xref: /openbmc/linux/net/ipv6/addrconf.c (revision 2702e0a46c2d28da92b32c9b068ee1291fc0de35)
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/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/in6.h>
47 #include <linux/netdevice.h>
48 #include <linux/if_addr.h>
49 #include <linux/if_arp.h>
50 #include <linux/if_arcnet.h>
51 #include <linux/if_infiniband.h>
52 #include <linux/route.h>
53 #include <linux/inetdevice.h>
54 #include <linux/init.h>
55 #ifdef CONFIG_SYSCTL
56 #include <linux/sysctl.h>
57 #endif
58 #include <linux/capability.h>
59 #include <linux/delay.h>
60 #include <linux/notifier.h>
61 #include <linux/string.h>
62 
63 #include <net/net_namespace.h>
64 #include <net/sock.h>
65 #include <net/snmp.h>
66 
67 #include <net/ipv6.h>
68 #include <net/protocol.h>
69 #include <net/ndisc.h>
70 #include <net/ip6_route.h>
71 #include <net/addrconf.h>
72 #include <net/tcp.h>
73 #include <net/ip.h>
74 #include <net/netlink.h>
75 #include <net/pkt_sched.h>
76 #include <linux/if_tunnel.h>
77 #include <linux/rtnetlink.h>
78 
79 #ifdef CONFIG_IPV6_PRIVACY
80 #include <linux/random.h>
81 #endif
82 
83 #include <asm/uaccess.h>
84 #include <asm/unaligned.h>
85 
86 #include <linux/proc_fs.h>
87 #include <linux/seq_file.h>
88 
89 /* Set to 3 to get tracing... */
90 #define ACONF_DEBUG 2
91 
92 #if ACONF_DEBUG >= 3
93 #define ADBG(x) printk x
94 #else
95 #define ADBG(x)
96 #endif
97 
98 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
100 
101 #ifdef CONFIG_SYSCTL
102 static void addrconf_sysctl_register(struct inet6_dev *idev);
103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
104 #else
105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
106 {
107 }
108 
109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
110 {
111 }
112 #endif
113 
114 #ifdef CONFIG_IPV6_PRIVACY
115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
117 static void ipv6_regen_rndid(unsigned long data);
118 
119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
120 #endif
121 
122 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
123 static int ipv6_count_addresses(struct inet6_dev *idev);
124 
125 /*
126  *	Configured unicast address hash table
127  */
128 static struct inet6_ifaddr		*inet6_addr_lst[IN6_ADDR_HSIZE];
129 static DEFINE_RWLOCK(addrconf_hash_lock);
130 
131 static void addrconf_verify(unsigned long);
132 
133 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
134 static DEFINE_SPINLOCK(addrconf_verify_lock);
135 
136 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
137 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
138 
139 static int addrconf_ifdown(struct net_device *dev, int how);
140 
141 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
142 static void addrconf_dad_timer(unsigned long data);
143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
144 static void addrconf_dad_run(struct inet6_dev *idev);
145 static void addrconf_rs_timer(unsigned long data);
146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
147 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 
149 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
150 				struct prefix_info *pinfo);
151 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
152 			      struct net_device *dev);
153 
154 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
155 
156 static struct ipv6_devconf ipv6_devconf __read_mostly = {
157 	.forwarding		= 0,
158 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
159 	.mtu6			= IPV6_MIN_MTU,
160 	.accept_ra		= 1,
161 	.accept_redirects	= 1,
162 	.autoconf		= 1,
163 	.force_mld_version	= 0,
164 	.dad_transmits		= 1,
165 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
166 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
167 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
168 #ifdef CONFIG_IPV6_PRIVACY
169 	.use_tempaddr 		= 0,
170 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
171 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
172 	.regen_max_retry	= REGEN_MAX_RETRY,
173 	.max_desync_factor	= MAX_DESYNC_FACTOR,
174 #endif
175 	.max_addresses		= IPV6_MAX_ADDRESSES,
176 	.accept_ra_defrtr	= 1,
177 	.accept_ra_pinfo	= 1,
178 #ifdef CONFIG_IPV6_ROUTER_PREF
179 	.accept_ra_rtr_pref	= 1,
180 	.rtr_probe_interval	= 60 * HZ,
181 #ifdef CONFIG_IPV6_ROUTE_INFO
182 	.accept_ra_rt_info_max_plen = 0,
183 #endif
184 #endif
185 	.proxy_ndp		= 0,
186 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
187 	.disable_ipv6		= 0,
188 	.accept_dad		= 1,
189 };
190 
191 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
192 	.forwarding		= 0,
193 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
194 	.mtu6			= IPV6_MIN_MTU,
195 	.accept_ra		= 1,
196 	.accept_redirects	= 1,
197 	.autoconf		= 1,
198 	.dad_transmits		= 1,
199 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
200 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
201 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
202 #ifdef CONFIG_IPV6_PRIVACY
203 	.use_tempaddr		= 0,
204 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
205 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
206 	.regen_max_retry	= REGEN_MAX_RETRY,
207 	.max_desync_factor	= MAX_DESYNC_FACTOR,
208 #endif
209 	.max_addresses		= IPV6_MAX_ADDRESSES,
210 	.accept_ra_defrtr	= 1,
211 	.accept_ra_pinfo	= 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213 	.accept_ra_rtr_pref	= 1,
214 	.rtr_probe_interval	= 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216 	.accept_ra_rt_info_max_plen = 0,
217 #endif
218 #endif
219 	.proxy_ndp		= 0,
220 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
221 	.disable_ipv6		= 0,
222 	.accept_dad		= 1,
223 };
224 
225 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
226 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
227 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
228 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
229 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
230 
231 /* Check if a valid qdisc is available */
232 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
233 {
234 	return !qdisc_tx_is_noop(dev);
235 }
236 
237 /* Check if a route is valid prefix route */
238 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
239 {
240 	return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
241 }
242 
243 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
244 {
245 	if (del_timer(&ifp->timer))
246 		__in6_ifa_put(ifp);
247 }
248 
249 enum addrconf_timer_t
250 {
251 	AC_NONE,
252 	AC_DAD,
253 	AC_RS,
254 };
255 
256 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
257 			       enum addrconf_timer_t what,
258 			       unsigned long when)
259 {
260 	if (!del_timer(&ifp->timer))
261 		in6_ifa_hold(ifp);
262 
263 	switch (what) {
264 	case AC_DAD:
265 		ifp->timer.function = addrconf_dad_timer;
266 		break;
267 	case AC_RS:
268 		ifp->timer.function = addrconf_rs_timer;
269 		break;
270 	default:;
271 	}
272 	ifp->timer.expires = jiffies + when;
273 	add_timer(&ifp->timer);
274 }
275 
276 static int snmp6_alloc_dev(struct inet6_dev *idev)
277 {
278 	if (snmp_mib_init((void **)idev->stats.ipv6,
279 			  sizeof(struct ipstats_mib)) < 0)
280 		goto err_ip;
281 	if (snmp_mib_init((void **)idev->stats.icmpv6,
282 			  sizeof(struct icmpv6_mib)) < 0)
283 		goto err_icmp;
284 	if (snmp_mib_init((void **)idev->stats.icmpv6msg,
285 			  sizeof(struct icmpv6msg_mib)) < 0)
286 		goto err_icmpmsg;
287 
288 	return 0;
289 
290 err_icmpmsg:
291 	snmp_mib_free((void **)idev->stats.icmpv6);
292 err_icmp:
293 	snmp_mib_free((void **)idev->stats.ipv6);
294 err_ip:
295 	return -ENOMEM;
296 }
297 
298 static void snmp6_free_dev(struct inet6_dev *idev)
299 {
300 	snmp_mib_free((void **)idev->stats.icmpv6msg);
301 	snmp_mib_free((void **)idev->stats.icmpv6);
302 	snmp_mib_free((void **)idev->stats.ipv6);
303 }
304 
305 /* Nobody refers to this device, we may destroy it. */
306 
307 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
308 {
309 	struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
310 	kfree(idev);
311 }
312 
313 void in6_dev_finish_destroy(struct inet6_dev *idev)
314 {
315 	struct net_device *dev = idev->dev;
316 
317 	WARN_ON(idev->addr_list != NULL);
318 	WARN_ON(idev->mc_list != NULL);
319 
320 #ifdef NET_REFCNT_DEBUG
321 	printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
322 #endif
323 	dev_put(dev);
324 	if (!idev->dead) {
325 		printk("Freeing alive inet6 device %p\n", idev);
326 		return;
327 	}
328 	snmp6_free_dev(idev);
329 	call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
330 }
331 
332 EXPORT_SYMBOL(in6_dev_finish_destroy);
333 
334 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
335 {
336 	struct inet6_dev *ndev;
337 
338 	ASSERT_RTNL();
339 
340 	if (dev->mtu < IPV6_MIN_MTU)
341 		return NULL;
342 
343 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
344 
345 	if (ndev == NULL)
346 		return NULL;
347 
348 	rwlock_init(&ndev->lock);
349 	ndev->dev = dev;
350 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
351 	ndev->cnf.mtu6 = dev->mtu;
352 	ndev->cnf.sysctl = NULL;
353 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
354 	if (ndev->nd_parms == NULL) {
355 		kfree(ndev);
356 		return NULL;
357 	}
358 	if (ndev->cnf.forwarding)
359 		dev_disable_lro(dev);
360 	/* We refer to the device */
361 	dev_hold(dev);
362 
363 	if (snmp6_alloc_dev(ndev) < 0) {
364 		ADBG((KERN_WARNING
365 			"%s(): cannot allocate memory for statistics; dev=%s.\n",
366 			__func__, dev->name));
367 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
368 		ndev->dead = 1;
369 		in6_dev_finish_destroy(ndev);
370 		return NULL;
371 	}
372 
373 	if (snmp6_register_dev(ndev) < 0) {
374 		ADBG((KERN_WARNING
375 			"%s(): cannot create /proc/net/dev_snmp6/%s\n",
376 			__func__, dev->name));
377 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 		ndev->dead = 1;
379 		in6_dev_finish_destroy(ndev);
380 		return NULL;
381 	}
382 
383 	/* One reference from device.  We must do this before
384 	 * we invoke __ipv6_regen_rndid().
385 	 */
386 	in6_dev_hold(ndev);
387 
388 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
389 		ndev->cnf.accept_dad = -1;
390 
391 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
392 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
393 		printk(KERN_INFO
394 		       "%s: Disabled Multicast RS\n",
395 		       dev->name);
396 		ndev->cnf.rtr_solicits = 0;
397 	}
398 #endif
399 
400 #ifdef CONFIG_IPV6_PRIVACY
401 	setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
402 	if ((dev->flags&IFF_LOOPBACK) ||
403 	    dev->type == ARPHRD_TUNNEL ||
404 	    dev->type == ARPHRD_TUNNEL6 ||
405 	    dev->type == ARPHRD_SIT ||
406 	    dev->type == ARPHRD_NONE) {
407 		printk(KERN_INFO
408 		       "%s: Disabled Privacy Extensions\n",
409 		       dev->name);
410 		ndev->cnf.use_tempaddr = -1;
411 	} else {
412 		in6_dev_hold(ndev);
413 		ipv6_regen_rndid((unsigned long) ndev);
414 	}
415 #endif
416 
417 	if (netif_running(dev) && addrconf_qdisc_ok(dev))
418 		ndev->if_flags |= IF_READY;
419 
420 	ipv6_mc_init_dev(ndev);
421 	ndev->tstamp = jiffies;
422 	addrconf_sysctl_register(ndev);
423 	/* protected by rtnl_lock */
424 	rcu_assign_pointer(dev->ip6_ptr, ndev);
425 
426 	/* Join all-node multicast group */
427 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
428 
429 	return ndev;
430 }
431 
432 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
433 {
434 	struct inet6_dev *idev;
435 
436 	ASSERT_RTNL();
437 
438 	if ((idev = __in6_dev_get(dev)) == NULL) {
439 		if ((idev = ipv6_add_dev(dev)) == NULL)
440 			return NULL;
441 	}
442 
443 	if (dev->flags&IFF_UP)
444 		ipv6_mc_up(idev);
445 	return idev;
446 }
447 
448 #ifdef CONFIG_SYSCTL
449 static void dev_forward_change(struct inet6_dev *idev)
450 {
451 	struct net_device *dev;
452 	struct inet6_ifaddr *ifa;
453 
454 	if (!idev)
455 		return;
456 	dev = idev->dev;
457 	if (idev->cnf.forwarding)
458 		dev_disable_lro(dev);
459 	if (dev && (dev->flags & IFF_MULTICAST)) {
460 		if (idev->cnf.forwarding)
461 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
462 		else
463 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
464 	}
465 	for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
466 		if (ifa->flags&IFA_F_TENTATIVE)
467 			continue;
468 		if (idev->cnf.forwarding)
469 			addrconf_join_anycast(ifa);
470 		else
471 			addrconf_leave_anycast(ifa);
472 	}
473 }
474 
475 
476 static void addrconf_forward_change(struct net *net, __s32 newf)
477 {
478 	struct net_device *dev;
479 	struct inet6_dev *idev;
480 
481 	read_lock(&dev_base_lock);
482 	for_each_netdev(net, dev) {
483 		rcu_read_lock();
484 		idev = __in6_dev_get(dev);
485 		if (idev) {
486 			int changed = (!idev->cnf.forwarding) ^ (!newf);
487 			idev->cnf.forwarding = newf;
488 			if (changed)
489 				dev_forward_change(idev);
490 		}
491 		rcu_read_unlock();
492 	}
493 	read_unlock(&dev_base_lock);
494 }
495 
496 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
497 {
498 	struct net *net;
499 
500 	net = (struct net *)table->extra2;
501 	if (p == &net->ipv6.devconf_dflt->forwarding)
502 		return;
503 
504 	rtnl_lock();
505 	if (p == &net->ipv6.devconf_all->forwarding) {
506 		__s32 newf = net->ipv6.devconf_all->forwarding;
507 		net->ipv6.devconf_dflt->forwarding = newf;
508 		addrconf_forward_change(net, newf);
509 	} else if ((!*p) ^ (!old))
510 		dev_forward_change((struct inet6_dev *)table->extra1);
511 	rtnl_unlock();
512 
513 	if (*p)
514 		rt6_purge_dflt_routers(net);
515 }
516 #endif
517 
518 /* Nobody refers to this ifaddr, destroy it */
519 
520 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
521 {
522 	WARN_ON(ifp->if_next != NULL);
523 	WARN_ON(ifp->lst_next != NULL);
524 
525 #ifdef NET_REFCNT_DEBUG
526 	printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
527 #endif
528 
529 	in6_dev_put(ifp->idev);
530 
531 	if (del_timer(&ifp->timer))
532 		printk("Timer is still running, when freeing ifa=%p\n", ifp);
533 
534 	if (!ifp->dead) {
535 		printk("Freeing alive inet6 address %p\n", ifp);
536 		return;
537 	}
538 	dst_release(&ifp->rt->u.dst);
539 
540 	kfree(ifp);
541 }
542 
543 static void
544 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
545 {
546 	struct inet6_ifaddr *ifa, **ifap;
547 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
548 
549 	/*
550 	 * Each device address list is sorted in order of scope -
551 	 * global before linklocal.
552 	 */
553 	for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
554 	     ifap = &ifa->if_next) {
555 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
556 			break;
557 	}
558 
559 	ifp->if_next = *ifap;
560 	*ifap = ifp;
561 }
562 
563 /*
564  *	Hash function taken from net_alias.c
565  */
566 static u8 ipv6_addr_hash(const struct in6_addr *addr)
567 {
568 	__u32 word;
569 
570 	/*
571 	 * We perform the hash function over the last 64 bits of the address
572 	 * This will include the IEEE address token on links that support it.
573 	 */
574 
575 	word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
576 	word ^= (word >> 16);
577 	word ^= (word >> 8);
578 
579 	return ((word ^ (word >> 4)) & 0x0f);
580 }
581 
582 /* On success it returns ifp with increased reference count */
583 
584 static struct inet6_ifaddr *
585 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
586 	      int scope, u32 flags)
587 {
588 	struct inet6_ifaddr *ifa = NULL;
589 	struct rt6_info *rt;
590 	int hash;
591 	int err = 0;
592 	int addr_type = ipv6_addr_type(addr);
593 
594 	if (addr_type == IPV6_ADDR_ANY ||
595 	    addr_type & IPV6_ADDR_MULTICAST ||
596 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
597 	     addr_type & IPV6_ADDR_LOOPBACK))
598 		return ERR_PTR(-EADDRNOTAVAIL);
599 
600 	rcu_read_lock_bh();
601 	if (idev->dead) {
602 		err = -ENODEV;			/*XXX*/
603 		goto out2;
604 	}
605 
606 	write_lock(&addrconf_hash_lock);
607 
608 	/* Ignore adding duplicate addresses on an interface */
609 	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
610 		ADBG(("ipv6_add_addr: already assigned\n"));
611 		err = -EEXIST;
612 		goto out;
613 	}
614 
615 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
616 
617 	if (ifa == NULL) {
618 		ADBG(("ipv6_add_addr: malloc failed\n"));
619 		err = -ENOBUFS;
620 		goto out;
621 	}
622 
623 	rt = addrconf_dst_alloc(idev, addr, 0);
624 	if (IS_ERR(rt)) {
625 		err = PTR_ERR(rt);
626 		goto out;
627 	}
628 
629 	ipv6_addr_copy(&ifa->addr, addr);
630 
631 	spin_lock_init(&ifa->lock);
632 	init_timer(&ifa->timer);
633 	ifa->timer.data = (unsigned long) ifa;
634 	ifa->scope = scope;
635 	ifa->prefix_len = pfxlen;
636 	ifa->flags = flags | IFA_F_TENTATIVE;
637 	ifa->cstamp = ifa->tstamp = jiffies;
638 
639 	ifa->rt = rt;
640 
641 	/*
642 	 * part one of RFC 4429, section 3.3
643 	 * We should not configure an address as
644 	 * optimistic if we do not yet know the link
645 	 * layer address of our nexhop router
646 	 */
647 
648 	if (rt->rt6i_nexthop == NULL)
649 		ifa->flags &= ~IFA_F_OPTIMISTIC;
650 
651 	ifa->idev = idev;
652 	in6_dev_hold(idev);
653 	/* For caller */
654 	in6_ifa_hold(ifa);
655 
656 	/* Add to big hash table */
657 	hash = ipv6_addr_hash(addr);
658 
659 	ifa->lst_next = inet6_addr_lst[hash];
660 	inet6_addr_lst[hash] = ifa;
661 	in6_ifa_hold(ifa);
662 	write_unlock(&addrconf_hash_lock);
663 
664 	write_lock(&idev->lock);
665 	/* Add to inet6_dev unicast addr list. */
666 	ipv6_link_dev_addr(idev, ifa);
667 
668 #ifdef CONFIG_IPV6_PRIVACY
669 	if (ifa->flags&IFA_F_TEMPORARY) {
670 		ifa->tmp_next = idev->tempaddr_list;
671 		idev->tempaddr_list = ifa;
672 		in6_ifa_hold(ifa);
673 	}
674 #endif
675 
676 	in6_ifa_hold(ifa);
677 	write_unlock(&idev->lock);
678 out2:
679 	rcu_read_unlock_bh();
680 
681 	if (likely(err == 0))
682 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
683 	else {
684 		kfree(ifa);
685 		ifa = ERR_PTR(err);
686 	}
687 
688 	return ifa;
689 out:
690 	write_unlock(&addrconf_hash_lock);
691 	goto out2;
692 }
693 
694 /* This function wants to get referenced ifp and releases it before return */
695 
696 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
697 {
698 	struct inet6_ifaddr *ifa, **ifap;
699 	struct inet6_dev *idev = ifp->idev;
700 	int hash;
701 	int deleted = 0, onlink = 0;
702 	unsigned long expires = jiffies;
703 
704 	hash = ipv6_addr_hash(&ifp->addr);
705 
706 	ifp->dead = 1;
707 
708 	write_lock_bh(&addrconf_hash_lock);
709 	for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
710 	     ifap = &ifa->lst_next) {
711 		if (ifa == ifp) {
712 			*ifap = ifa->lst_next;
713 			__in6_ifa_put(ifp);
714 			ifa->lst_next = NULL;
715 			break;
716 		}
717 	}
718 	write_unlock_bh(&addrconf_hash_lock);
719 
720 	write_lock_bh(&idev->lock);
721 #ifdef CONFIG_IPV6_PRIVACY
722 	if (ifp->flags&IFA_F_TEMPORARY) {
723 		for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
724 		     ifap = &ifa->tmp_next) {
725 			if (ifa == ifp) {
726 				*ifap = ifa->tmp_next;
727 				if (ifp->ifpub) {
728 					in6_ifa_put(ifp->ifpub);
729 					ifp->ifpub = NULL;
730 				}
731 				__in6_ifa_put(ifp);
732 				ifa->tmp_next = NULL;
733 				break;
734 			}
735 		}
736 	}
737 #endif
738 
739 	for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
740 		if (ifa == ifp) {
741 			*ifap = ifa->if_next;
742 			__in6_ifa_put(ifp);
743 			ifa->if_next = NULL;
744 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
745 				break;
746 			deleted = 1;
747 			continue;
748 		} else if (ifp->flags & IFA_F_PERMANENT) {
749 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
750 					      ifp->prefix_len)) {
751 				if (ifa->flags & IFA_F_PERMANENT) {
752 					onlink = 1;
753 					if (deleted)
754 						break;
755 				} else {
756 					unsigned long lifetime;
757 
758 					if (!onlink)
759 						onlink = -1;
760 
761 					spin_lock(&ifa->lock);
762 
763 					lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
764 					/*
765 					 * Note: Because this address is
766 					 * not permanent, lifetime <
767 					 * LONG_MAX / HZ here.
768 					 */
769 					if (time_before(expires,
770 							ifa->tstamp + lifetime * HZ))
771 						expires = ifa->tstamp + lifetime * HZ;
772 					spin_unlock(&ifa->lock);
773 				}
774 			}
775 		}
776 		ifap = &ifa->if_next;
777 	}
778 	write_unlock_bh(&idev->lock);
779 
780 	addrconf_del_timer(ifp);
781 
782 	ipv6_ifa_notify(RTM_DELADDR, ifp);
783 
784 	atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
785 
786 	/*
787 	 * Purge or update corresponding prefix
788 	 *
789 	 * 1) we don't purge prefix here if address was not permanent.
790 	 *    prefix is managed by its own lifetime.
791 	 * 2) if there're no addresses, delete prefix.
792 	 * 3) if there're still other permanent address(es),
793 	 *    corresponding prefix is still permanent.
794 	 * 4) otherwise, update prefix lifetime to the
795 	 *    longest valid lifetime among the corresponding
796 	 *    addresses on the device.
797 	 *    Note: subsequent RA will update lifetime.
798 	 *
799 	 * --yoshfuji
800 	 */
801 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
802 		struct in6_addr prefix;
803 		struct rt6_info *rt;
804 		struct net *net = dev_net(ifp->idev->dev);
805 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
806 		rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
807 
808 		if (rt && addrconf_is_prefix_route(rt)) {
809 			if (onlink == 0) {
810 				ip6_del_rt(rt);
811 				rt = NULL;
812 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
813 				rt->rt6i_expires = expires;
814 				rt->rt6i_flags |= RTF_EXPIRES;
815 			}
816 		}
817 		dst_release(&rt->u.dst);
818 	}
819 
820 	in6_ifa_put(ifp);
821 }
822 
823 #ifdef CONFIG_IPV6_PRIVACY
824 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
825 {
826 	struct inet6_dev *idev = ifp->idev;
827 	struct in6_addr addr, *tmpaddr;
828 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
829 	unsigned long regen_advance;
830 	int tmp_plen;
831 	int ret = 0;
832 	int max_addresses;
833 	u32 addr_flags;
834 
835 	write_lock(&idev->lock);
836 	if (ift) {
837 		spin_lock_bh(&ift->lock);
838 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
839 		spin_unlock_bh(&ift->lock);
840 		tmpaddr = &addr;
841 	} else {
842 		tmpaddr = NULL;
843 	}
844 retry:
845 	in6_dev_hold(idev);
846 	if (idev->cnf.use_tempaddr <= 0) {
847 		write_unlock(&idev->lock);
848 		printk(KERN_INFO
849 			"ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
850 		in6_dev_put(idev);
851 		ret = -1;
852 		goto out;
853 	}
854 	spin_lock_bh(&ifp->lock);
855 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
856 		idev->cnf.use_tempaddr = -1;	/*XXX*/
857 		spin_unlock_bh(&ifp->lock);
858 		write_unlock(&idev->lock);
859 		printk(KERN_WARNING
860 			"ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
861 		in6_dev_put(idev);
862 		ret = -1;
863 		goto out;
864 	}
865 	in6_ifa_hold(ifp);
866 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
867 	if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
868 		spin_unlock_bh(&ifp->lock);
869 		write_unlock(&idev->lock);
870 		printk(KERN_WARNING
871 			"ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
872 		in6_ifa_put(ifp);
873 		in6_dev_put(idev);
874 		ret = -1;
875 		goto out;
876 	}
877 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
878 	tmp_valid_lft = min_t(__u32,
879 			      ifp->valid_lft,
880 			      idev->cnf.temp_valid_lft);
881 	tmp_prefered_lft = min_t(__u32,
882 				 ifp->prefered_lft,
883 				 idev->cnf.temp_prefered_lft - desync_factor / HZ);
884 	tmp_plen = ifp->prefix_len;
885 	max_addresses = idev->cnf.max_addresses;
886 	tmp_cstamp = ifp->cstamp;
887 	tmp_tstamp = ifp->tstamp;
888 	spin_unlock_bh(&ifp->lock);
889 
890 	regen_advance = idev->cnf.regen_max_retry *
891 	                idev->cnf.dad_transmits *
892 	                idev->nd_parms->retrans_time / HZ;
893 	write_unlock(&idev->lock);
894 
895 	/* A temporary address is created only if this calculated Preferred
896 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
897 	 * an implementation must not create a temporary address with a zero
898 	 * Preferred Lifetime.
899 	 */
900 	if (tmp_prefered_lft <= regen_advance) {
901 		in6_ifa_put(ifp);
902 		in6_dev_put(idev);
903 		ret = -1;
904 		goto out;
905 	}
906 
907 	addr_flags = IFA_F_TEMPORARY;
908 	/* set in addrconf_prefix_rcv() */
909 	if (ifp->flags & IFA_F_OPTIMISTIC)
910 		addr_flags |= IFA_F_OPTIMISTIC;
911 
912 	ift = !max_addresses ||
913 	      ipv6_count_addresses(idev) < max_addresses ?
914 		ipv6_add_addr(idev, &addr, tmp_plen,
915 			      ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
916 			      addr_flags) : NULL;
917 	if (!ift || IS_ERR(ift)) {
918 		in6_ifa_put(ifp);
919 		in6_dev_put(idev);
920 		printk(KERN_INFO
921 			"ipv6_create_tempaddr(): retry temporary address regeneration.\n");
922 		tmpaddr = &addr;
923 		write_lock(&idev->lock);
924 		goto retry;
925 	}
926 
927 	spin_lock_bh(&ift->lock);
928 	ift->ifpub = ifp;
929 	ift->valid_lft = tmp_valid_lft;
930 	ift->prefered_lft = tmp_prefered_lft;
931 	ift->cstamp = tmp_cstamp;
932 	ift->tstamp = tmp_tstamp;
933 	spin_unlock_bh(&ift->lock);
934 
935 	addrconf_dad_start(ift, 0);
936 	in6_ifa_put(ift);
937 	in6_dev_put(idev);
938 out:
939 	return ret;
940 }
941 #endif
942 
943 /*
944  *	Choose an appropriate source address (RFC3484)
945  */
946 enum {
947 	IPV6_SADDR_RULE_INIT = 0,
948 	IPV6_SADDR_RULE_LOCAL,
949 	IPV6_SADDR_RULE_SCOPE,
950 	IPV6_SADDR_RULE_PREFERRED,
951 #ifdef CONFIG_IPV6_MIP6
952 	IPV6_SADDR_RULE_HOA,
953 #endif
954 	IPV6_SADDR_RULE_OIF,
955 	IPV6_SADDR_RULE_LABEL,
956 #ifdef CONFIG_IPV6_PRIVACY
957 	IPV6_SADDR_RULE_PRIVACY,
958 #endif
959 	IPV6_SADDR_RULE_ORCHID,
960 	IPV6_SADDR_RULE_PREFIX,
961 	IPV6_SADDR_RULE_MAX
962 };
963 
964 struct ipv6_saddr_score {
965 	int			rule;
966 	int			addr_type;
967 	struct inet6_ifaddr	*ifa;
968 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
969 	int			scopedist;
970 	int			matchlen;
971 };
972 
973 struct ipv6_saddr_dst {
974 	const struct in6_addr *addr;
975 	int ifindex;
976 	int scope;
977 	int label;
978 	unsigned int prefs;
979 };
980 
981 static inline int ipv6_saddr_preferred(int type)
982 {
983 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
984 		    IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
985 		return 1;
986 	return 0;
987 }
988 
989 static int ipv6_get_saddr_eval(struct net *net,
990 			       struct ipv6_saddr_score *score,
991 			       struct ipv6_saddr_dst *dst,
992 			       int i)
993 {
994 	int ret;
995 
996 	if (i <= score->rule) {
997 		switch (i) {
998 		case IPV6_SADDR_RULE_SCOPE:
999 			ret = score->scopedist;
1000 			break;
1001 		case IPV6_SADDR_RULE_PREFIX:
1002 			ret = score->matchlen;
1003 			break;
1004 		default:
1005 			ret = !!test_bit(i, score->scorebits);
1006 		}
1007 		goto out;
1008 	}
1009 
1010 	switch (i) {
1011 	case IPV6_SADDR_RULE_INIT:
1012 		/* Rule 0: remember if hiscore is not ready yet */
1013 		ret = !!score->ifa;
1014 		break;
1015 	case IPV6_SADDR_RULE_LOCAL:
1016 		/* Rule 1: Prefer same address */
1017 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018 		break;
1019 	case IPV6_SADDR_RULE_SCOPE:
1020 		/* Rule 2: Prefer appropriate scope
1021 		 *
1022 		 *      ret
1023 		 *       ^
1024 		 *    -1 |  d 15
1025 		 *    ---+--+-+---> scope
1026 		 *       |
1027 		 *       |             d is scope of the destination.
1028 		 *  B-d  |  \
1029 		 *       |   \      <- smaller scope is better if
1030 		 *  B-15 |    \        if scope is enough for destinaion.
1031 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1032 		 * d-C-1 | /
1033 		 *       |/         <- greater is better
1034 		 *   -C  /             if scope is not enough for destination.
1035 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1036 		 *
1037 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039 		 * Assume B = 0 and we get C > 29.
1040 		 */
1041 		ret = __ipv6_addr_src_scope(score->addr_type);
1042 		if (ret >= dst->scope)
1043 			ret = -ret;
1044 		else
1045 			ret -= 128;	/* 30 is enough */
1046 		score->scopedist = ret;
1047 		break;
1048 	case IPV6_SADDR_RULE_PREFERRED:
1049 		/* Rule 3: Avoid deprecated and optimistic addresses */
1050 		ret = ipv6_saddr_preferred(score->addr_type) ||
1051 		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052 		break;
1053 #ifdef CONFIG_IPV6_MIP6
1054 	case IPV6_SADDR_RULE_HOA:
1055 	    {
1056 		/* Rule 4: Prefer home address */
1057 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059 		break;
1060 	    }
1061 #endif
1062 	case IPV6_SADDR_RULE_OIF:
1063 		/* Rule 5: Prefer outgoing interface */
1064 		ret = (!dst->ifindex ||
1065 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1066 		break;
1067 	case IPV6_SADDR_RULE_LABEL:
1068 		/* Rule 6: Prefer matching label */
1069 		ret = ipv6_addr_label(net,
1070 				      &score->ifa->addr, score->addr_type,
1071 				      score->ifa->idev->dev->ifindex) == dst->label;
1072 		break;
1073 #ifdef CONFIG_IPV6_PRIVACY
1074 	case IPV6_SADDR_RULE_PRIVACY:
1075 	    {
1076 		/* Rule 7: Prefer public address
1077 		 * Note: prefer temprary address if use_tempaddr >= 2
1078 		 */
1079 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081 				score->ifa->idev->cnf.use_tempaddr >= 2;
1082 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083 		break;
1084 	    }
1085 #endif
1086 	case IPV6_SADDR_RULE_ORCHID:
1087 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1088 		 *	    non-ORCHID vs non-ORCHID
1089 		 */
1090 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091 			ipv6_addr_orchid(dst->addr));
1092 		break;
1093 	case IPV6_SADDR_RULE_PREFIX:
1094 		/* Rule 8: Use longest matching prefix */
1095 		score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096 						       dst->addr);
1097 		break;
1098 	default:
1099 		ret = 0;
1100 	}
1101 
1102 	if (ret)
1103 		__set_bit(i, score->scorebits);
1104 	score->rule = i;
1105 out:
1106 	return ret;
1107 }
1108 
1109 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110 		       const struct in6_addr *daddr, unsigned int prefs,
1111 		       struct in6_addr *saddr)
1112 {
1113 	struct ipv6_saddr_score scores[2],
1114 				*score = &scores[0], *hiscore = &scores[1];
1115 	struct ipv6_saddr_dst dst;
1116 	struct net_device *dev;
1117 	int dst_type;
1118 
1119 	dst_type = __ipv6_addr_type(daddr);
1120 	dst.addr = daddr;
1121 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122 	dst.scope = __ipv6_addr_src_scope(dst_type);
1123 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124 	dst.prefs = prefs;
1125 
1126 	hiscore->rule = -1;
1127 	hiscore->ifa = NULL;
1128 
1129 	read_lock(&dev_base_lock);
1130 	rcu_read_lock();
1131 
1132 	for_each_netdev(net, dev) {
1133 		struct inet6_dev *idev;
1134 
1135 		/* Candidate Source Address (section 4)
1136 		 *  - multicast and link-local destination address,
1137 		 *    the set of candidate source address MUST only
1138 		 *    include addresses assigned to interfaces
1139 		 *    belonging to the same link as the outgoing
1140 		 *    interface.
1141 		 * (- For site-local destination addresses, the
1142 		 *    set of candidate source addresses MUST only
1143 		 *    include addresses assigned to interfaces
1144 		 *    belonging to the same site as the outgoing
1145 		 *    interface.)
1146 		 */
1147 		if (((dst_type & IPV6_ADDR_MULTICAST) ||
1148 		     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1149 		    dst.ifindex && dev->ifindex != dst.ifindex)
1150 			continue;
1151 
1152 		idev = __in6_dev_get(dev);
1153 		if (!idev)
1154 			continue;
1155 
1156 		read_lock_bh(&idev->lock);
1157 		for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1158 			int i;
1159 
1160 			/*
1161 			 * - Tentative Address (RFC2462 section 5.4)
1162 			 *  - A tentative address is not considered
1163 			 *    "assigned to an interface" in the traditional
1164 			 *    sense, unless it is also flagged as optimistic.
1165 			 * - Candidate Source Address (section 4)
1166 			 *  - In any case, anycast addresses, multicast
1167 			 *    addresses, and the unspecified address MUST
1168 			 *    NOT be included in a candidate set.
1169 			 */
1170 			if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1171 			    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1172 				continue;
1173 
1174 			score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1175 
1176 			if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1177 				     score->addr_type & IPV6_ADDR_MULTICAST)) {
1178 				LIMIT_NETDEBUG(KERN_DEBUG
1179 					       "ADDRCONF: unspecified / multicast address "
1180 					       "assigned as unicast address on %s",
1181 					       dev->name);
1182 				continue;
1183 			}
1184 
1185 			score->rule = -1;
1186 			bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1187 
1188 			for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1189 				int minihiscore, miniscore;
1190 
1191 				minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1192 				miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1193 
1194 				if (minihiscore > miniscore) {
1195 					if (i == IPV6_SADDR_RULE_SCOPE &&
1196 					    score->scopedist > 0) {
1197 						/*
1198 						 * special case:
1199 						 * each remaining entry
1200 						 * has too small (not enough)
1201 						 * scope, because ifa entries
1202 						 * are sorted by their scope
1203 						 * values.
1204 						 */
1205 						goto try_nextdev;
1206 					}
1207 					break;
1208 				} else if (minihiscore < miniscore) {
1209 					struct ipv6_saddr_score *tmp;
1210 
1211 					if (hiscore->ifa)
1212 						in6_ifa_put(hiscore->ifa);
1213 
1214 					in6_ifa_hold(score->ifa);
1215 
1216 					tmp = hiscore;
1217 					hiscore = score;
1218 					score = tmp;
1219 
1220 					/* restore our iterator */
1221 					score->ifa = hiscore->ifa;
1222 
1223 					break;
1224 				}
1225 			}
1226 		}
1227 try_nextdev:
1228 		read_unlock_bh(&idev->lock);
1229 	}
1230 	rcu_read_unlock();
1231 	read_unlock(&dev_base_lock);
1232 
1233 	if (!hiscore->ifa)
1234 		return -EADDRNOTAVAIL;
1235 
1236 	ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1237 	in6_ifa_put(hiscore->ifa);
1238 	return 0;
1239 }
1240 
1241 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1242 
1243 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1244 		    unsigned char banned_flags)
1245 {
1246 	struct inet6_dev *idev;
1247 	int err = -EADDRNOTAVAIL;
1248 
1249 	rcu_read_lock();
1250 	if ((idev = __in6_dev_get(dev)) != NULL) {
1251 		struct inet6_ifaddr *ifp;
1252 
1253 		read_lock_bh(&idev->lock);
1254 		for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1255 			if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1256 				ipv6_addr_copy(addr, &ifp->addr);
1257 				err = 0;
1258 				break;
1259 			}
1260 		}
1261 		read_unlock_bh(&idev->lock);
1262 	}
1263 	rcu_read_unlock();
1264 	return err;
1265 }
1266 
1267 static int ipv6_count_addresses(struct inet6_dev *idev)
1268 {
1269 	int cnt = 0;
1270 	struct inet6_ifaddr *ifp;
1271 
1272 	read_lock_bh(&idev->lock);
1273 	for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1274 		cnt++;
1275 	read_unlock_bh(&idev->lock);
1276 	return cnt;
1277 }
1278 
1279 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1280 		  struct net_device *dev, int strict)
1281 {
1282 	struct inet6_ifaddr * ifp;
1283 	u8 hash = ipv6_addr_hash(addr);
1284 
1285 	read_lock_bh(&addrconf_hash_lock);
1286 	for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1287 		if (!net_eq(dev_net(ifp->idev->dev), net))
1288 			continue;
1289 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1290 		    !(ifp->flags&IFA_F_TENTATIVE)) {
1291 			if (dev == NULL || ifp->idev->dev == dev ||
1292 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1293 				break;
1294 		}
1295 	}
1296 	read_unlock_bh(&addrconf_hash_lock);
1297 	return ifp != NULL;
1298 }
1299 EXPORT_SYMBOL(ipv6_chk_addr);
1300 
1301 static
1302 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1303 		       struct net_device *dev)
1304 {
1305 	struct inet6_ifaddr * ifp;
1306 	u8 hash = ipv6_addr_hash(addr);
1307 
1308 	for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1309 		if (!net_eq(dev_net(ifp->idev->dev), net))
1310 			continue;
1311 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1312 			if (dev == NULL || ifp->idev->dev == dev)
1313 				break;
1314 		}
1315 	}
1316 	return ifp != NULL;
1317 }
1318 
1319 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1320 {
1321 	struct inet6_dev *idev;
1322 	struct inet6_ifaddr *ifa;
1323 	int	onlink;
1324 
1325 	onlink = 0;
1326 	rcu_read_lock();
1327 	idev = __in6_dev_get(dev);
1328 	if (idev) {
1329 		read_lock_bh(&idev->lock);
1330 		for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1331 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1332 						   ifa->prefix_len);
1333 			if (onlink)
1334 				break;
1335 		}
1336 		read_unlock_bh(&idev->lock);
1337 	}
1338 	rcu_read_unlock();
1339 	return onlink;
1340 }
1341 
1342 EXPORT_SYMBOL(ipv6_chk_prefix);
1343 
1344 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1345 				     struct net_device *dev, int strict)
1346 {
1347 	struct inet6_ifaddr * ifp;
1348 	u8 hash = ipv6_addr_hash(addr);
1349 
1350 	read_lock_bh(&addrconf_hash_lock);
1351 	for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1352 		if (!net_eq(dev_net(ifp->idev->dev), net))
1353 			continue;
1354 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1355 			if (dev == NULL || ifp->idev->dev == dev ||
1356 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1357 				in6_ifa_hold(ifp);
1358 				break;
1359 			}
1360 		}
1361 	}
1362 	read_unlock_bh(&addrconf_hash_lock);
1363 
1364 	return ifp;
1365 }
1366 
1367 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1368 {
1369 	const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1370 	const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1371 	__be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1372 	__be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1373 	int sk_ipv6only = ipv6_only_sock(sk);
1374 	int sk2_ipv6only = inet_v6_ipv6only(sk2);
1375 	int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1376 	int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1377 
1378 	if (!sk2_rcv_saddr && !sk_ipv6only)
1379 		return 1;
1380 
1381 	if (addr_type2 == IPV6_ADDR_ANY &&
1382 	    !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1383 		return 1;
1384 
1385 	if (addr_type == IPV6_ADDR_ANY &&
1386 	    !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1387 		return 1;
1388 
1389 	if (sk2_rcv_saddr6 &&
1390 	    ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1391 		return 1;
1392 
1393 	if (addr_type == IPV6_ADDR_MAPPED &&
1394 	    !sk2_ipv6only &&
1395 	    (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1396 		return 1;
1397 
1398 	return 0;
1399 }
1400 
1401 /* Gets referenced address, destroys ifaddr */
1402 
1403 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1404 {
1405 	if (ifp->flags&IFA_F_PERMANENT) {
1406 		spin_lock_bh(&ifp->lock);
1407 		addrconf_del_timer(ifp);
1408 		ifp->flags |= IFA_F_TENTATIVE;
1409 		spin_unlock_bh(&ifp->lock);
1410 		in6_ifa_put(ifp);
1411 #ifdef CONFIG_IPV6_PRIVACY
1412 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1413 		struct inet6_ifaddr *ifpub;
1414 		spin_lock_bh(&ifp->lock);
1415 		ifpub = ifp->ifpub;
1416 		if (ifpub) {
1417 			in6_ifa_hold(ifpub);
1418 			spin_unlock_bh(&ifp->lock);
1419 			ipv6_create_tempaddr(ifpub, ifp);
1420 			in6_ifa_put(ifpub);
1421 		} else {
1422 			spin_unlock_bh(&ifp->lock);
1423 		}
1424 		ipv6_del_addr(ifp);
1425 #endif
1426 	} else
1427 		ipv6_del_addr(ifp);
1428 }
1429 
1430 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1431 {
1432 	struct inet6_dev *idev = ifp->idev;
1433 	if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1434 		struct in6_addr addr;
1435 
1436 		addr.s6_addr32[0] = htonl(0xfe800000);
1437 		addr.s6_addr32[1] = 0;
1438 
1439 		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1440 		    ipv6_addr_equal(&ifp->addr, &addr)) {
1441 			/* DAD failed for link-local based on MAC address */
1442 			idev->cnf.disable_ipv6 = 1;
1443 		}
1444 	}
1445 
1446 	if (net_ratelimit())
1447 		printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1448 	addrconf_dad_stop(ifp);
1449 }
1450 
1451 /* Join to solicited addr multicast group. */
1452 
1453 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1454 {
1455 	struct in6_addr maddr;
1456 
1457 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1458 		return;
1459 
1460 	addrconf_addr_solict_mult(addr, &maddr);
1461 	ipv6_dev_mc_inc(dev, &maddr);
1462 }
1463 
1464 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1465 {
1466 	struct in6_addr maddr;
1467 
1468 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1469 		return;
1470 
1471 	addrconf_addr_solict_mult(addr, &maddr);
1472 	__ipv6_dev_mc_dec(idev, &maddr);
1473 }
1474 
1475 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1476 {
1477 	struct in6_addr addr;
1478 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1479 	if (ipv6_addr_any(&addr))
1480 		return;
1481 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1482 }
1483 
1484 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1485 {
1486 	struct in6_addr addr;
1487 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1488 	if (ipv6_addr_any(&addr))
1489 		return;
1490 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1491 }
1492 
1493 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1494 {
1495 	if (dev->addr_len != ETH_ALEN)
1496 		return -1;
1497 	memcpy(eui, dev->dev_addr, 3);
1498 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1499 
1500 	/*
1501 	 * The zSeries OSA network cards can be shared among various
1502 	 * OS instances, but the OSA cards have only one MAC address.
1503 	 * This leads to duplicate address conflicts in conjunction
1504 	 * with IPv6 if more than one instance uses the same card.
1505 	 *
1506 	 * The driver for these cards can deliver a unique 16-bit
1507 	 * identifier for each instance sharing the same card.  It is
1508 	 * placed instead of 0xFFFE in the interface identifier.  The
1509 	 * "u" bit of the interface identifier is not inverted in this
1510 	 * case.  Hence the resulting interface identifier has local
1511 	 * scope according to RFC2373.
1512 	 */
1513 	if (dev->dev_id) {
1514 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1515 		eui[4] = dev->dev_id & 0xFF;
1516 	} else {
1517 		eui[3] = 0xFF;
1518 		eui[4] = 0xFE;
1519 		eui[0] ^= 2;
1520 	}
1521 	return 0;
1522 }
1523 
1524 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1525 {
1526 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1527 	if (dev->addr_len != ARCNET_ALEN)
1528 		return -1;
1529 	memset(eui, 0, 7);
1530 	eui[7] = *(u8*)dev->dev_addr;
1531 	return 0;
1532 }
1533 
1534 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1535 {
1536 	if (dev->addr_len != INFINIBAND_ALEN)
1537 		return -1;
1538 	memcpy(eui, dev->dev_addr + 12, 8);
1539 	eui[0] |= 2;
1540 	return 0;
1541 }
1542 
1543 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1544 {
1545 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1546 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1547 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1548 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1549 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1550 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1551 	eui[1] = 0;
1552 	eui[2] = 0x5E;
1553 	eui[3] = 0xFE;
1554 	memcpy(eui + 4, &addr, 4);
1555 	return 0;
1556 }
1557 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1558 
1559 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1560 {
1561 	if (dev->priv_flags & IFF_ISATAP)
1562 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563 	return -1;
1564 }
1565 
1566 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1567 {
1568 	switch (dev->type) {
1569 	case ARPHRD_ETHER:
1570 	case ARPHRD_FDDI:
1571 	case ARPHRD_IEEE802_TR:
1572 		return addrconf_ifid_eui48(eui, dev);
1573 	case ARPHRD_ARCNET:
1574 		return addrconf_ifid_arcnet(eui, dev);
1575 	case ARPHRD_INFINIBAND:
1576 		return addrconf_ifid_infiniband(eui, dev);
1577 	case ARPHRD_SIT:
1578 		return addrconf_ifid_sit(eui, dev);
1579 	}
1580 	return -1;
1581 }
1582 
1583 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1584 {
1585 	int err = -1;
1586 	struct inet6_ifaddr *ifp;
1587 
1588 	read_lock_bh(&idev->lock);
1589 	for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1590 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1591 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1592 			err = 0;
1593 			break;
1594 		}
1595 	}
1596 	read_unlock_bh(&idev->lock);
1597 	return err;
1598 }
1599 
1600 #ifdef CONFIG_IPV6_PRIVACY
1601 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1602 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1603 {
1604 regen:
1605 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1606 	idev->rndid[0] &= ~0x02;
1607 
1608 	/*
1609 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1610 	 * check if generated address is not inappropriate
1611 	 *
1612 	 *  - Reserved subnet anycast (RFC 2526)
1613 	 *	11111101 11....11 1xxxxxxx
1614 	 *  - ISATAP (RFC4214) 6.1
1615 	 *	00-00-5E-FE-xx-xx-xx-xx
1616 	 *  - value 0
1617 	 *  - XXX: already assigned to an address on the device
1618 	 */
1619 	if (idev->rndid[0] == 0xfd &&
1620 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1621 	    (idev->rndid[7]&0x80))
1622 		goto regen;
1623 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1624 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1625 			goto regen;
1626 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627 			goto regen;
1628 	}
1629 
1630 	return 0;
1631 }
1632 
1633 static void ipv6_regen_rndid(unsigned long data)
1634 {
1635 	struct inet6_dev *idev = (struct inet6_dev *) data;
1636 	unsigned long expires;
1637 
1638 	rcu_read_lock_bh();
1639 	write_lock_bh(&idev->lock);
1640 
1641 	if (idev->dead)
1642 		goto out;
1643 
1644 	if (__ipv6_regen_rndid(idev) < 0)
1645 		goto out;
1646 
1647 	expires = jiffies +
1648 		idev->cnf.temp_prefered_lft * HZ -
1649 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1650 	if (time_before(expires, jiffies)) {
1651 		printk(KERN_WARNING
1652 			"ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1653 			idev->dev->name);
1654 		goto out;
1655 	}
1656 
1657 	if (!mod_timer(&idev->regen_timer, expires))
1658 		in6_dev_hold(idev);
1659 
1660 out:
1661 	write_unlock_bh(&idev->lock);
1662 	rcu_read_unlock_bh();
1663 	in6_dev_put(idev);
1664 }
1665 
1666 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1667 	int ret = 0;
1668 
1669 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1670 		ret = __ipv6_regen_rndid(idev);
1671 	return ret;
1672 }
1673 #endif
1674 
1675 /*
1676  *	Add prefix route.
1677  */
1678 
1679 static void
1680 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1681 		      unsigned long expires, u32 flags)
1682 {
1683 	struct fib6_config cfg = {
1684 		.fc_table = RT6_TABLE_PREFIX,
1685 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1686 		.fc_ifindex = dev->ifindex,
1687 		.fc_expires = expires,
1688 		.fc_dst_len = plen,
1689 		.fc_flags = RTF_UP | flags,
1690 		.fc_nlinfo.nl_net = dev_net(dev),
1691 		.fc_protocol = RTPROT_KERNEL,
1692 	};
1693 
1694 	ipv6_addr_copy(&cfg.fc_dst, pfx);
1695 
1696 	/* Prevent useless cloning on PtP SIT.
1697 	   This thing is done here expecting that the whole
1698 	   class of non-broadcast devices need not cloning.
1699 	 */
1700 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1701 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1702 		cfg.fc_flags |= RTF_NONEXTHOP;
1703 #endif
1704 
1705 	ip6_route_add(&cfg);
1706 }
1707 
1708 /* Create "default" multicast route to the interface */
1709 
1710 static void addrconf_add_mroute(struct net_device *dev)
1711 {
1712 	struct fib6_config cfg = {
1713 		.fc_table = RT6_TABLE_LOCAL,
1714 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1715 		.fc_ifindex = dev->ifindex,
1716 		.fc_dst_len = 8,
1717 		.fc_flags = RTF_UP,
1718 		.fc_nlinfo.nl_net = dev_net(dev),
1719 	};
1720 
1721 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1722 
1723 	ip6_route_add(&cfg);
1724 }
1725 
1726 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1727 static void sit_route_add(struct net_device *dev)
1728 {
1729 	struct fib6_config cfg = {
1730 		.fc_table = RT6_TABLE_MAIN,
1731 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1732 		.fc_ifindex = dev->ifindex,
1733 		.fc_dst_len = 96,
1734 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
1735 		.fc_nlinfo.nl_net = dev_net(dev),
1736 	};
1737 
1738 	/* prefix length - 96 bits "::d.d.d.d" */
1739 	ip6_route_add(&cfg);
1740 }
1741 #endif
1742 
1743 static void addrconf_add_lroute(struct net_device *dev)
1744 {
1745 	struct in6_addr addr;
1746 
1747 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1748 	addrconf_prefix_route(&addr, 64, dev, 0, 0);
1749 }
1750 
1751 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1752 {
1753 	struct inet6_dev *idev;
1754 
1755 	ASSERT_RTNL();
1756 
1757 	if ((idev = ipv6_find_idev(dev)) == NULL)
1758 		return NULL;
1759 
1760 	/* Add default multicast route */
1761 	addrconf_add_mroute(dev);
1762 
1763 	/* Add link local route */
1764 	addrconf_add_lroute(dev);
1765 	return idev;
1766 }
1767 
1768 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1769 {
1770 	struct prefix_info *pinfo;
1771 	__u32 valid_lft;
1772 	__u32 prefered_lft;
1773 	int addr_type;
1774 	struct inet6_dev *in6_dev;
1775 
1776 	pinfo = (struct prefix_info *) opt;
1777 
1778 	if (len < sizeof(struct prefix_info)) {
1779 		ADBG(("addrconf: prefix option too short\n"));
1780 		return;
1781 	}
1782 
1783 	/*
1784 	 *	Validation checks ([ADDRCONF], page 19)
1785 	 */
1786 
1787 	addr_type = ipv6_addr_type(&pinfo->prefix);
1788 
1789 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1790 		return;
1791 
1792 	valid_lft = ntohl(pinfo->valid);
1793 	prefered_lft = ntohl(pinfo->prefered);
1794 
1795 	if (prefered_lft > valid_lft) {
1796 		if (net_ratelimit())
1797 			printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1798 		return;
1799 	}
1800 
1801 	in6_dev = in6_dev_get(dev);
1802 
1803 	if (in6_dev == NULL) {
1804 		if (net_ratelimit())
1805 			printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1806 		return;
1807 	}
1808 
1809 	/*
1810 	 *	Two things going on here:
1811 	 *	1) Add routes for on-link prefixes
1812 	 *	2) Configure prefixes with the auto flag set
1813 	 */
1814 
1815 	if (pinfo->onlink) {
1816 		struct rt6_info *rt;
1817 		unsigned long rt_expires;
1818 
1819 		/* Avoid arithmetic overflow. Really, we could
1820 		 * save rt_expires in seconds, likely valid_lft,
1821 		 * but it would require division in fib gc, that it
1822 		 * not good.
1823 		 */
1824 		if (HZ > USER_HZ)
1825 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1826 		else
1827 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1828 
1829 		if (addrconf_finite_timeout(rt_expires))
1830 			rt_expires *= HZ;
1831 
1832 		rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1833 				dev->ifindex, 1);
1834 
1835 		if (rt && addrconf_is_prefix_route(rt)) {
1836 			/* Autoconf prefix route */
1837 			if (valid_lft == 0) {
1838 				ip6_del_rt(rt);
1839 				rt = NULL;
1840 			} else if (addrconf_finite_timeout(rt_expires)) {
1841 				/* not infinity */
1842 				rt->rt6i_expires = jiffies + rt_expires;
1843 				rt->rt6i_flags |= RTF_EXPIRES;
1844 			} else {
1845 				rt->rt6i_flags &= ~RTF_EXPIRES;
1846 				rt->rt6i_expires = 0;
1847 			}
1848 		} else if (valid_lft) {
1849 			clock_t expires = 0;
1850 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1851 			if (addrconf_finite_timeout(rt_expires)) {
1852 				/* not infinity */
1853 				flags |= RTF_EXPIRES;
1854 				expires = jiffies_to_clock_t(rt_expires);
1855 			}
1856 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1857 					      dev, expires, flags);
1858 		}
1859 		if (rt)
1860 			dst_release(&rt->u.dst);
1861 	}
1862 
1863 	/* Try to figure out our local address for this prefix */
1864 
1865 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1866 		struct inet6_ifaddr * ifp;
1867 		struct in6_addr addr;
1868 		int create = 0, update_lft = 0;
1869 		struct net *net = dev_net(dev);
1870 
1871 		if (pinfo->prefix_len == 64) {
1872 			memcpy(&addr, &pinfo->prefix, 8);
1873 			if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1874 			    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1875 				in6_dev_put(in6_dev);
1876 				return;
1877 			}
1878 			goto ok;
1879 		}
1880 		if (net_ratelimit())
1881 			printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1882 			       pinfo->prefix_len);
1883 		in6_dev_put(in6_dev);
1884 		return;
1885 
1886 ok:
1887 
1888 		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1889 
1890 		if (ifp == NULL && valid_lft) {
1891 			int max_addresses = in6_dev->cnf.max_addresses;
1892 			u32 addr_flags = 0;
1893 
1894 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1895 			if (in6_dev->cnf.optimistic_dad &&
1896 			    !net->ipv6.devconf_all->forwarding)
1897 				addr_flags = IFA_F_OPTIMISTIC;
1898 #endif
1899 
1900 			/* Do not allow to create too much of autoconfigured
1901 			 * addresses; this would be too easy way to crash kernel.
1902 			 */
1903 			if (!max_addresses ||
1904 			    ipv6_count_addresses(in6_dev) < max_addresses)
1905 				ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1906 						    addr_type&IPV6_ADDR_SCOPE_MASK,
1907 						    addr_flags);
1908 
1909 			if (!ifp || IS_ERR(ifp)) {
1910 				in6_dev_put(in6_dev);
1911 				return;
1912 			}
1913 
1914 			update_lft = create = 1;
1915 			ifp->cstamp = jiffies;
1916 			addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1917 		}
1918 
1919 		if (ifp) {
1920 			int flags;
1921 			unsigned long now;
1922 #ifdef CONFIG_IPV6_PRIVACY
1923 			struct inet6_ifaddr *ift;
1924 #endif
1925 			u32 stored_lft;
1926 
1927 			/* update lifetime (RFC2462 5.5.3 e) */
1928 			spin_lock(&ifp->lock);
1929 			now = jiffies;
1930 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1931 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1932 			else
1933 				stored_lft = 0;
1934 			if (!update_lft && stored_lft) {
1935 				if (valid_lft > MIN_VALID_LIFETIME ||
1936 				    valid_lft > stored_lft)
1937 					update_lft = 1;
1938 				else if (stored_lft <= MIN_VALID_LIFETIME) {
1939 					/* valid_lft <= stored_lft is always true */
1940 					/* XXX: IPsec */
1941 					update_lft = 0;
1942 				} else {
1943 					valid_lft = MIN_VALID_LIFETIME;
1944 					if (valid_lft < prefered_lft)
1945 						prefered_lft = valid_lft;
1946 					update_lft = 1;
1947 				}
1948 			}
1949 
1950 			if (update_lft) {
1951 				ifp->valid_lft = valid_lft;
1952 				ifp->prefered_lft = prefered_lft;
1953 				ifp->tstamp = now;
1954 				flags = ifp->flags;
1955 				ifp->flags &= ~IFA_F_DEPRECATED;
1956 				spin_unlock(&ifp->lock);
1957 
1958 				if (!(flags&IFA_F_TENTATIVE))
1959 					ipv6_ifa_notify(0, ifp);
1960 			} else
1961 				spin_unlock(&ifp->lock);
1962 
1963 #ifdef CONFIG_IPV6_PRIVACY
1964 			read_lock_bh(&in6_dev->lock);
1965 			/* update all temporary addresses in the list */
1966 			for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1967 				/*
1968 				 * When adjusting the lifetimes of an existing
1969 				 * temporary address, only lower the lifetimes.
1970 				 * Implementations must not increase the
1971 				 * lifetimes of an existing temporary address
1972 				 * when processing a Prefix Information Option.
1973 				 */
1974 				if (ifp != ift->ifpub)
1975 					continue;
1976 
1977 				spin_lock(&ift->lock);
1978 				flags = ift->flags;
1979 				if (ift->valid_lft > valid_lft &&
1980 				    ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1981 					ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1982 				if (ift->prefered_lft > prefered_lft &&
1983 				    ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1984 					ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1985 				spin_unlock(&ift->lock);
1986 				if (!(flags&IFA_F_TENTATIVE))
1987 					ipv6_ifa_notify(0, ift);
1988 			}
1989 
1990 			if (create && in6_dev->cnf.use_tempaddr > 0) {
1991 				/*
1992 				 * When a new public address is created as described in [ADDRCONF],
1993 				 * also create a new temporary address.
1994 				 */
1995 				read_unlock_bh(&in6_dev->lock);
1996 				ipv6_create_tempaddr(ifp, NULL);
1997 			} else {
1998 				read_unlock_bh(&in6_dev->lock);
1999 			}
2000 #endif
2001 			in6_ifa_put(ifp);
2002 			addrconf_verify(0);
2003 		}
2004 	}
2005 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2006 	in6_dev_put(in6_dev);
2007 }
2008 
2009 /*
2010  *	Set destination address.
2011  *	Special case for SIT interfaces where we create a new "virtual"
2012  *	device.
2013  */
2014 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2015 {
2016 	struct in6_ifreq ireq;
2017 	struct net_device *dev;
2018 	int err = -EINVAL;
2019 
2020 	rtnl_lock();
2021 
2022 	err = -EFAULT;
2023 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2024 		goto err_exit;
2025 
2026 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2027 
2028 	err = -ENODEV;
2029 	if (dev == NULL)
2030 		goto err_exit;
2031 
2032 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2033 	if (dev->type == ARPHRD_SIT) {
2034 		const struct net_device_ops *ops = dev->netdev_ops;
2035 		struct ifreq ifr;
2036 		struct ip_tunnel_parm p;
2037 
2038 		err = -EADDRNOTAVAIL;
2039 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2040 			goto err_exit;
2041 
2042 		memset(&p, 0, sizeof(p));
2043 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2044 		p.iph.saddr = 0;
2045 		p.iph.version = 4;
2046 		p.iph.ihl = 5;
2047 		p.iph.protocol = IPPROTO_IPV6;
2048 		p.iph.ttl = 64;
2049 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2050 
2051 		if (ops->ndo_do_ioctl) {
2052 			mm_segment_t oldfs = get_fs();
2053 
2054 			set_fs(KERNEL_DS);
2055 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2056 			set_fs(oldfs);
2057 		} else
2058 			err = -EOPNOTSUPP;
2059 
2060 		if (err == 0) {
2061 			err = -ENOBUFS;
2062 			dev = __dev_get_by_name(net, p.name);
2063 			if (!dev)
2064 				goto err_exit;
2065 			err = dev_open(dev);
2066 		}
2067 	}
2068 #endif
2069 
2070 err_exit:
2071 	rtnl_unlock();
2072 	return err;
2073 }
2074 
2075 /*
2076  *	Manual configuration of address on an interface
2077  */
2078 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2079 			  unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2080 			  __u32 valid_lft)
2081 {
2082 	struct inet6_ifaddr *ifp;
2083 	struct inet6_dev *idev;
2084 	struct net_device *dev;
2085 	int scope;
2086 	u32 flags;
2087 	clock_t expires;
2088 	unsigned long timeout;
2089 
2090 	ASSERT_RTNL();
2091 
2092 	if (plen > 128)
2093 		return -EINVAL;
2094 
2095 	/* check the lifetime */
2096 	if (!valid_lft || prefered_lft > valid_lft)
2097 		return -EINVAL;
2098 
2099 	dev = __dev_get_by_index(net, ifindex);
2100 	if (!dev)
2101 		return -ENODEV;
2102 
2103 	if ((idev = addrconf_add_dev(dev)) == NULL)
2104 		return -ENOBUFS;
2105 
2106 	scope = ipv6_addr_scope(pfx);
2107 
2108 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2109 	if (addrconf_finite_timeout(timeout)) {
2110 		expires = jiffies_to_clock_t(timeout * HZ);
2111 		valid_lft = timeout;
2112 		flags = RTF_EXPIRES;
2113 	} else {
2114 		expires = 0;
2115 		flags = 0;
2116 		ifa_flags |= IFA_F_PERMANENT;
2117 	}
2118 
2119 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2120 	if (addrconf_finite_timeout(timeout)) {
2121 		if (timeout == 0)
2122 			ifa_flags |= IFA_F_DEPRECATED;
2123 		prefered_lft = timeout;
2124 	}
2125 
2126 	ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2127 
2128 	if (!IS_ERR(ifp)) {
2129 		spin_lock_bh(&ifp->lock);
2130 		ifp->valid_lft = valid_lft;
2131 		ifp->prefered_lft = prefered_lft;
2132 		ifp->tstamp = jiffies;
2133 		spin_unlock_bh(&ifp->lock);
2134 
2135 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2136 				      expires, flags);
2137 		/*
2138 		 * Note that section 3.1 of RFC 4429 indicates
2139 		 * that the Optimistic flag should not be set for
2140 		 * manually configured addresses
2141 		 */
2142 		addrconf_dad_start(ifp, 0);
2143 		in6_ifa_put(ifp);
2144 		addrconf_verify(0);
2145 		return 0;
2146 	}
2147 
2148 	return PTR_ERR(ifp);
2149 }
2150 
2151 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2152 			  unsigned int plen)
2153 {
2154 	struct inet6_ifaddr *ifp;
2155 	struct inet6_dev *idev;
2156 	struct net_device *dev;
2157 
2158 	if (plen > 128)
2159 		return -EINVAL;
2160 
2161 	dev = __dev_get_by_index(net, ifindex);
2162 	if (!dev)
2163 		return -ENODEV;
2164 
2165 	if ((idev = __in6_dev_get(dev)) == NULL)
2166 		return -ENXIO;
2167 
2168 	read_lock_bh(&idev->lock);
2169 	for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2170 		if (ifp->prefix_len == plen &&
2171 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2172 			in6_ifa_hold(ifp);
2173 			read_unlock_bh(&idev->lock);
2174 
2175 			ipv6_del_addr(ifp);
2176 
2177 			/* If the last address is deleted administratively,
2178 			   disable IPv6 on this interface.
2179 			 */
2180 			if (idev->addr_list == NULL)
2181 				addrconf_ifdown(idev->dev, 1);
2182 			return 0;
2183 		}
2184 	}
2185 	read_unlock_bh(&idev->lock);
2186 	return -EADDRNOTAVAIL;
2187 }
2188 
2189 
2190 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2191 {
2192 	struct in6_ifreq ireq;
2193 	int err;
2194 
2195 	if (!capable(CAP_NET_ADMIN))
2196 		return -EPERM;
2197 
2198 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2199 		return -EFAULT;
2200 
2201 	rtnl_lock();
2202 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2203 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2204 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2205 	rtnl_unlock();
2206 	return err;
2207 }
2208 
2209 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2210 {
2211 	struct in6_ifreq ireq;
2212 	int err;
2213 
2214 	if (!capable(CAP_NET_ADMIN))
2215 		return -EPERM;
2216 
2217 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2218 		return -EFAULT;
2219 
2220 	rtnl_lock();
2221 	err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2222 			     ireq.ifr6_prefixlen);
2223 	rtnl_unlock();
2224 	return err;
2225 }
2226 
2227 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2228 static void sit_add_v4_addrs(struct inet6_dev *idev)
2229 {
2230 	struct inet6_ifaddr * ifp;
2231 	struct in6_addr addr;
2232 	struct net_device *dev;
2233 	struct net *net = dev_net(idev->dev);
2234 	int scope;
2235 
2236 	ASSERT_RTNL();
2237 
2238 	memset(&addr, 0, sizeof(struct in6_addr));
2239 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2240 
2241 	if (idev->dev->flags&IFF_POINTOPOINT) {
2242 		addr.s6_addr32[0] = htonl(0xfe800000);
2243 		scope = IFA_LINK;
2244 	} else {
2245 		scope = IPV6_ADDR_COMPATv4;
2246 	}
2247 
2248 	if (addr.s6_addr32[3]) {
2249 		ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2250 		if (!IS_ERR(ifp)) {
2251 			spin_lock_bh(&ifp->lock);
2252 			ifp->flags &= ~IFA_F_TENTATIVE;
2253 			spin_unlock_bh(&ifp->lock);
2254 			ipv6_ifa_notify(RTM_NEWADDR, ifp);
2255 			in6_ifa_put(ifp);
2256 		}
2257 		return;
2258 	}
2259 
2260 	for_each_netdev(net, dev) {
2261 		struct in_device * in_dev = __in_dev_get_rtnl(dev);
2262 		if (in_dev && (dev->flags & IFF_UP)) {
2263 			struct in_ifaddr * ifa;
2264 
2265 			int flag = scope;
2266 
2267 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2268 				int plen;
2269 
2270 				addr.s6_addr32[3] = ifa->ifa_local;
2271 
2272 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2273 					continue;
2274 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2275 					if (idev->dev->flags&IFF_POINTOPOINT)
2276 						continue;
2277 					flag |= IFA_HOST;
2278 				}
2279 				if (idev->dev->flags&IFF_POINTOPOINT)
2280 					plen = 64;
2281 				else
2282 					plen = 96;
2283 
2284 				ifp = ipv6_add_addr(idev, &addr, plen, flag,
2285 						    IFA_F_PERMANENT);
2286 				if (!IS_ERR(ifp)) {
2287 					spin_lock_bh(&ifp->lock);
2288 					ifp->flags &= ~IFA_F_TENTATIVE;
2289 					spin_unlock_bh(&ifp->lock);
2290 					ipv6_ifa_notify(RTM_NEWADDR, ifp);
2291 					in6_ifa_put(ifp);
2292 				}
2293 			}
2294 		}
2295 	}
2296 }
2297 #endif
2298 
2299 static void init_loopback(struct net_device *dev)
2300 {
2301 	struct inet6_dev  *idev;
2302 	struct inet6_ifaddr * ifp;
2303 
2304 	/* ::1 */
2305 
2306 	ASSERT_RTNL();
2307 
2308 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2309 		printk(KERN_DEBUG "init loopback: add_dev failed\n");
2310 		return;
2311 	}
2312 
2313 	ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2314 	if (!IS_ERR(ifp)) {
2315 		spin_lock_bh(&ifp->lock);
2316 		ifp->flags &= ~IFA_F_TENTATIVE;
2317 		spin_unlock_bh(&ifp->lock);
2318 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2319 		in6_ifa_put(ifp);
2320 	}
2321 }
2322 
2323 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2324 {
2325 	struct inet6_ifaddr * ifp;
2326 	u32 addr_flags = IFA_F_PERMANENT;
2327 
2328 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2329 	if (idev->cnf.optimistic_dad &&
2330 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2331 		addr_flags |= IFA_F_OPTIMISTIC;
2332 #endif
2333 
2334 
2335 	ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2336 	if (!IS_ERR(ifp)) {
2337 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2338 		addrconf_dad_start(ifp, 0);
2339 		in6_ifa_put(ifp);
2340 	}
2341 }
2342 
2343 static void addrconf_dev_config(struct net_device *dev)
2344 {
2345 	struct in6_addr addr;
2346 	struct inet6_dev    * idev;
2347 
2348 	ASSERT_RTNL();
2349 
2350 	if ((dev->type != ARPHRD_ETHER) &&
2351 	    (dev->type != ARPHRD_FDDI) &&
2352 	    (dev->type != ARPHRD_IEEE802_TR) &&
2353 	    (dev->type != ARPHRD_ARCNET) &&
2354 	    (dev->type != ARPHRD_INFINIBAND)) {
2355 		/* Alas, we support only Ethernet autoconfiguration. */
2356 		return;
2357 	}
2358 
2359 	idev = addrconf_add_dev(dev);
2360 	if (idev == NULL)
2361 		return;
2362 
2363 	memset(&addr, 0, sizeof(struct in6_addr));
2364 	addr.s6_addr32[0] = htonl(0xFE800000);
2365 
2366 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2367 		addrconf_add_linklocal(idev, &addr);
2368 }
2369 
2370 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2371 static void addrconf_sit_config(struct net_device *dev)
2372 {
2373 	struct inet6_dev *idev;
2374 
2375 	ASSERT_RTNL();
2376 
2377 	/*
2378 	 * Configure the tunnel with one of our IPv4
2379 	 * addresses... we should configure all of
2380 	 * our v4 addrs in the tunnel
2381 	 */
2382 
2383 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2384 		printk(KERN_DEBUG "init sit: add_dev failed\n");
2385 		return;
2386 	}
2387 
2388 	if (dev->priv_flags & IFF_ISATAP) {
2389 		struct in6_addr addr;
2390 
2391 		ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2392 		addrconf_prefix_route(&addr, 64, dev, 0, 0);
2393 		if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2394 			addrconf_add_linklocal(idev, &addr);
2395 		return;
2396 	}
2397 
2398 	sit_add_v4_addrs(idev);
2399 
2400 	if (dev->flags&IFF_POINTOPOINT) {
2401 		addrconf_add_mroute(dev);
2402 		addrconf_add_lroute(dev);
2403 	} else
2404 		sit_route_add(dev);
2405 }
2406 #endif
2407 
2408 static inline int
2409 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2410 {
2411 	struct in6_addr lladdr;
2412 
2413 	if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2414 		addrconf_add_linklocal(idev, &lladdr);
2415 		return 0;
2416 	}
2417 	return -1;
2418 }
2419 
2420 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2421 {
2422 	struct net_device *link_dev;
2423 	struct net *net = dev_net(idev->dev);
2424 
2425 	/* first try to inherit the link-local address from the link device */
2426 	if (idev->dev->iflink &&
2427 	    (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2428 		if (!ipv6_inherit_linklocal(idev, link_dev))
2429 			return;
2430 	}
2431 	/* then try to inherit it from any device */
2432 	for_each_netdev(net, link_dev) {
2433 		if (!ipv6_inherit_linklocal(idev, link_dev))
2434 			return;
2435 	}
2436 	printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2437 }
2438 
2439 /*
2440  * Autoconfigure tunnel with a link-local address so routing protocols,
2441  * DHCPv6, MLD etc. can be run over the virtual link
2442  */
2443 
2444 static void addrconf_ip6_tnl_config(struct net_device *dev)
2445 {
2446 	struct inet6_dev *idev;
2447 
2448 	ASSERT_RTNL();
2449 
2450 	if ((idev = addrconf_add_dev(dev)) == NULL) {
2451 		printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2452 		return;
2453 	}
2454 	ip6_tnl_add_linklocal(idev);
2455 }
2456 
2457 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2458 			   void * data)
2459 {
2460 	struct net_device *dev = (struct net_device *) data;
2461 	struct inet6_dev *idev = __in6_dev_get(dev);
2462 	int run_pending = 0;
2463 	int err;
2464 
2465 	switch(event) {
2466 	case NETDEV_REGISTER:
2467 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2468 			idev = ipv6_add_dev(dev);
2469 			if (!idev)
2470 				return notifier_from_errno(-ENOMEM);
2471 		}
2472 		break;
2473 	case NETDEV_UP:
2474 	case NETDEV_CHANGE:
2475 		if (dev->flags & IFF_SLAVE)
2476 			break;
2477 
2478 		if (event == NETDEV_UP) {
2479 			if (!addrconf_qdisc_ok(dev)) {
2480 				/* device is not ready yet. */
2481 				printk(KERN_INFO
2482 					"ADDRCONF(NETDEV_UP): %s: "
2483 					"link is not ready\n",
2484 					dev->name);
2485 				break;
2486 			}
2487 
2488 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
2489 				idev = ipv6_add_dev(dev);
2490 
2491 			if (idev) {
2492 				idev->if_flags |= IF_READY;
2493 				run_pending = 1;
2494 			}
2495 		} else {
2496 			if (!addrconf_qdisc_ok(dev)) {
2497 				/* device is still not ready. */
2498 				break;
2499 			}
2500 
2501 			if (idev) {
2502 				if (idev->if_flags & IF_READY) {
2503 					/* device is already configured. */
2504 					break;
2505 				}
2506 				idev->if_flags |= IF_READY;
2507 			}
2508 
2509 			printk(KERN_INFO
2510 					"ADDRCONF(NETDEV_CHANGE): %s: "
2511 					"link becomes ready\n",
2512 					dev->name);
2513 
2514 			run_pending = 1;
2515 		}
2516 
2517 		switch(dev->type) {
2518 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2519 		case ARPHRD_SIT:
2520 			addrconf_sit_config(dev);
2521 			break;
2522 #endif
2523 		case ARPHRD_TUNNEL6:
2524 			addrconf_ip6_tnl_config(dev);
2525 			break;
2526 		case ARPHRD_LOOPBACK:
2527 			init_loopback(dev);
2528 			break;
2529 
2530 		default:
2531 			addrconf_dev_config(dev);
2532 			break;
2533 		}
2534 		if (idev) {
2535 			if (run_pending)
2536 				addrconf_dad_run(idev);
2537 
2538 			/* If the MTU changed during the interface down, when the
2539 			   interface up, the changed MTU must be reflected in the
2540 			   idev as well as routers.
2541 			 */
2542 			if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2543 				rt6_mtu_change(dev, dev->mtu);
2544 				idev->cnf.mtu6 = dev->mtu;
2545 			}
2546 			idev->tstamp = jiffies;
2547 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2548 			/* If the changed mtu during down is lower than IPV6_MIN_MTU
2549 			   stop IPv6 on this interface.
2550 			 */
2551 			if (dev->mtu < IPV6_MIN_MTU)
2552 				addrconf_ifdown(dev, event != NETDEV_DOWN);
2553 		}
2554 		break;
2555 
2556 	case NETDEV_CHANGEMTU:
2557 		if (idev && dev->mtu >= IPV6_MIN_MTU) {
2558 			rt6_mtu_change(dev, dev->mtu);
2559 			idev->cnf.mtu6 = dev->mtu;
2560 			break;
2561 		}
2562 
2563 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2564 			idev = ipv6_add_dev(dev);
2565 			if (idev)
2566 				break;
2567 		}
2568 
2569 		/* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2570 
2571 	case NETDEV_DOWN:
2572 	case NETDEV_UNREGISTER:
2573 		/*
2574 		 *	Remove all addresses from this interface.
2575 		 */
2576 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2577 		break;
2578 
2579 	case NETDEV_CHANGENAME:
2580 		if (idev) {
2581 			snmp6_unregister_dev(idev);
2582 			addrconf_sysctl_unregister(idev);
2583 			addrconf_sysctl_register(idev);
2584 			err = snmp6_register_dev(idev);
2585 			if (err)
2586 				return notifier_from_errno(err);
2587 		}
2588 		break;
2589 	}
2590 
2591 	return NOTIFY_OK;
2592 }
2593 
2594 /*
2595  *	addrconf module should be notified of a device going up
2596  */
2597 static struct notifier_block ipv6_dev_notf = {
2598 	.notifier_call = addrconf_notify,
2599 	.priority = 0
2600 };
2601 
2602 static int addrconf_ifdown(struct net_device *dev, int how)
2603 {
2604 	struct inet6_dev *idev;
2605 	struct inet6_ifaddr *ifa, **bifa;
2606 	struct net *net = dev_net(dev);
2607 	int i;
2608 
2609 	ASSERT_RTNL();
2610 
2611 	if ((dev->flags & IFF_LOOPBACK) && how == 1)
2612 		how = 0;
2613 
2614 	rt6_ifdown(net, dev);
2615 	neigh_ifdown(&nd_tbl, dev);
2616 
2617 	idev = __in6_dev_get(dev);
2618 	if (idev == NULL)
2619 		return -ENODEV;
2620 
2621 	/* Step 1: remove reference to ipv6 device from parent device.
2622 		   Do not dev_put!
2623 	 */
2624 	if (how) {
2625 		idev->dead = 1;
2626 
2627 		/* protected by rtnl_lock */
2628 		rcu_assign_pointer(dev->ip6_ptr, NULL);
2629 
2630 		/* Step 1.5: remove snmp6 entry */
2631 		snmp6_unregister_dev(idev);
2632 
2633 	}
2634 
2635 	/* Step 2: clear hash table */
2636 	for (i=0; i<IN6_ADDR_HSIZE; i++) {
2637 		bifa = &inet6_addr_lst[i];
2638 
2639 		write_lock_bh(&addrconf_hash_lock);
2640 		while ((ifa = *bifa) != NULL) {
2641 			if (ifa->idev == idev) {
2642 				*bifa = ifa->lst_next;
2643 				ifa->lst_next = NULL;
2644 				addrconf_del_timer(ifa);
2645 				in6_ifa_put(ifa);
2646 				continue;
2647 			}
2648 			bifa = &ifa->lst_next;
2649 		}
2650 		write_unlock_bh(&addrconf_hash_lock);
2651 	}
2652 
2653 	write_lock_bh(&idev->lock);
2654 
2655 	/* Step 3: clear flags for stateless addrconf */
2656 	if (!how)
2657 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2658 
2659 	/* Step 4: clear address list */
2660 #ifdef CONFIG_IPV6_PRIVACY
2661 	if (how && del_timer(&idev->regen_timer))
2662 		in6_dev_put(idev);
2663 
2664 	/* clear tempaddr list */
2665 	while ((ifa = idev->tempaddr_list) != NULL) {
2666 		idev->tempaddr_list = ifa->tmp_next;
2667 		ifa->tmp_next = NULL;
2668 		ifa->dead = 1;
2669 		write_unlock_bh(&idev->lock);
2670 		spin_lock_bh(&ifa->lock);
2671 
2672 		if (ifa->ifpub) {
2673 			in6_ifa_put(ifa->ifpub);
2674 			ifa->ifpub = NULL;
2675 		}
2676 		spin_unlock_bh(&ifa->lock);
2677 		in6_ifa_put(ifa);
2678 		write_lock_bh(&idev->lock);
2679 	}
2680 #endif
2681 	while ((ifa = idev->addr_list) != NULL) {
2682 		idev->addr_list = ifa->if_next;
2683 		ifa->if_next = NULL;
2684 		ifa->dead = 1;
2685 		addrconf_del_timer(ifa);
2686 		write_unlock_bh(&idev->lock);
2687 
2688 		__ipv6_ifa_notify(RTM_DELADDR, ifa);
2689 		atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2690 		in6_ifa_put(ifa);
2691 
2692 		write_lock_bh(&idev->lock);
2693 	}
2694 	write_unlock_bh(&idev->lock);
2695 
2696 	/* Step 5: Discard multicast list */
2697 
2698 	if (how)
2699 		ipv6_mc_destroy_dev(idev);
2700 	else
2701 		ipv6_mc_down(idev);
2702 
2703 	idev->tstamp = jiffies;
2704 
2705 	/* Shot the device (if unregistered) */
2706 
2707 	if (how) {
2708 		addrconf_sysctl_unregister(idev);
2709 		neigh_parms_release(&nd_tbl, idev->nd_parms);
2710 		neigh_ifdown(&nd_tbl, dev);
2711 		in6_dev_put(idev);
2712 	}
2713 	return 0;
2714 }
2715 
2716 static void addrconf_rs_timer(unsigned long data)
2717 {
2718 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2719 
2720 	if (ifp->idev->cnf.forwarding)
2721 		goto out;
2722 
2723 	if (ifp->idev->if_flags & IF_RA_RCVD) {
2724 		/*
2725 		 *	Announcement received after solicitation
2726 		 *	was sent
2727 		 */
2728 		goto out;
2729 	}
2730 
2731 	spin_lock(&ifp->lock);
2732 	if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2733 		/* The wait after the last probe can be shorter */
2734 		addrconf_mod_timer(ifp, AC_RS,
2735 				   (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2736 				   ifp->idev->cnf.rtr_solicit_delay :
2737 				   ifp->idev->cnf.rtr_solicit_interval);
2738 		spin_unlock(&ifp->lock);
2739 
2740 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2741 	} else {
2742 		spin_unlock(&ifp->lock);
2743 		/*
2744 		 * Note: we do not support deprecated "all on-link"
2745 		 * assumption any longer.
2746 		 */
2747 		printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2748 		       ifp->idev->dev->name);
2749 	}
2750 
2751 out:
2752 	in6_ifa_put(ifp);
2753 }
2754 
2755 /*
2756  *	Duplicate Address Detection
2757  */
2758 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2759 {
2760 	unsigned long rand_num;
2761 	struct inet6_dev *idev = ifp->idev;
2762 
2763 	if (ifp->flags & IFA_F_OPTIMISTIC)
2764 		rand_num = 0;
2765 	else
2766 		rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2767 
2768 	ifp->probes = idev->cnf.dad_transmits;
2769 	addrconf_mod_timer(ifp, AC_DAD, rand_num);
2770 }
2771 
2772 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2773 {
2774 	struct inet6_dev *idev = ifp->idev;
2775 	struct net_device *dev = idev->dev;
2776 
2777 	addrconf_join_solict(dev, &ifp->addr);
2778 
2779 	net_srandom(ifp->addr.s6_addr32[3]);
2780 
2781 	read_lock_bh(&idev->lock);
2782 	if (ifp->dead)
2783 		goto out;
2784 	spin_lock_bh(&ifp->lock);
2785 
2786 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2787 	    idev->cnf.accept_dad < 1 ||
2788 	    !(ifp->flags&IFA_F_TENTATIVE) ||
2789 	    ifp->flags & IFA_F_NODAD) {
2790 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2791 		spin_unlock_bh(&ifp->lock);
2792 		read_unlock_bh(&idev->lock);
2793 
2794 		addrconf_dad_completed(ifp);
2795 		return;
2796 	}
2797 
2798 	if (!(idev->if_flags & IF_READY)) {
2799 		spin_unlock_bh(&ifp->lock);
2800 		read_unlock_bh(&idev->lock);
2801 		/*
2802 		 * If the defice is not ready:
2803 		 * - keep it tentative if it is a permanent address.
2804 		 * - otherwise, kill it.
2805 		 */
2806 		in6_ifa_hold(ifp);
2807 		addrconf_dad_stop(ifp);
2808 		return;
2809 	}
2810 
2811 	/*
2812 	 * Optimistic nodes can start receiving
2813 	 * Frames right away
2814 	 */
2815 	if(ifp->flags & IFA_F_OPTIMISTIC)
2816 		ip6_ins_rt(ifp->rt);
2817 
2818 	addrconf_dad_kick(ifp);
2819 	spin_unlock_bh(&ifp->lock);
2820 out:
2821 	read_unlock_bh(&idev->lock);
2822 }
2823 
2824 static void addrconf_dad_timer(unsigned long data)
2825 {
2826 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2827 	struct inet6_dev *idev = ifp->idev;
2828 	struct in6_addr mcaddr;
2829 
2830 	read_lock_bh(&idev->lock);
2831 	if (idev->dead) {
2832 		read_unlock_bh(&idev->lock);
2833 		goto out;
2834 	}
2835 	if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2836 		read_unlock_bh(&idev->lock);
2837 		addrconf_dad_failure(ifp);
2838 		return;
2839 	}
2840 	spin_lock_bh(&ifp->lock);
2841 	if (ifp->probes == 0) {
2842 		/*
2843 		 * DAD was successful
2844 		 */
2845 
2846 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2847 		spin_unlock_bh(&ifp->lock);
2848 		read_unlock_bh(&idev->lock);
2849 
2850 		addrconf_dad_completed(ifp);
2851 
2852 		goto out;
2853 	}
2854 
2855 	ifp->probes--;
2856 	addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2857 	spin_unlock_bh(&ifp->lock);
2858 	read_unlock_bh(&idev->lock);
2859 
2860 	/* send a neighbour solicitation for our addr */
2861 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2862 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2863 out:
2864 	in6_ifa_put(ifp);
2865 }
2866 
2867 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2868 {
2869 	struct net_device *	dev = ifp->idev->dev;
2870 
2871 	/*
2872 	 *	Configure the address for reception. Now it is valid.
2873 	 */
2874 
2875 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
2876 
2877 	/* If added prefix is link local and forwarding is off,
2878 	   start sending router solicitations.
2879 	 */
2880 
2881 	if (ifp->idev->cnf.forwarding == 0 &&
2882 	    ifp->idev->cnf.rtr_solicits > 0 &&
2883 	    (dev->flags&IFF_LOOPBACK) == 0 &&
2884 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2885 		/*
2886 		 *	If a host as already performed a random delay
2887 		 *	[...] as part of DAD [...] there is no need
2888 		 *	to delay again before sending the first RS
2889 		 */
2890 		ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2891 
2892 		spin_lock_bh(&ifp->lock);
2893 		ifp->probes = 1;
2894 		ifp->idev->if_flags |= IF_RS_SENT;
2895 		addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2896 		spin_unlock_bh(&ifp->lock);
2897 	}
2898 }
2899 
2900 static void addrconf_dad_run(struct inet6_dev *idev) {
2901 	struct inet6_ifaddr *ifp;
2902 
2903 	read_lock_bh(&idev->lock);
2904 	for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2905 		spin_lock_bh(&ifp->lock);
2906 		if (!(ifp->flags & IFA_F_TENTATIVE)) {
2907 			spin_unlock_bh(&ifp->lock);
2908 			continue;
2909 		}
2910 		spin_unlock_bh(&ifp->lock);
2911 		addrconf_dad_kick(ifp);
2912 	}
2913 	read_unlock_bh(&idev->lock);
2914 }
2915 
2916 #ifdef CONFIG_PROC_FS
2917 struct if6_iter_state {
2918 	struct seq_net_private p;
2919 	int bucket;
2920 };
2921 
2922 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2923 {
2924 	struct inet6_ifaddr *ifa = NULL;
2925 	struct if6_iter_state *state = seq->private;
2926 	struct net *net = seq_file_net(seq);
2927 
2928 	for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2929 		ifa = inet6_addr_lst[state->bucket];
2930 
2931 		while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2932 			ifa = ifa->lst_next;
2933 		if (ifa)
2934 			break;
2935 	}
2936 	return ifa;
2937 }
2938 
2939 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2940 {
2941 	struct if6_iter_state *state = seq->private;
2942 	struct net *net = seq_file_net(seq);
2943 
2944 	ifa = ifa->lst_next;
2945 try_again:
2946 	if (ifa) {
2947 		if (!net_eq(dev_net(ifa->idev->dev), net)) {
2948 			ifa = ifa->lst_next;
2949 			goto try_again;
2950 		}
2951 	}
2952 
2953 	if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2954 		ifa = inet6_addr_lst[state->bucket];
2955 		goto try_again;
2956 	}
2957 
2958 	return ifa;
2959 }
2960 
2961 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2962 {
2963 	struct inet6_ifaddr *ifa = if6_get_first(seq);
2964 
2965 	if (ifa)
2966 		while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2967 			--pos;
2968 	return pos ? NULL : ifa;
2969 }
2970 
2971 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2972 	__acquires(addrconf_hash_lock)
2973 {
2974 	read_lock_bh(&addrconf_hash_lock);
2975 	return if6_get_idx(seq, *pos);
2976 }
2977 
2978 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2979 {
2980 	struct inet6_ifaddr *ifa;
2981 
2982 	ifa = if6_get_next(seq, v);
2983 	++*pos;
2984 	return ifa;
2985 }
2986 
2987 static void if6_seq_stop(struct seq_file *seq, void *v)
2988 	__releases(addrconf_hash_lock)
2989 {
2990 	read_unlock_bh(&addrconf_hash_lock);
2991 }
2992 
2993 static int if6_seq_show(struct seq_file *seq, void *v)
2994 {
2995 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2996 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2997 		   &ifp->addr,
2998 		   ifp->idev->dev->ifindex,
2999 		   ifp->prefix_len,
3000 		   ifp->scope,
3001 		   ifp->flags,
3002 		   ifp->idev->dev->name);
3003 	return 0;
3004 }
3005 
3006 static const struct seq_operations if6_seq_ops = {
3007 	.start	= if6_seq_start,
3008 	.next	= if6_seq_next,
3009 	.show	= if6_seq_show,
3010 	.stop	= if6_seq_stop,
3011 };
3012 
3013 static int if6_seq_open(struct inode *inode, struct file *file)
3014 {
3015 	return seq_open_net(inode, file, &if6_seq_ops,
3016 			    sizeof(struct if6_iter_state));
3017 }
3018 
3019 static const struct file_operations if6_fops = {
3020 	.owner		= THIS_MODULE,
3021 	.open		= if6_seq_open,
3022 	.read		= seq_read,
3023 	.llseek		= seq_lseek,
3024 	.release	= seq_release_net,
3025 };
3026 
3027 static int if6_proc_net_init(struct net *net)
3028 {
3029 	if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3030 		return -ENOMEM;
3031 	return 0;
3032 }
3033 
3034 static void if6_proc_net_exit(struct net *net)
3035 {
3036        proc_net_remove(net, "if_inet6");
3037 }
3038 
3039 static struct pernet_operations if6_proc_net_ops = {
3040        .init = if6_proc_net_init,
3041        .exit = if6_proc_net_exit,
3042 };
3043 
3044 int __init if6_proc_init(void)
3045 {
3046 	return register_pernet_subsys(&if6_proc_net_ops);
3047 }
3048 
3049 void if6_proc_exit(void)
3050 {
3051 	unregister_pernet_subsys(&if6_proc_net_ops);
3052 }
3053 #endif	/* CONFIG_PROC_FS */
3054 
3055 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3056 /* Check if address is a home address configured on any interface. */
3057 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3058 {
3059 	int ret = 0;
3060 	struct inet6_ifaddr * ifp;
3061 	u8 hash = ipv6_addr_hash(addr);
3062 	read_lock_bh(&addrconf_hash_lock);
3063 	for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3064 		if (!net_eq(dev_net(ifp->idev->dev), net))
3065 			continue;
3066 		if (ipv6_addr_equal(&ifp->addr, addr) &&
3067 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3068 			ret = 1;
3069 			break;
3070 		}
3071 	}
3072 	read_unlock_bh(&addrconf_hash_lock);
3073 	return ret;
3074 }
3075 #endif
3076 
3077 /*
3078  *	Periodic address status verification
3079  */
3080 
3081 static void addrconf_verify(unsigned long foo)
3082 {
3083 	struct inet6_ifaddr *ifp;
3084 	unsigned long now, next;
3085 	int i;
3086 
3087 	spin_lock_bh(&addrconf_verify_lock);
3088 	now = jiffies;
3089 	next = now + ADDR_CHECK_FREQUENCY;
3090 
3091 	del_timer(&addr_chk_timer);
3092 
3093 	for (i=0; i < IN6_ADDR_HSIZE; i++) {
3094 
3095 restart:
3096 		read_lock(&addrconf_hash_lock);
3097 		for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3098 			unsigned long age;
3099 #ifdef CONFIG_IPV6_PRIVACY
3100 			unsigned long regen_advance;
3101 #endif
3102 
3103 			if (ifp->flags & IFA_F_PERMANENT)
3104 				continue;
3105 
3106 			spin_lock(&ifp->lock);
3107 			age = (now - ifp->tstamp) / HZ;
3108 
3109 #ifdef CONFIG_IPV6_PRIVACY
3110 			regen_advance = ifp->idev->cnf.regen_max_retry *
3111 					ifp->idev->cnf.dad_transmits *
3112 					ifp->idev->nd_parms->retrans_time / HZ;
3113 #endif
3114 
3115 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3116 			    age >= ifp->valid_lft) {
3117 				spin_unlock(&ifp->lock);
3118 				in6_ifa_hold(ifp);
3119 				read_unlock(&addrconf_hash_lock);
3120 				ipv6_del_addr(ifp);
3121 				goto restart;
3122 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3123 				spin_unlock(&ifp->lock);
3124 				continue;
3125 			} else if (age >= ifp->prefered_lft) {
3126 				/* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3127 				int deprecate = 0;
3128 
3129 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3130 					deprecate = 1;
3131 					ifp->flags |= IFA_F_DEPRECATED;
3132 				}
3133 
3134 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3135 					next = ifp->tstamp + ifp->valid_lft * HZ;
3136 
3137 				spin_unlock(&ifp->lock);
3138 
3139 				if (deprecate) {
3140 					in6_ifa_hold(ifp);
3141 					read_unlock(&addrconf_hash_lock);
3142 
3143 					ipv6_ifa_notify(0, ifp);
3144 					in6_ifa_put(ifp);
3145 					goto restart;
3146 				}
3147 #ifdef CONFIG_IPV6_PRIVACY
3148 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3149 				   !(ifp->flags&IFA_F_TENTATIVE)) {
3150 				if (age >= ifp->prefered_lft - regen_advance) {
3151 					struct inet6_ifaddr *ifpub = ifp->ifpub;
3152 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3153 						next = ifp->tstamp + ifp->prefered_lft * HZ;
3154 					if (!ifp->regen_count && ifpub) {
3155 						ifp->regen_count++;
3156 						in6_ifa_hold(ifp);
3157 						in6_ifa_hold(ifpub);
3158 						spin_unlock(&ifp->lock);
3159 						read_unlock(&addrconf_hash_lock);
3160 						spin_lock(&ifpub->lock);
3161 						ifpub->regen_count = 0;
3162 						spin_unlock(&ifpub->lock);
3163 						ipv6_create_tempaddr(ifpub, ifp);
3164 						in6_ifa_put(ifpub);
3165 						in6_ifa_put(ifp);
3166 						goto restart;
3167 					}
3168 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3169 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3170 				spin_unlock(&ifp->lock);
3171 #endif
3172 			} else {
3173 				/* ifp->prefered_lft <= ifp->valid_lft */
3174 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3175 					next = ifp->tstamp + ifp->prefered_lft * HZ;
3176 				spin_unlock(&ifp->lock);
3177 			}
3178 		}
3179 		read_unlock(&addrconf_hash_lock);
3180 	}
3181 
3182 	addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3183 	add_timer(&addr_chk_timer);
3184 	spin_unlock_bh(&addrconf_verify_lock);
3185 }
3186 
3187 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3188 {
3189 	struct in6_addr *pfx = NULL;
3190 
3191 	if (addr)
3192 		pfx = nla_data(addr);
3193 
3194 	if (local) {
3195 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3196 			pfx = NULL;
3197 		else
3198 			pfx = nla_data(local);
3199 	}
3200 
3201 	return pfx;
3202 }
3203 
3204 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3205 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
3206 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
3207 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
3208 };
3209 
3210 static int
3211 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3212 {
3213 	struct net *net = sock_net(skb->sk);
3214 	struct ifaddrmsg *ifm;
3215 	struct nlattr *tb[IFA_MAX+1];
3216 	struct in6_addr *pfx;
3217 	int err;
3218 
3219 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3220 	if (err < 0)
3221 		return err;
3222 
3223 	ifm = nlmsg_data(nlh);
3224 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3225 	if (pfx == NULL)
3226 		return -EINVAL;
3227 
3228 	return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3229 }
3230 
3231 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3232 			     u32 prefered_lft, u32 valid_lft)
3233 {
3234 	u32 flags;
3235 	clock_t expires;
3236 	unsigned long timeout;
3237 
3238 	if (!valid_lft || (prefered_lft > valid_lft))
3239 		return -EINVAL;
3240 
3241 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
3242 	if (addrconf_finite_timeout(timeout)) {
3243 		expires = jiffies_to_clock_t(timeout * HZ);
3244 		valid_lft = timeout;
3245 		flags = RTF_EXPIRES;
3246 	} else {
3247 		expires = 0;
3248 		flags = 0;
3249 		ifa_flags |= IFA_F_PERMANENT;
3250 	}
3251 
3252 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3253 	if (addrconf_finite_timeout(timeout)) {
3254 		if (timeout == 0)
3255 			ifa_flags |= IFA_F_DEPRECATED;
3256 		prefered_lft = timeout;
3257 	}
3258 
3259 	spin_lock_bh(&ifp->lock);
3260 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3261 	ifp->tstamp = jiffies;
3262 	ifp->valid_lft = valid_lft;
3263 	ifp->prefered_lft = prefered_lft;
3264 
3265 	spin_unlock_bh(&ifp->lock);
3266 	if (!(ifp->flags&IFA_F_TENTATIVE))
3267 		ipv6_ifa_notify(0, ifp);
3268 
3269 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3270 			      expires, flags);
3271 	addrconf_verify(0);
3272 
3273 	return 0;
3274 }
3275 
3276 static int
3277 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3278 {
3279 	struct net *net = sock_net(skb->sk);
3280 	struct ifaddrmsg *ifm;
3281 	struct nlattr *tb[IFA_MAX+1];
3282 	struct in6_addr *pfx;
3283 	struct inet6_ifaddr *ifa;
3284 	struct net_device *dev;
3285 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3286 	u8 ifa_flags;
3287 	int err;
3288 
3289 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3290 	if (err < 0)
3291 		return err;
3292 
3293 	ifm = nlmsg_data(nlh);
3294 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3295 	if (pfx == NULL)
3296 		return -EINVAL;
3297 
3298 	if (tb[IFA_CACHEINFO]) {
3299 		struct ifa_cacheinfo *ci;
3300 
3301 		ci = nla_data(tb[IFA_CACHEINFO]);
3302 		valid_lft = ci->ifa_valid;
3303 		preferred_lft = ci->ifa_prefered;
3304 	} else {
3305 		preferred_lft = INFINITY_LIFE_TIME;
3306 		valid_lft = INFINITY_LIFE_TIME;
3307 	}
3308 
3309 	dev =  __dev_get_by_index(net, ifm->ifa_index);
3310 	if (dev == NULL)
3311 		return -ENODEV;
3312 
3313 	/* We ignore other flags so far. */
3314 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3315 
3316 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3317 	if (ifa == NULL) {
3318 		/*
3319 		 * It would be best to check for !NLM_F_CREATE here but
3320 		 * userspace alreay relies on not having to provide this.
3321 		 */
3322 		return inet6_addr_add(net, ifm->ifa_index, pfx,
3323 				      ifm->ifa_prefixlen, ifa_flags,
3324 				      preferred_lft, valid_lft);
3325 	}
3326 
3327 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3328 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3329 		err = -EEXIST;
3330 	else
3331 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3332 
3333 	in6_ifa_put(ifa);
3334 
3335 	return err;
3336 }
3337 
3338 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3339 			  u8 scope, int ifindex)
3340 {
3341 	struct ifaddrmsg *ifm;
3342 
3343 	ifm = nlmsg_data(nlh);
3344 	ifm->ifa_family = AF_INET6;
3345 	ifm->ifa_prefixlen = prefixlen;
3346 	ifm->ifa_flags = flags;
3347 	ifm->ifa_scope = scope;
3348 	ifm->ifa_index = ifindex;
3349 }
3350 
3351 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3352 			 unsigned long tstamp, u32 preferred, u32 valid)
3353 {
3354 	struct ifa_cacheinfo ci;
3355 
3356 	ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3357 			+ TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3358 	ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3359 			+ TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3360 	ci.ifa_prefered = preferred;
3361 	ci.ifa_valid = valid;
3362 
3363 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3364 }
3365 
3366 static inline int rt_scope(int ifa_scope)
3367 {
3368 	if (ifa_scope & IFA_HOST)
3369 		return RT_SCOPE_HOST;
3370 	else if (ifa_scope & IFA_LINK)
3371 		return RT_SCOPE_LINK;
3372 	else if (ifa_scope & IFA_SITE)
3373 		return RT_SCOPE_SITE;
3374 	else
3375 		return RT_SCOPE_UNIVERSE;
3376 }
3377 
3378 static inline int inet6_ifaddr_msgsize(void)
3379 {
3380 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3381 	       + nla_total_size(16) /* IFA_ADDRESS */
3382 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3383 }
3384 
3385 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3386 			     u32 pid, u32 seq, int event, unsigned int flags)
3387 {
3388 	struct nlmsghdr  *nlh;
3389 	u32 preferred, valid;
3390 
3391 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3392 	if (nlh == NULL)
3393 		return -EMSGSIZE;
3394 
3395 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3396 		      ifa->idev->dev->ifindex);
3397 
3398 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3399 		preferred = ifa->prefered_lft;
3400 		valid = ifa->valid_lft;
3401 		if (preferred != INFINITY_LIFE_TIME) {
3402 			long tval = (jiffies - ifa->tstamp)/HZ;
3403 			preferred -= tval;
3404 			if (valid != INFINITY_LIFE_TIME)
3405 				valid -= tval;
3406 		}
3407 	} else {
3408 		preferred = INFINITY_LIFE_TIME;
3409 		valid = INFINITY_LIFE_TIME;
3410 	}
3411 
3412 	if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3413 	    put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3414 		nlmsg_cancel(skb, nlh);
3415 		return -EMSGSIZE;
3416 	}
3417 
3418 	return nlmsg_end(skb, nlh);
3419 }
3420 
3421 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3422 				u32 pid, u32 seq, int event, u16 flags)
3423 {
3424 	struct nlmsghdr  *nlh;
3425 	u8 scope = RT_SCOPE_UNIVERSE;
3426 	int ifindex = ifmca->idev->dev->ifindex;
3427 
3428 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3429 		scope = RT_SCOPE_SITE;
3430 
3431 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3432 	if (nlh == NULL)
3433 		return -EMSGSIZE;
3434 
3435 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3436 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3437 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3438 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3439 		nlmsg_cancel(skb, nlh);
3440 		return -EMSGSIZE;
3441 	}
3442 
3443 	return nlmsg_end(skb, nlh);
3444 }
3445 
3446 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3447 				u32 pid, u32 seq, int event, unsigned int flags)
3448 {
3449 	struct nlmsghdr  *nlh;
3450 	u8 scope = RT_SCOPE_UNIVERSE;
3451 	int ifindex = ifaca->aca_idev->dev->ifindex;
3452 
3453 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3454 		scope = RT_SCOPE_SITE;
3455 
3456 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3457 	if (nlh == NULL)
3458 		return -EMSGSIZE;
3459 
3460 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3461 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3462 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3463 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3464 		nlmsg_cancel(skb, nlh);
3465 		return -EMSGSIZE;
3466 	}
3467 
3468 	return nlmsg_end(skb, nlh);
3469 }
3470 
3471 enum addr_type_t
3472 {
3473 	UNICAST_ADDR,
3474 	MULTICAST_ADDR,
3475 	ANYCAST_ADDR,
3476 };
3477 
3478 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3479 			   enum addr_type_t type)
3480 {
3481 	int idx, ip_idx;
3482 	int s_idx, s_ip_idx;
3483 	int err = 1;
3484 	struct net_device *dev;
3485 	struct inet6_dev *idev = NULL;
3486 	struct inet6_ifaddr *ifa;
3487 	struct ifmcaddr6 *ifmca;
3488 	struct ifacaddr6 *ifaca;
3489 	struct net *net = sock_net(skb->sk);
3490 
3491 	s_idx = cb->args[0];
3492 	s_ip_idx = ip_idx = cb->args[1];
3493 
3494 	idx = 0;
3495 	for_each_netdev(net, dev) {
3496 		if (idx < s_idx)
3497 			goto cont;
3498 		if (idx > s_idx)
3499 			s_ip_idx = 0;
3500 		ip_idx = 0;
3501 		if ((idev = in6_dev_get(dev)) == NULL)
3502 			goto cont;
3503 		read_lock_bh(&idev->lock);
3504 		switch (type) {
3505 		case UNICAST_ADDR:
3506 			/* unicast address incl. temp addr */
3507 			for (ifa = idev->addr_list; ifa;
3508 			     ifa = ifa->if_next, ip_idx++) {
3509 				if (ip_idx < s_ip_idx)
3510 					continue;
3511 				err = inet6_fill_ifaddr(skb, ifa,
3512 							NETLINK_CB(cb->skb).pid,
3513 							cb->nlh->nlmsg_seq,
3514 							RTM_NEWADDR,
3515 							NLM_F_MULTI);
3516 			}
3517 			break;
3518 		case MULTICAST_ADDR:
3519 			/* multicast address */
3520 			for (ifmca = idev->mc_list; ifmca;
3521 			     ifmca = ifmca->next, ip_idx++) {
3522 				if (ip_idx < s_ip_idx)
3523 					continue;
3524 				err = inet6_fill_ifmcaddr(skb, ifmca,
3525 							  NETLINK_CB(cb->skb).pid,
3526 							  cb->nlh->nlmsg_seq,
3527 							  RTM_GETMULTICAST,
3528 							  NLM_F_MULTI);
3529 			}
3530 			break;
3531 		case ANYCAST_ADDR:
3532 			/* anycast address */
3533 			for (ifaca = idev->ac_list; ifaca;
3534 			     ifaca = ifaca->aca_next, ip_idx++) {
3535 				if (ip_idx < s_ip_idx)
3536 					continue;
3537 				err = inet6_fill_ifacaddr(skb, ifaca,
3538 							  NETLINK_CB(cb->skb).pid,
3539 							  cb->nlh->nlmsg_seq,
3540 							  RTM_GETANYCAST,
3541 							  NLM_F_MULTI);
3542 			}
3543 			break;
3544 		default:
3545 			break;
3546 		}
3547 		read_unlock_bh(&idev->lock);
3548 		in6_dev_put(idev);
3549 
3550 		if (err <= 0)
3551 			break;
3552 cont:
3553 		idx++;
3554 	}
3555 	cb->args[0] = idx;
3556 	cb->args[1] = ip_idx;
3557 	return skb->len;
3558 }
3559 
3560 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3561 {
3562 	enum addr_type_t type = UNICAST_ADDR;
3563 
3564 	return inet6_dump_addr(skb, cb, type);
3565 }
3566 
3567 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3568 {
3569 	enum addr_type_t type = MULTICAST_ADDR;
3570 
3571 	return inet6_dump_addr(skb, cb, type);
3572 }
3573 
3574 
3575 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3576 {
3577 	enum addr_type_t type = ANYCAST_ADDR;
3578 
3579 	return inet6_dump_addr(skb, cb, type);
3580 }
3581 
3582 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3583 			     void *arg)
3584 {
3585 	struct net *net = sock_net(in_skb->sk);
3586 	struct ifaddrmsg *ifm;
3587 	struct nlattr *tb[IFA_MAX+1];
3588 	struct in6_addr *addr = NULL;
3589 	struct net_device *dev = NULL;
3590 	struct inet6_ifaddr *ifa;
3591 	struct sk_buff *skb;
3592 	int err;
3593 
3594 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3595 	if (err < 0)
3596 		goto errout;
3597 
3598 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3599 	if (addr == NULL) {
3600 		err = -EINVAL;
3601 		goto errout;
3602 	}
3603 
3604 	ifm = nlmsg_data(nlh);
3605 	if (ifm->ifa_index)
3606 		dev = __dev_get_by_index(net, ifm->ifa_index);
3607 
3608 	if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3609 		err = -EADDRNOTAVAIL;
3610 		goto errout;
3611 	}
3612 
3613 	if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3614 		err = -ENOBUFS;
3615 		goto errout_ifa;
3616 	}
3617 
3618 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3619 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
3620 	if (err < 0) {
3621 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3622 		WARN_ON(err == -EMSGSIZE);
3623 		kfree_skb(skb);
3624 		goto errout_ifa;
3625 	}
3626 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3627 errout_ifa:
3628 	in6_ifa_put(ifa);
3629 errout:
3630 	return err;
3631 }
3632 
3633 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3634 {
3635 	struct sk_buff *skb;
3636 	struct net *net = dev_net(ifa->idev->dev);
3637 	int err = -ENOBUFS;
3638 
3639 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3640 	if (skb == NULL)
3641 		goto errout;
3642 
3643 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3644 	if (err < 0) {
3645 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3646 		WARN_ON(err == -EMSGSIZE);
3647 		kfree_skb(skb);
3648 		goto errout;
3649 	}
3650 	err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3651 errout:
3652 	if (err < 0)
3653 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3654 }
3655 
3656 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3657 				__s32 *array, int bytes)
3658 {
3659 	BUG_ON(bytes < (DEVCONF_MAX * 4));
3660 
3661 	memset(array, 0, bytes);
3662 	array[DEVCONF_FORWARDING] = cnf->forwarding;
3663 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3664 	array[DEVCONF_MTU6] = cnf->mtu6;
3665 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3666 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3667 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
3668 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3669 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3670 	array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3671 	array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3672 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3673 #ifdef CONFIG_IPV6_PRIVACY
3674 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3675 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3676 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3677 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3678 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3679 #endif
3680 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3681 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3682 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3683 #ifdef CONFIG_IPV6_ROUTER_PREF
3684 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3685 	array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3686 #ifdef CONFIG_IPV6_ROUTE_INFO
3687 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3688 #endif
3689 #endif
3690 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3691 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3692 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3693 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3694 #endif
3695 #ifdef CONFIG_IPV6_MROUTE
3696 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3697 #endif
3698 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3699 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3700 }
3701 
3702 static inline size_t inet6_if_nlmsg_size(void)
3703 {
3704 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3705 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3706 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3707 	       + nla_total_size(4) /* IFLA_MTU */
3708 	       + nla_total_size(4) /* IFLA_LINK */
3709 	       + nla_total_size( /* IFLA_PROTINFO */
3710 			nla_total_size(4) /* IFLA_INET6_FLAGS */
3711 			+ nla_total_size(sizeof(struct ifla_cacheinfo))
3712 			+ nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3713 			+ nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3714 			+ nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3715 		 );
3716 }
3717 
3718 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3719 				      int bytes)
3720 {
3721 	int i;
3722 	int pad = bytes - sizeof(u64) * items;
3723 	BUG_ON(pad < 0);
3724 
3725 	/* Use put_unaligned() because stats may not be aligned for u64. */
3726 	put_unaligned(items, &stats[0]);
3727 	for (i = 1; i < items; i++)
3728 		put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3729 
3730 	memset(&stats[items], 0, pad);
3731 }
3732 
3733 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3734 			     int bytes)
3735 {
3736 	switch(attrtype) {
3737 	case IFLA_INET6_STATS:
3738 		__snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3739 		break;
3740 	case IFLA_INET6_ICMP6STATS:
3741 		__snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3742 		break;
3743 	}
3744 }
3745 
3746 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3747 			     u32 pid, u32 seq, int event, unsigned int flags)
3748 {
3749 	struct net_device *dev = idev->dev;
3750 	struct nlattr *nla;
3751 	struct ifinfomsg *hdr;
3752 	struct nlmsghdr *nlh;
3753 	void *protoinfo;
3754 	struct ifla_cacheinfo ci;
3755 
3756 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3757 	if (nlh == NULL)
3758 		return -EMSGSIZE;
3759 
3760 	hdr = nlmsg_data(nlh);
3761 	hdr->ifi_family = AF_INET6;
3762 	hdr->__ifi_pad = 0;
3763 	hdr->ifi_type = dev->type;
3764 	hdr->ifi_index = dev->ifindex;
3765 	hdr->ifi_flags = dev_get_flags(dev);
3766 	hdr->ifi_change = 0;
3767 
3768 	NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3769 
3770 	if (dev->addr_len)
3771 		NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3772 
3773 	NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3774 	if (dev->ifindex != dev->iflink)
3775 		NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3776 
3777 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3778 	if (protoinfo == NULL)
3779 		goto nla_put_failure;
3780 
3781 	NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3782 
3783 	ci.max_reasm_len = IPV6_MAXPLEN;
3784 	ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3785 		    + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3786 	ci.reachable_time = idev->nd_parms->reachable_time;
3787 	ci.retrans_time = idev->nd_parms->retrans_time;
3788 	NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3789 
3790 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3791 	if (nla == NULL)
3792 		goto nla_put_failure;
3793 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3794 
3795 	/* XXX - MC not implemented */
3796 
3797 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3798 	if (nla == NULL)
3799 		goto nla_put_failure;
3800 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3801 
3802 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3803 	if (nla == NULL)
3804 		goto nla_put_failure;
3805 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3806 
3807 	nla_nest_end(skb, protoinfo);
3808 	return nlmsg_end(skb, nlh);
3809 
3810 nla_put_failure:
3811 	nlmsg_cancel(skb, nlh);
3812 	return -EMSGSIZE;
3813 }
3814 
3815 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3816 {
3817 	struct net *net = sock_net(skb->sk);
3818 	int idx, err;
3819 	int s_idx = cb->args[0];
3820 	struct net_device *dev;
3821 	struct inet6_dev *idev;
3822 
3823 	read_lock(&dev_base_lock);
3824 	idx = 0;
3825 	for_each_netdev(net, dev) {
3826 		if (idx < s_idx)
3827 			goto cont;
3828 		if ((idev = in6_dev_get(dev)) == NULL)
3829 			goto cont;
3830 		err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3831 				cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3832 		in6_dev_put(idev);
3833 		if (err <= 0)
3834 			break;
3835 cont:
3836 		idx++;
3837 	}
3838 	read_unlock(&dev_base_lock);
3839 	cb->args[0] = idx;
3840 
3841 	return skb->len;
3842 }
3843 
3844 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3845 {
3846 	struct sk_buff *skb;
3847 	struct net *net = dev_net(idev->dev);
3848 	int err = -ENOBUFS;
3849 
3850 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3851 	if (skb == NULL)
3852 		goto errout;
3853 
3854 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3855 	if (err < 0) {
3856 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3857 		WARN_ON(err == -EMSGSIZE);
3858 		kfree_skb(skb);
3859 		goto errout;
3860 	}
3861 	err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3862 errout:
3863 	if (err < 0)
3864 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3865 }
3866 
3867 static inline size_t inet6_prefix_nlmsg_size(void)
3868 {
3869 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
3870 	       + nla_total_size(sizeof(struct in6_addr))
3871 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
3872 }
3873 
3874 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3875 			     struct prefix_info *pinfo, u32 pid, u32 seq,
3876 			     int event, unsigned int flags)
3877 {
3878 	struct prefixmsg *pmsg;
3879 	struct nlmsghdr *nlh;
3880 	struct prefix_cacheinfo	ci;
3881 
3882 	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3883 	if (nlh == NULL)
3884 		return -EMSGSIZE;
3885 
3886 	pmsg = nlmsg_data(nlh);
3887 	pmsg->prefix_family = AF_INET6;
3888 	pmsg->prefix_pad1 = 0;
3889 	pmsg->prefix_pad2 = 0;
3890 	pmsg->prefix_ifindex = idev->dev->ifindex;
3891 	pmsg->prefix_len = pinfo->prefix_len;
3892 	pmsg->prefix_type = pinfo->type;
3893 	pmsg->prefix_pad3 = 0;
3894 	pmsg->prefix_flags = 0;
3895 	if (pinfo->onlink)
3896 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3897 	if (pinfo->autoconf)
3898 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3899 
3900 	NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3901 
3902 	ci.preferred_time = ntohl(pinfo->prefered);
3903 	ci.valid_time = ntohl(pinfo->valid);
3904 	NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3905 
3906 	return nlmsg_end(skb, nlh);
3907 
3908 nla_put_failure:
3909 	nlmsg_cancel(skb, nlh);
3910 	return -EMSGSIZE;
3911 }
3912 
3913 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3914 			 struct prefix_info *pinfo)
3915 {
3916 	struct sk_buff *skb;
3917 	struct net *net = dev_net(idev->dev);
3918 	int err = -ENOBUFS;
3919 
3920 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3921 	if (skb == NULL)
3922 		goto errout;
3923 
3924 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3925 	if (err < 0) {
3926 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3927 		WARN_ON(err == -EMSGSIZE);
3928 		kfree_skb(skb);
3929 		goto errout;
3930 	}
3931 	err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3932 errout:
3933 	if (err < 0)
3934 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3935 }
3936 
3937 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3938 {
3939 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3940 
3941 	switch (event) {
3942 	case RTM_NEWADDR:
3943 		/*
3944 		 * If the address was optimistic
3945 		 * we inserted the route at the start of
3946 		 * our DAD process, so we don't need
3947 		 * to do it again
3948 		 */
3949 		if (!(ifp->rt->rt6i_node))
3950 			ip6_ins_rt(ifp->rt);
3951 		if (ifp->idev->cnf.forwarding)
3952 			addrconf_join_anycast(ifp);
3953 		break;
3954 	case RTM_DELADDR:
3955 		if (ifp->idev->cnf.forwarding)
3956 			addrconf_leave_anycast(ifp);
3957 		addrconf_leave_solict(ifp->idev, &ifp->addr);
3958 		dst_hold(&ifp->rt->u.dst);
3959 		if (ip6_del_rt(ifp->rt))
3960 			dst_free(&ifp->rt->u.dst);
3961 		break;
3962 	}
3963 }
3964 
3965 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3966 {
3967 	rcu_read_lock_bh();
3968 	if (likely(ifp->idev->dead == 0))
3969 		__ipv6_ifa_notify(event, ifp);
3970 	rcu_read_unlock_bh();
3971 }
3972 
3973 #ifdef CONFIG_SYSCTL
3974 
3975 static
3976 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3977 			   void __user *buffer, size_t *lenp, loff_t *ppos)
3978 {
3979 	int *valp = ctl->data;
3980 	int val = *valp;
3981 	int ret;
3982 
3983 	ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3984 
3985 	if (write)
3986 		addrconf_fixup_forwarding(ctl, valp, val);
3987 	return ret;
3988 }
3989 
3990 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3991 					    void __user *oldval,
3992 					    size_t __user *oldlenp,
3993 					    void __user *newval, size_t newlen)
3994 {
3995 	int *valp = table->data;
3996 	int val = *valp;
3997 	int new;
3998 
3999 	if (!newval || !newlen)
4000 		return 0;
4001 	if (newlen != sizeof(int))
4002 		return -EINVAL;
4003 	if (get_user(new, (int __user *)newval))
4004 		return -EFAULT;
4005 	if (new == *valp)
4006 		return 0;
4007 	if (oldval && oldlenp) {
4008 		size_t len;
4009 		if (get_user(len, oldlenp))
4010 			return -EFAULT;
4011 		if (len) {
4012 			if (len > table->maxlen)
4013 				len = table->maxlen;
4014 			if (copy_to_user(oldval, valp, len))
4015 				return -EFAULT;
4016 			if (put_user(len, oldlenp))
4017 				return -EFAULT;
4018 		}
4019 	}
4020 
4021 	*valp = new;
4022 	addrconf_fixup_forwarding(table, valp, val);
4023 	return 1;
4024 }
4025 
4026 static struct addrconf_sysctl_table
4027 {
4028 	struct ctl_table_header *sysctl_header;
4029 	ctl_table addrconf_vars[DEVCONF_MAX+1];
4030 	char *dev_name;
4031 } addrconf_sysctl __read_mostly = {
4032 	.sysctl_header = NULL,
4033 	.addrconf_vars = {
4034 		{
4035 			.ctl_name	=	NET_IPV6_FORWARDING,
4036 			.procname	=	"forwarding",
4037 			.data		=	&ipv6_devconf.forwarding,
4038 			.maxlen		=	sizeof(int),
4039 			.mode		=	0644,
4040 			.proc_handler	=	addrconf_sysctl_forward,
4041 			.strategy	=	addrconf_sysctl_forward_strategy,
4042 		},
4043 		{
4044 			.ctl_name	=	NET_IPV6_HOP_LIMIT,
4045 			.procname	=	"hop_limit",
4046 			.data		=	&ipv6_devconf.hop_limit,
4047 			.maxlen		=	sizeof(int),
4048 			.mode		=	0644,
4049 			.proc_handler	=	proc_dointvec,
4050 		},
4051 		{
4052 			.ctl_name	=	NET_IPV6_MTU,
4053 			.procname	=	"mtu",
4054 			.data		=	&ipv6_devconf.mtu6,
4055 			.maxlen		=	sizeof(int),
4056 			.mode		=	0644,
4057 			.proc_handler	=	proc_dointvec,
4058 		},
4059 		{
4060 			.ctl_name	=	NET_IPV6_ACCEPT_RA,
4061 			.procname	=	"accept_ra",
4062 			.data		=	&ipv6_devconf.accept_ra,
4063 			.maxlen		=	sizeof(int),
4064 			.mode		=	0644,
4065 			.proc_handler	=	proc_dointvec,
4066 		},
4067 		{
4068 			.ctl_name	=	NET_IPV6_ACCEPT_REDIRECTS,
4069 			.procname	=	"accept_redirects",
4070 			.data		=	&ipv6_devconf.accept_redirects,
4071 			.maxlen		=	sizeof(int),
4072 			.mode		=	0644,
4073 			.proc_handler	=	proc_dointvec,
4074 		},
4075 		{
4076 			.ctl_name	=	NET_IPV6_AUTOCONF,
4077 			.procname	=	"autoconf",
4078 			.data		=	&ipv6_devconf.autoconf,
4079 			.maxlen		=	sizeof(int),
4080 			.mode		=	0644,
4081 			.proc_handler	=	proc_dointvec,
4082 		},
4083 		{
4084 			.ctl_name	=	NET_IPV6_DAD_TRANSMITS,
4085 			.procname	=	"dad_transmits",
4086 			.data		=	&ipv6_devconf.dad_transmits,
4087 			.maxlen		=	sizeof(int),
4088 			.mode		=	0644,
4089 			.proc_handler	=	proc_dointvec,
4090 		},
4091 		{
4092 			.ctl_name	=	NET_IPV6_RTR_SOLICITS,
4093 			.procname	=	"router_solicitations",
4094 			.data		=	&ipv6_devconf.rtr_solicits,
4095 			.maxlen		=	sizeof(int),
4096 			.mode		=	0644,
4097 			.proc_handler	=	proc_dointvec,
4098 		},
4099 		{
4100 			.ctl_name	=	NET_IPV6_RTR_SOLICIT_INTERVAL,
4101 			.procname	=	"router_solicitation_interval",
4102 			.data		=	&ipv6_devconf.rtr_solicit_interval,
4103 			.maxlen		=	sizeof(int),
4104 			.mode		=	0644,
4105 			.proc_handler	=	proc_dointvec_jiffies,
4106 			.strategy	=	sysctl_jiffies,
4107 		},
4108 		{
4109 			.ctl_name	=	NET_IPV6_RTR_SOLICIT_DELAY,
4110 			.procname	=	"router_solicitation_delay",
4111 			.data		=	&ipv6_devconf.rtr_solicit_delay,
4112 			.maxlen		=	sizeof(int),
4113 			.mode		=	0644,
4114 			.proc_handler	=	proc_dointvec_jiffies,
4115 			.strategy	=	sysctl_jiffies,
4116 		},
4117 		{
4118 			.ctl_name	=	NET_IPV6_FORCE_MLD_VERSION,
4119 			.procname	=	"force_mld_version",
4120 			.data		=	&ipv6_devconf.force_mld_version,
4121 			.maxlen		=	sizeof(int),
4122 			.mode		=	0644,
4123 			.proc_handler	=	proc_dointvec,
4124 		},
4125 #ifdef CONFIG_IPV6_PRIVACY
4126 		{
4127 			.ctl_name	=	NET_IPV6_USE_TEMPADDR,
4128 			.procname	=	"use_tempaddr",
4129 			.data		=	&ipv6_devconf.use_tempaddr,
4130 			.maxlen		=	sizeof(int),
4131 			.mode		=	0644,
4132 			.proc_handler	=	proc_dointvec,
4133 		},
4134 		{
4135 			.ctl_name	=	NET_IPV6_TEMP_VALID_LFT,
4136 			.procname	=	"temp_valid_lft",
4137 			.data		=	&ipv6_devconf.temp_valid_lft,
4138 			.maxlen		=	sizeof(int),
4139 			.mode		=	0644,
4140 			.proc_handler	=	proc_dointvec,
4141 		},
4142 		{
4143 			.ctl_name	=	NET_IPV6_TEMP_PREFERED_LFT,
4144 			.procname	=	"temp_prefered_lft",
4145 			.data		=	&ipv6_devconf.temp_prefered_lft,
4146 			.maxlen		=	sizeof(int),
4147 			.mode		=	0644,
4148 			.proc_handler	=	proc_dointvec,
4149 		},
4150 		{
4151 			.ctl_name	=	NET_IPV6_REGEN_MAX_RETRY,
4152 			.procname	=	"regen_max_retry",
4153 			.data		=	&ipv6_devconf.regen_max_retry,
4154 			.maxlen		=	sizeof(int),
4155 			.mode		=	0644,
4156 			.proc_handler	=	proc_dointvec,
4157 		},
4158 		{
4159 			.ctl_name	=	NET_IPV6_MAX_DESYNC_FACTOR,
4160 			.procname	=	"max_desync_factor",
4161 			.data		=	&ipv6_devconf.max_desync_factor,
4162 			.maxlen		=	sizeof(int),
4163 			.mode		=	0644,
4164 			.proc_handler	=	proc_dointvec,
4165 		},
4166 #endif
4167 		{
4168 			.ctl_name	=	NET_IPV6_MAX_ADDRESSES,
4169 			.procname	=	"max_addresses",
4170 			.data		=	&ipv6_devconf.max_addresses,
4171 			.maxlen		=	sizeof(int),
4172 			.mode		=	0644,
4173 			.proc_handler	=	proc_dointvec,
4174 		},
4175 		{
4176 			.ctl_name	=	NET_IPV6_ACCEPT_RA_DEFRTR,
4177 			.procname	=	"accept_ra_defrtr",
4178 			.data		=	&ipv6_devconf.accept_ra_defrtr,
4179 			.maxlen		=	sizeof(int),
4180 			.mode		=	0644,
4181 			.proc_handler	=	proc_dointvec,
4182 		},
4183 		{
4184 			.ctl_name	=	NET_IPV6_ACCEPT_RA_PINFO,
4185 			.procname	=	"accept_ra_pinfo",
4186 			.data		=	&ipv6_devconf.accept_ra_pinfo,
4187 			.maxlen		=	sizeof(int),
4188 			.mode		=	0644,
4189 			.proc_handler	=	proc_dointvec,
4190 		},
4191 #ifdef CONFIG_IPV6_ROUTER_PREF
4192 		{
4193 			.ctl_name	=	NET_IPV6_ACCEPT_RA_RTR_PREF,
4194 			.procname	=	"accept_ra_rtr_pref",
4195 			.data		=	&ipv6_devconf.accept_ra_rtr_pref,
4196 			.maxlen		=	sizeof(int),
4197 			.mode		=	0644,
4198 			.proc_handler	=	proc_dointvec,
4199 		},
4200 		{
4201 			.ctl_name	=	NET_IPV6_RTR_PROBE_INTERVAL,
4202 			.procname	=	"router_probe_interval",
4203 			.data		=	&ipv6_devconf.rtr_probe_interval,
4204 			.maxlen		=	sizeof(int),
4205 			.mode		=	0644,
4206 			.proc_handler	=	proc_dointvec_jiffies,
4207 			.strategy	=	sysctl_jiffies,
4208 		},
4209 #ifdef CONFIG_IPV6_ROUTE_INFO
4210 		{
4211 			.ctl_name	=	NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4212 			.procname	=	"accept_ra_rt_info_max_plen",
4213 			.data		=	&ipv6_devconf.accept_ra_rt_info_max_plen,
4214 			.maxlen		=	sizeof(int),
4215 			.mode		=	0644,
4216 			.proc_handler	=	proc_dointvec,
4217 		},
4218 #endif
4219 #endif
4220 		{
4221 			.ctl_name	=	NET_IPV6_PROXY_NDP,
4222 			.procname	=	"proxy_ndp",
4223 			.data		=	&ipv6_devconf.proxy_ndp,
4224 			.maxlen		=	sizeof(int),
4225 			.mode		=	0644,
4226 			.proc_handler	=	proc_dointvec,
4227 		},
4228 		{
4229 			.ctl_name	=	NET_IPV6_ACCEPT_SOURCE_ROUTE,
4230 			.procname	=	"accept_source_route",
4231 			.data		=	&ipv6_devconf.accept_source_route,
4232 			.maxlen		=	sizeof(int),
4233 			.mode		=	0644,
4234 			.proc_handler	=	proc_dointvec,
4235 		},
4236 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4237 		{
4238 			.ctl_name	=	CTL_UNNUMBERED,
4239 			.procname       =       "optimistic_dad",
4240 			.data           =       &ipv6_devconf.optimistic_dad,
4241 			.maxlen         =       sizeof(int),
4242 			.mode           =       0644,
4243 			.proc_handler   =       proc_dointvec,
4244 
4245 		},
4246 #endif
4247 #ifdef CONFIG_IPV6_MROUTE
4248 		{
4249 			.ctl_name	=	CTL_UNNUMBERED,
4250 			.procname	=	"mc_forwarding",
4251 			.data		=	&ipv6_devconf.mc_forwarding,
4252 			.maxlen		=	sizeof(int),
4253 			.mode		=	0444,
4254 			.proc_handler	=	proc_dointvec,
4255 		},
4256 #endif
4257 		{
4258 			.ctl_name	=	CTL_UNNUMBERED,
4259 			.procname	=	"disable_ipv6",
4260 			.data		=	&ipv6_devconf.disable_ipv6,
4261 			.maxlen		=	sizeof(int),
4262 			.mode		=	0644,
4263 			.proc_handler	=	proc_dointvec,
4264 		},
4265 		{
4266 			.ctl_name	=	CTL_UNNUMBERED,
4267 			.procname	=	"accept_dad",
4268 			.data		=	&ipv6_devconf.accept_dad,
4269 			.maxlen		=	sizeof(int),
4270 			.mode		=	0644,
4271 			.proc_handler	=	proc_dointvec,
4272 		},
4273 		{
4274 			.ctl_name	=	0,	/* sentinel */
4275 		}
4276 	},
4277 };
4278 
4279 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4280 		int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4281 {
4282 	int i;
4283 	struct addrconf_sysctl_table *t;
4284 
4285 #define ADDRCONF_CTL_PATH_DEV	3
4286 
4287 	struct ctl_path addrconf_ctl_path[] = {
4288 		{ .procname = "net", .ctl_name = CTL_NET, },
4289 		{ .procname = "ipv6", .ctl_name = NET_IPV6, },
4290 		{ .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4291 		{ /* to be set */ },
4292 		{ },
4293 	};
4294 
4295 
4296 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4297 	if (t == NULL)
4298 		goto out;
4299 
4300 	for (i=0; t->addrconf_vars[i].data; i++) {
4301 		t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4302 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4303 		t->addrconf_vars[i].extra2 = net;
4304 	}
4305 
4306 	/*
4307 	 * Make a copy of dev_name, because '.procname' is regarded as const
4308 	 * by sysctl and we wouldn't want anyone to change it under our feet
4309 	 * (see SIOCSIFNAME).
4310 	 */
4311 	t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4312 	if (!t->dev_name)
4313 		goto free;
4314 
4315 	addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4316 	addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4317 
4318 	t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4319 			t->addrconf_vars);
4320 	if (t->sysctl_header == NULL)
4321 		goto free_procname;
4322 
4323 	p->sysctl = t;
4324 	return 0;
4325 
4326 free_procname:
4327 	kfree(t->dev_name);
4328 free:
4329 	kfree(t);
4330 out:
4331 	return -ENOBUFS;
4332 }
4333 
4334 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4335 {
4336 	struct addrconf_sysctl_table *t;
4337 
4338 	if (p->sysctl == NULL)
4339 		return;
4340 
4341 	t = p->sysctl;
4342 	p->sysctl = NULL;
4343 	unregister_sysctl_table(t->sysctl_header);
4344 	kfree(t->dev_name);
4345 	kfree(t);
4346 }
4347 
4348 static void addrconf_sysctl_register(struct inet6_dev *idev)
4349 {
4350 	neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4351 			      NET_IPV6_NEIGH, "ipv6",
4352 			      &ndisc_ifinfo_sysctl_change,
4353 			      ndisc_ifinfo_sysctl_strategy);
4354 	__addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4355 			idev->dev->ifindex, idev, &idev->cnf);
4356 }
4357 
4358 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4359 {
4360 	__addrconf_sysctl_unregister(&idev->cnf);
4361 	neigh_sysctl_unregister(idev->nd_parms);
4362 }
4363 
4364 
4365 #endif
4366 
4367 static int addrconf_init_net(struct net *net)
4368 {
4369 	int err;
4370 	struct ipv6_devconf *all, *dflt;
4371 
4372 	err = -ENOMEM;
4373 	all = &ipv6_devconf;
4374 	dflt = &ipv6_devconf_dflt;
4375 
4376 	if (net != &init_net) {
4377 		all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4378 		if (all == NULL)
4379 			goto err_alloc_all;
4380 
4381 		dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4382 		if (dflt == NULL)
4383 			goto err_alloc_dflt;
4384 	}
4385 
4386 	net->ipv6.devconf_all = all;
4387 	net->ipv6.devconf_dflt = dflt;
4388 
4389 #ifdef CONFIG_SYSCTL
4390 	err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4391 			NULL, all);
4392 	if (err < 0)
4393 		goto err_reg_all;
4394 
4395 	err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4396 			NULL, dflt);
4397 	if (err < 0)
4398 		goto err_reg_dflt;
4399 #endif
4400 	return 0;
4401 
4402 #ifdef CONFIG_SYSCTL
4403 err_reg_dflt:
4404 	__addrconf_sysctl_unregister(all);
4405 err_reg_all:
4406 	kfree(dflt);
4407 #endif
4408 err_alloc_dflt:
4409 	kfree(all);
4410 err_alloc_all:
4411 	return err;
4412 }
4413 
4414 static void addrconf_exit_net(struct net *net)
4415 {
4416 #ifdef CONFIG_SYSCTL
4417 	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4418 	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
4419 #endif
4420 	if (net != &init_net) {
4421 		kfree(net->ipv6.devconf_dflt);
4422 		kfree(net->ipv6.devconf_all);
4423 	}
4424 }
4425 
4426 static struct pernet_operations addrconf_ops = {
4427 	.init = addrconf_init_net,
4428 	.exit = addrconf_exit_net,
4429 };
4430 
4431 /*
4432  *      Device notifier
4433  */
4434 
4435 int register_inet6addr_notifier(struct notifier_block *nb)
4436 {
4437 	return atomic_notifier_chain_register(&inet6addr_chain, nb);
4438 }
4439 
4440 EXPORT_SYMBOL(register_inet6addr_notifier);
4441 
4442 int unregister_inet6addr_notifier(struct notifier_block *nb)
4443 {
4444 	return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4445 }
4446 
4447 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4448 
4449 static void addrconf_net_exit(struct net *net)
4450 {
4451 	struct net_device *dev;
4452 
4453 	rtnl_lock();
4454 	/* clean dev list */
4455 	for_each_netdev(net, dev) {
4456 		if (__in6_dev_get(dev) == NULL)
4457 			continue;
4458 		addrconf_ifdown(dev, 1);
4459 	}
4460 	addrconf_ifdown(net->loopback_dev, 2);
4461 	rtnl_unlock();
4462 }
4463 
4464 static struct pernet_operations addrconf_net_ops = {
4465 	.exit = addrconf_net_exit,
4466 };
4467 
4468 /*
4469  *	Init / cleanup code
4470  */
4471 
4472 int __init addrconf_init(void)
4473 {
4474 	int err;
4475 
4476 	if ((err = ipv6_addr_label_init()) < 0) {
4477 		printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4478 			err);
4479 		return err;
4480 	}
4481 
4482 	register_pernet_subsys(&addrconf_ops);
4483 
4484 	/* The addrconf netdev notifier requires that loopback_dev
4485 	 * has it's ipv6 private information allocated and setup
4486 	 * before it can bring up and give link-local addresses
4487 	 * to other devices which are up.
4488 	 *
4489 	 * Unfortunately, loopback_dev is not necessarily the first
4490 	 * entry in the global dev_base list of net devices.  In fact,
4491 	 * it is likely to be the very last entry on that list.
4492 	 * So this causes the notifier registry below to try and
4493 	 * give link-local addresses to all devices besides loopback_dev
4494 	 * first, then loopback_dev, which cases all the non-loopback_dev
4495 	 * devices to fail to get a link-local address.
4496 	 *
4497 	 * So, as a temporary fix, allocate the ipv6 structure for
4498 	 * loopback_dev first by hand.
4499 	 * Longer term, all of the dependencies ipv6 has upon the loopback
4500 	 * device and it being up should be removed.
4501 	 */
4502 	rtnl_lock();
4503 	if (!ipv6_add_dev(init_net.loopback_dev))
4504 		err = -ENOMEM;
4505 	rtnl_unlock();
4506 	if (err)
4507 		goto errlo;
4508 
4509 	err = register_pernet_device(&addrconf_net_ops);
4510 	if (err)
4511 		return err;
4512 
4513 	register_netdevice_notifier(&ipv6_dev_notf);
4514 
4515 	addrconf_verify(0);
4516 
4517 	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4518 	if (err < 0)
4519 		goto errout;
4520 
4521 	/* Only the first call to __rtnl_register can fail */
4522 	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4523 	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4524 	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4525 	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4526 	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4527 
4528 	ipv6_addr_label_rtnl_register();
4529 
4530 	return 0;
4531 errout:
4532 	unregister_netdevice_notifier(&ipv6_dev_notf);
4533 errlo:
4534 	unregister_pernet_subsys(&addrconf_ops);
4535 
4536 	return err;
4537 }
4538 
4539 void addrconf_cleanup(void)
4540 {
4541 	struct inet6_ifaddr *ifa;
4542 	int i;
4543 
4544 	unregister_netdevice_notifier(&ipv6_dev_notf);
4545 	unregister_pernet_device(&addrconf_net_ops);
4546 
4547 	unregister_pernet_subsys(&addrconf_ops);
4548 
4549 	rtnl_lock();
4550 
4551 	/*
4552 	 *	Check hash table.
4553 	 */
4554 	write_lock_bh(&addrconf_hash_lock);
4555 	for (i=0; i < IN6_ADDR_HSIZE; i++) {
4556 		for (ifa=inet6_addr_lst[i]; ifa; ) {
4557 			struct inet6_ifaddr *bifa;
4558 
4559 			bifa = ifa;
4560 			ifa = ifa->lst_next;
4561 			printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4562 			/* Do not free it; something is wrong.
4563 			   Now we can investigate it with debugger.
4564 			 */
4565 		}
4566 	}
4567 	write_unlock_bh(&addrconf_hash_lock);
4568 
4569 	del_timer(&addr_chk_timer);
4570 	rtnl_unlock();
4571 
4572 	unregister_pernet_subsys(&addrconf_net_ops);
4573 }
4574