s_idx = cb->args[1];
3979 
3980 	rcu_read_lock();
3981 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3982 		idx = 0;
3983 		head = &net->dev_index_head[h];
3984 		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3985 			if (idx < s_idx)
3986 				goto cont;
3987 			idev = __in6_dev_get(dev);
3988 			if (!idev)
3989 				goto cont;
3990 			if (inet6_fill_ifinfo(skb, idev,
3991 					      NETLINK_CB(cb->skb).pid,
3992 					      cb->nlh->nlmsg_seq,
3993 					      RTM_NEWLINK, NLM_F_MULTI) <= 0)
3994 				goto out;
3995 cont:
3996 			idx++;
3997 		}
3998 	}
3999 out:
4000 	rcu_read_unlock();
4001 	cb->args[1] = idx;
4002 	cb->args[0] = h;
4003 
4004 	return skb->len;
4005 }
4006 
4007 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4008 {
4009 	struct sk_buff *skb;
4010 	struct net *net = dev_net(idev->dev);
4011 	int err = -ENOBUFS;
4012 
4013 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4014 	if (skb == NULL)
4015 		goto errout;
4016 
4017 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4018 	if (err < 0) {
4019 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4020 		WARN_ON(err == -EMSGSIZE);
4021 		kfree_skb(skb);
4022 		goto errout;
4023 	}
4024 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4025 	return;
4026 errout:
4027 	if (err < 0)
4028 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4029 }
4030 
4031 static inline size_t inet6_prefix_nlmsg_size(void)
4032 {
4033 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
4034 	       + nla_total_size(sizeof(struct in6_addr))
4035 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
4036 }
4037 
4038 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4039 			     struct prefix_info *pinfo, u32 pid, u32 seq,
4040 			     int event, unsigned int flags)
4041 {
4042 	struct prefixmsg *pmsg;
4043 	struct nlmsghdr *nlh;
4044 	struct prefix_cacheinfo	ci;
4045 
4046 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4047 	if (nlh == NULL)
4048 		return -EMSGSIZE;
4049 
4050 	pmsg = nlmsg_data(nlh);
4051 	pmsg->prefix_family = AF_INET6;
4052 	pmsg->prefix_pad1 = 0;
4053 	pmsg->prefix_pad2 = 0;
4054 	pmsg->prefix_ifindex = idev->dev->ifindex;
4055 	pmsg->prefix_len = pinfo->prefix_len;
4056 	pmsg->prefix_type = pinfo->type;
4057 	pmsg->prefix_pad3 = 0;
4058 	pmsg->prefix_flags = 0;
4059 	if (pinfo->onlink)
4060 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4061 	if (pinfo->autoconf)
4062 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4063 
4064 	NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4065 
4066 	ci.preferred_time = ntohl(pinfo->prefered);
4067 	ci.valid_time = ntohl(pinfo->valid);
4068 	NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4069 
4070 	return nlmsg_end(skb, nlh);
4071 
4072 nla_put_failure:
4073 	nlmsg_cancel(skb, nlh);
4074 	return -EMSGSIZE;
4075 }
4076 
4077 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4078 			 struct prefix_info *pinfo)
4079 {
4080 	struct sk_buff *skb;
4081 	struct net *net = dev_net(idev->dev);
4082 	int err = -ENOBUFS;
4083 
4084 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4085 	if (skb == NULL)
4086 		goto errout;
4087 
4088 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4089 	if (err < 0) {
4090 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4091 		WARN_ON(err == -EMSGSIZE);
4092 		kfree_skb(skb);
4093 		goto errout;
4094 	}
4095 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4096 	return;
4097 errout:
4098 	if (err < 0)
4099 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4100 }
4101 
4102 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4103 {
4104 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4105 
4106 	switch (event) {
4107 	case RTM_NEWADDR:
4108 		/*
4109 		 * If the address was optimistic
4110 		 * we inserted the route at the start of
4111 		 * our DAD process, so we don't need
4112 		 * to do it again
4113 		 */
4114 		if (!(ifp->rt->rt6i_node))
4115 			ip6_ins_rt(ifp->rt);
4116 		if (ifp->idev->cnf.forwarding)
4117 			addrconf_join_anycast(ifp);
4118 		break;
4119 	case RTM_DELADDR:
4120 		if (ifp->idev->cnf.forwarding)
4121 			addrconf_leave_anycast(ifp);
4122 		addrconf_leave_solict(ifp->idev, &ifp->addr);
4123 		dst_hold(&ifp->rt->dst);
4124 
4125 		if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4126 		    ip6_del_rt(ifp->rt))
4127 			dst_free(&ifp->rt->dst);
4128 		break;
4129 	}
4130 }
4131 
4132 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4133 {
4134 	rcu_read_lock_bh();
4135 	if (likely(ifp->idev->dead == 0))
4136 		__ipv6_ifa_notify(event, ifp);
4137 	rcu_read_unlock_bh();
4138 }
4139 
4140 #ifdef CONFIG_SYSCTL
4141 
4142 static
4143 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4144 			   void __user *buffer, size_t *lenp, loff_t *ppos)
4145 {
4146 	int *valp = ctl->data;
4147 	int val = *valp;
4148 	loff_t pos = *ppos;
4149 	int ret;
4150 
4151 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4152 
4153 	if (write)
4154 		ret = addrconf_fixup_forwarding(ctl, valp, val);
4155 	if (ret)
4156 		*ppos = pos;
4157 	return ret;
4158 }
4159 
4160 static void dev_disable_change(struct inet6_dev *idev)
4161 {
4162 	if (!idev || !idev->dev)
4163 		return;
4164 
4165 	if (idev->cnf.disable_ipv6)
4166 		addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4167 	else
4168 		addrconf_notify(NULL, NETDEV_UP, idev->dev);
4169 }
4170 
4171 static void addrconf_disable_change(struct net *net, __s32 newf)
4172 {
4173 	struct net_device *dev;
4174 	struct inet6_dev *idev;
4175 
4176 	rcu_read_lock();
4177 	for_each_netdev_rcu(net, dev) {
4178 		idev = __in6_dev_get(dev);
4179 		if (idev) {
4180 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4181 			idev->cnf.disable_ipv6 = newf;
4182 			if (changed)
4183 				dev_disable_change(idev);
4184 		}
4185 	}
4186 	rcu_read_unlock();
4187 }
4188 
4189 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4190 {
4191 	struct net *net;
4192 
4193 	net = (struct net *)table->extra2;
4194 
4195 	if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4196 		return 0;
4197 
4198 	if (!rtnl_trylock()) {
4199 		/* Restore the original values before restarting */
4200 		*p = old;
4201 		return restart_syscall();
4202 	}
4203 
4204 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
4205 		__s32 newf = net->ipv6.devconf_all->disable_ipv6;
4206 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
4207 		addrconf_disable_change(net, newf);
4208 	} else if ((!*p) ^ (!old))
4209 		dev_disable_change((struct inet6_dev *)table->extra1);
4210 
4211 	rtnl_unlock();
4212 	return 0;
4213 }
4214 
4215 static
4216 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4217 			    void __user *buffer, size_t *lenp, loff_t *ppos)
4218 {
4219 	int *valp = ctl->data;
4220 	int val = *valp;
4221 	loff_t pos = *ppos;
4222 	int ret;
4223 
4224 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4225 
4226 	if (write)
4227 		ret = addrconf_disable_ipv6(ctl, valp, val);
4228 	if (ret)
4229 		*ppos = pos;
4230 	return ret;
4231 }
4232 
4233 static struct addrconf_sysctl_table
4234 {
4235 	struct ctl_table_header *sysctl_header;
4236 	ctl_table addrconf_vars[DEVCONF_MAX+1];
4237 	char *dev_name;
4238 } addrconf_sysctl __read_mostly = {
4239 	.sysctl_header = NULL,
4240 	.addrconf_vars = {
4241 		{
4242 			.procname	= "forwarding",
4243 			.data		= &ipv6_devconf.forwarding,
4244 			.maxlen		= sizeof(int),
4245 			.mode		= 0644,
4246 			.proc_handler	= addrconf_sysctl_forward,
4247 		},
4248 		{
4249 			.procname	= "hop_limit",
4250 			.data		= &ipv6_devconf.hop_limit,
4251 			.maxlen		= sizeof(int),
4252 			.mode		= 0644,
4253 			.proc_handler	= proc_dointvec,
4254 		},
4255 		{
4256 			.procname	= "mtu",
4257 			.data		= &ipv6_devconf.mtu6,
4258 			.maxlen		= sizeof(int),
4259 			.mode		= 0644,
4260 			.proc_handler	= proc_dointvec,
4261 		},
4262 		{
4263 			.procname	= "accept_ra",
4264 			.data		= &ipv6_devconf.accept_ra,
4265 			.maxlen		= sizeof(int),
4266 			.mode		= 0644,
4267 			.proc_handler	= proc_dointvec,
4268 		},
4269 		{
4270 			.procname	= "accept_redirects",
4271 			.data		= &ipv6_devconf.accept_redirects,
4272 			.maxlen		= sizeof(int),
4273 			.mode		= 0644,
4274 			.proc_handler	= proc_dointvec,
4275 		},
4276 		{
4277 			.procname	= "autoconf",
4278 			.data		= &ipv6_devconf.autoconf,
4279 			.maxlen		= sizeof(int),
4280 			.mode		= 0644,
4281 			.proc_handler	= proc_dointvec,
4282 		},
4283 		{
4284 			.procname	= "dad_transmits",
4285 			.data		= &ipv6_devconf.dad_transmits,
4286 			.maxlen		= sizeof(int),
4287 			.mode		= 0644,
4288 			.proc_handler	= proc_dointvec,
4289 		},
4290 		{
4291 			.procname	= "router_solicitations",
4292 			.data		= &ipv6_devconf.rtr_solicits,
4293 			.maxlen		= sizeof(int),
4294 			.mode		= 0644,
4295 			.proc_handler	= proc_dointvec,
4296 		},
4297 		{
4298 			.procname	= "router_solicitation_interval",
4299 			.data		= &ipv6_devconf.rtr_solicit_interval,
4300 			.maxlen		= sizeof(int),
4301 			.mode		= 0644,
4302 			.proc_handler	= proc_dointvec_jiffies,
4303 		},
4304 		{
4305 			.procname	= "router_solicitation_delay",
4306 			.data		= &ipv6_devconf.rtr_solicit_delay,
4307 			.maxlen		= sizeof(int),
4308 			.mode		= 0644,
4309 			.proc_handler	= proc_dointvec_jiffies,
4310 		},
4311 		{
4312 			.procname	= "force_mld_version",
4313 			.data		= &ipv6_devconf.force_mld_version,
4314 			.maxlen		= sizeof(int),
4315 			.mode		= 0644,
4316 			.proc_handler	= proc_dointvec,
4317 		},
4318 #ifdef CONFIG_IPV6_PRIVACY
4319 		{
4320 			.procname	= "use_tempaddr",
4321 			.data		= &ipv6_devconf.use_tempaddr,
4322 			.maxlen		= sizeof(int),
4323 			.mode		= 0644,
4324 			.proc_handler	= proc_dointvec,
4325 		},
4326 		{
4327 			.procname	= "temp_valid_lft",
4328 			.data		= &ipv6_devconf.temp_valid_lft,
4329 			.maxlen		= sizeof(int),
4330 			.mode		= 0644,
4331 			.proc_handler	= proc_dointvec,
4332 		},
4333 		{
4334 			.procname	= "temp_prefered_lft",
4335 			.data		= &ipv6_devconf.temp_prefered_lft,
4336 			.maxlen		= sizeof(int),
4337 			.mode		= 0644,
4338 			.proc_handler	= proc_dointvec,
4339 		},
4340 		{
4341 			.procname	= "regen_max_retry",
4342 			.data		= &ipv6_devconf.regen_max_retry,
4343 			.maxlen		= sizeof(int),
4344 			.mode		= 0644,
4345 			.proc_handler	= proc_dointvec,
4346 		},
4347 		{
4348 			.procname	= "max_desync_factor",
4349 			.data		= &ipv6_devconf.max_desync_factor,
4350 			.maxlen		= sizeof(int),
4351 			.mode		= 0644,
4352 			.proc_handler	= proc_dointvec,
4353 		},
4354 #endif
4355 		{
4356 			.procname	= "max_addresses",
4357 			.data		= &ipv6_devconf.max_addresses,
4358 			.maxlen		= sizeof(int),
4359 			.mode		= 0644,
4360 			.proc_handler	= proc_dointvec,
4361 		},
4362 		{
4363 			.procname	= "accept_ra_defrtr",
4364 			.data		= &ipv6_devconf.accept_ra_defrtr,
4365 			.maxlen		= sizeof(int),
4366 			.mode		= 0644,
4367 			.proc_handler	= proc_dointvec,
4368 		},
4369 		{
4370 			.procname	= "accept_ra_pinfo",
4371 			.data		= &ipv6_devconf.accept_ra_pinfo,
4372 			.maxlen		= sizeof(int),
4373 			.mode		= 0644,
4374 			.proc_handler	= proc_dointvec,
4375 		},
4376 #ifdef CONFIG_IPV6_ROUTER_PREF
4377 		{
4378 			.procname	= "accept_ra_rtr_pref",
4379 			.data		= &ipv6_devconf.accept_ra_rtr_pref,
4380 			.maxlen		= sizeof(int),
4381 			.mode		= 0644,
4382 			.proc_handler	= proc_dointvec,
4383 		},
4384 		{
4385 			.procname	= "router_probe_interval",
4386 			.data		= &ipv6_devconf.rtr_probe_interval,
4387 			.maxlen		= sizeof(int),
4388 			.mode		= 0644,
4389 			.proc_handler	= proc_dointvec_jiffies,
4390 		},
4391 #ifdef CONFIG_IPV6_ROUTE_INFO
4392 		{
4393 			.procname	= "accept_ra_rt_info_max_plen",
4394 			.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
4395 			.maxlen		= sizeof(int),
4396 			.mode		= 0644,
4397 			.proc_handler	= proc_dointvec,
4398 		},
4399 #endif
4400 #endif
4401 		{
4402 			.procname	= "proxy_ndp",
4403 			.data		= &ipv6_devconf.proxy_ndp,
4404 			.maxlen		= sizeof(int),
4405 			.mode		= 0644,
4406 			.proc_handler	= proc_dointvec,
4407 		},
4408 		{
4409 			.procname	= "accept_source_route",
4410 			.data		= &ipv6_devconf.accept_source_route,
4411 			.maxlen		= sizeof(int),
4412 			.mode		= 0644,
4413 			.proc_handler	= proc_dointvec,
4414 		},
4415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4416 		{
4417 			.procname       = "optimistic_dad",
4418 			.data           = &ipv6_devconf.optimistic_dad,
4419 			.maxlen         = sizeof(int),
4420 			.mode           = 0644,
4421 			.proc_handler   = proc_dointvec,
4422 
4423 		},
4424 #endif
4425 #ifdef CONFIG_IPV6_MROUTE
4426 		{
4427 			.procname	= "mc_forwarding",
4428 			.data		= &ipv6_devconf.mc_forwarding,
4429 			.maxlen		= sizeof(int),
4430 			.mode		= 0444,
4431 			.proc_handler	= proc_dointvec,
4432 		},
4433 #endif
4434 		{
4435 			.procname	= "disable_ipv6",
4436 			.data		= &ipv6_devconf.disable_ipv6,
4437 			.maxlen		= sizeof(int),
4438 			.mode		= 0644,
4439 			.proc_handler	= addrconf_sysctl_disable,
4440 		},
4441 		{
4442 			.procname	= "accept_dad",
4443 			.data		= &ipv6_devconf.accept_dad,
4444 			.maxlen		= sizeof(int),
4445 			.mode		= 0644,
4446 			.proc_handler	= proc_dointvec,
4447 		},
4448 		{
4449 			.procname       = "force_tllao",
4450 			.data           = &ipv6_devconf.force_tllao,
4451 			.maxlen         = sizeof(int),
4452 			.mode           = 0644,
4453 			.proc_handler   = proc_dointvec
4454 		},
4455 		{
4456 			/* sentinel */
4457 		}
4458 	},
4459 };
4460 
4461 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4462 		struct inet6_dev *idev, struct ipv6_devconf *p)
4463 {
4464 	int i;
4465 	struct addrconf_sysctl_table *t;
4466 
4467 #define ADDRCONF_CTL_PATH_DEV	3
4468 
4469 	struct ctl_path addrconf_ctl_path[] = {
4470 		{ .procname = "net", },
4471 		{ .procname = "ipv6", },
4472 		{ .procname = "conf", },
4473 		{ /* to be set */ },
4474 		{ },
4475 	};
4476 
4477 
4478 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4479 	if (t == NULL)
4480 		goto out;
4481 
4482 	for (i = 0; t->addrconf_vars[i].data; i++) {
4483 		t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4484 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4485 		t->addrconf_vars[i].extra2 = net;
4486 	}
4487 
4488 	/*
4489 	 * Make a copy of dev_name, because '.procname' is regarded as const
4490 	 * by sysctl and we wouldn't want anyone to change it under our feet
4491 	 * (see SIOCSIFNAME).
4492 	 */
4493 	t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4494 	if (!t->dev_name)
4495 		goto free;
4496 
4497 	addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4498 
4499 	t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4500 			t->addrconf_vars);
4501 	if (t->sysctl_header == NULL)
4502 		goto free_procname;
4503 
4504 	p->sysctl = t;
4505 	return 0;
4506 
4507 free_procname:
4508 	kfree(t->dev_name);
4509 free:
4510 	kfree(t);
4511 out:
4512 	return -ENOBUFS;
4513 }
4514 
4515 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4516 {
4517 	struct addrconf_sysctl_table *t;
4518 
4519 	if (p->sysctl == NULL)
4520 		return;
4521 
4522 	t = p->sysctl;
4523 	p->sysctl = NULL;
4524 	unregister_sysctl_table(t->sysctl_header);
4525 	kfree(t->dev_name);
4526 	kfree(t);
4527 }
4528 
4529 static void addrconf_sysctl_register(struct inet6_dev *idev)
4530 {
4531 	neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4532 			      &ndisc_ifinfo_sysctl_change);
4533 	__addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4534 					idev, &idev->cnf);
4535 }
4536 
4537 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4538 {
4539 	__addrconf_sysctl_unregister(&idev->cnf);
4540 	neigh_sysctl_unregister(idev->nd_parms);
4541 }
4542 
4543 
4544 #endif
4545 
4546 static int __net_init addrconf_init_net(struct net *net)
4547 {
4548 	int err;
4549 	struct ipv6_devconf *all, *dflt;
4550 
4551 	err = -ENOMEM;
4552 	all = &ipv6_devconf;
4553 	dflt = &ipv6_devconf_dflt;
4554 
4555 	if (!net_eq(net, &init_net)) {
4556 		all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4557 		if (all == NULL)
4558 			goto err_alloc_all;
4559 
4560 		dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4561 		if (dflt == NULL)
4562 			goto err_alloc_dflt;
4563 	} else {
4564 		/* these will be inherited by all namespaces */
4565 		dflt->autoconf = ipv6_defaults.autoconf;
4566 		dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4567 	}
4568 
4569 	net->ipv6.devconf_all = all;
4570 	net->ipv6.devconf_dflt = dflt;
4571 
4572 #ifdef CONFIG_SYSCTL
4573 	err = __addrconf_sysctl_register(net, "all", NULL, all);
4574 	if (err < 0)
4575 		goto err_reg_all;
4576 
4577 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4578 	if (err < 0)
4579 		goto err_reg_dflt;
4580 #endif
4581 	return 0;
4582 
4583 #ifdef CONFIG_SYSCTL
4584 err_reg_dflt:
4585 	__addrconf_sysctl_unregister(all);
4586 err_reg_all:
4587 	kfree(dflt);
4588 #endif
4589 err_alloc_dflt:
4590 	kfree(all);
4591 err_alloc_all:
4592 	return err;
4593 }
4594 
4595 static void __net_exit addrconf_exit_net(struct net *net)
4596 {
4597 #ifdef CONFIG_SYSCTL
4598 	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4599 	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
4600 #endif
4601 	if (!net_eq(net, &init_net)) {
4602 		kfree(net->ipv6.devconf_dflt);
4603 		kfree(net->ipv6.devconf_all);
4604 	}
4605 }
4606 
4607 static struct pernet_operations addrconf_ops = {
4608 	.init = addrconf_init_net,
4609 	.exit = addrconf_exit_net,
4610 };
4611 
4612 /*
4613  *      Device notifier
4614  */
4615 
4616 int register_inet6addr_notifier(struct notifier_block *nb)
4617 {
4618 	return atomic_notifier_chain_register(&inet6addr_chain, nb);
4619 }
4620 EXPORT_SYMBOL(register_inet6addr_notifier);
4621 
4622 int unregister_inet6addr_notifier(struct notifier_block *nb)
4623 {
4624 	return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4625 }
4626 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4627 
4628 /*
4629  *	Init / cleanup code
4630  */
4631 
4632 int __init addrconf_init(void)
4633 {
4634 	int i, err;
4635 
4636 	err = ipv6_addr_label_init();
4637 	if (err < 0) {
4638 		printk(KERN_CRIT "IPv6 Addrconf:"
4639 		       " cannot initialize default policy table: %d.\n", err);
4640 		return err;
4641 	}
4642 
4643 	register_pernet_subsys(&addrconf_ops);
4644 
4645 	/* The addrconf netdev notifier requires that loopback_dev
4646 	 * has it's ipv6 private information allocated and setup
4647 	 * before it can bring up and give link-local addresses
4648 	 * to other devices which are up.
4649 	 *
4650 	 * Unfortunately, loopback_dev is not necessarily the first
4651 	 * entry in the global dev_base list of net devices.  In fact,
4652 	 * it is likely to be the very last entry on that list.
4653 	 * So this causes the notifier registry below to try and
4654 	 * give link-local addresses to all devices besides loopback_dev
4655 	 * first, then loopback_dev, which cases all the non-loopback_dev
4656 	 * devices to fail to get a link-local address.
4657 	 *
4658 	 * So, as a temporary fix, allocate the ipv6 structure for
4659 	 * loopback_dev first by hand.
4660 	 * Longer term, all of the dependencies ipv6 has upon the loopback
4661 	 * device and it being up should be removed.
4662 	 */
4663 	rtnl_lock();
4664 	if (!ipv6_add_dev(init_net.loopback_dev))
4665 		err = -ENOMEM;
4666 	rtnl_unlock();
4667 	if (err)
4668 		goto errlo;
4669 
4670 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4671 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4672 
4673 	register_netdevice_notifier(&ipv6_dev_notf);
4674 
4675 	addrconf_verify(0);
4676 
4677 	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4678 	if (err < 0)
4679 		goto errout;
4680 
4681 	/* Only the first call to __rtnl_register can fail */
4682 	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4683 	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4684 	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4685 	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4686 	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4687 
4688 	ipv6_addr_label_rtnl_register();
4689 
4690 	return 0;
4691 errout:
4692 	unregister_netdevice_notifier(&ipv6_dev_notf);
4693 errlo:
4694 	unregister_pernet_subsys(&addrconf_ops);
4695 
4696 	return err;
4697 }
4698 
4699 void addrconf_cleanup(void)
4700 {
4701 	struct net_device *dev;
4702 	int i;
4703 
4704 	unregister_netdevice_notifier(&ipv6_dev_notf);
4705 	unregister_pernet_subsys(&addrconf_ops);
4706 
4707 	rtnl_lock();
4708 
4709 	/* clean dev list */
4710 	for_each_netdev(&init_net, dev) {
4711 		if (__in6_dev_get(dev) == NULL)
4712 			continue;
4713 		addrconf_ifdown(dev, 1);
4714 	}
4715 	addrconf_ifdown(init_net.loopback_dev, 2);
4716 
4717 	/*
4718 	 *	Check hash table.
4719 	 */
4720 	spin_lock_bh(&addrconf_hash_lock);
4721 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
4722 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4723 	spin_unlock_bh(&addrconf_hash_lock);
4724 
4725 	del_timer(&addr_chk_timer);
4726 	rtnl_unlock();
4727 }
4728