xref: /linux/net/ipv6/addrconf.c (revision c4ee0af3fa0dc65f690fc908f02b8355f9576ea0)
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 #define pr_fmt(fmt) "IPv6: " fmt
42 
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67 
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71 
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
74 #include <net/ipv6.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
79 #include <net/tcp.h>
80 #include <net/ip.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
89 
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
93 
94 /* Set to 3 to get tracing... */
95 #define ACONF_DEBUG 2
96 
97 #if ACONF_DEBUG >= 3
98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
99 #else
100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
101 #endif
102 
103 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
104 
105 static inline u32 cstamp_delta(unsigned long cstamp)
106 {
107 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
108 }
109 
110 #ifdef CONFIG_SYSCTL
111 static void addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
113 #else
114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
115 {
116 }
117 
118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
119 {
120 }
121 #endif
122 
123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
125 static void ipv6_regen_rndid(unsigned long data);
126 
127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
128 static int ipv6_count_addresses(struct inet6_dev *idev);
129 
130 /*
131  *	Configured unicast address hash table
132  */
133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
134 static DEFINE_SPINLOCK(addrconf_hash_lock);
135 
136 static void addrconf_verify(unsigned long);
137 
138 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
139 static DEFINE_SPINLOCK(addrconf_verify_lock);
140 
141 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
142 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
143 
144 static void addrconf_type_change(struct net_device *dev,
145 				 unsigned long event);
146 static int addrconf_ifdown(struct net_device *dev, int how);
147 
148 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
149 						  int plen,
150 						  const struct net_device *dev,
151 						  u32 flags, u32 noflags);
152 
153 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
154 static void addrconf_dad_timer(unsigned long data);
155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
156 static void addrconf_dad_run(struct inet6_dev *idev);
157 static void addrconf_rs_timer(unsigned long data);
158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
159 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
160 
161 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
162 				struct prefix_info *pinfo);
163 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
164 			       struct net_device *dev);
165 
166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
167 	.forwarding		= 0,
168 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
169 	.mtu6			= IPV6_MIN_MTU,
170 	.accept_ra		= 1,
171 	.accept_redirects	= 1,
172 	.autoconf		= 1,
173 	.force_mld_version	= 0,
174 	.mldv1_unsolicited_report_interval = 10 * HZ,
175 	.mldv2_unsolicited_report_interval = HZ,
176 	.dad_transmits		= 1,
177 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
178 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
179 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
180 	.use_tempaddr 		= 0,
181 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
182 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
183 	.regen_max_retry	= REGEN_MAX_RETRY,
184 	.max_desync_factor	= MAX_DESYNC_FACTOR,
185 	.max_addresses		= IPV6_MAX_ADDRESSES,
186 	.accept_ra_defrtr	= 1,
187 	.accept_ra_pinfo	= 1,
188 #ifdef CONFIG_IPV6_ROUTER_PREF
189 	.accept_ra_rtr_pref	= 1,
190 	.rtr_probe_interval	= 60 * HZ,
191 #ifdef CONFIG_IPV6_ROUTE_INFO
192 	.accept_ra_rt_info_max_plen = 0,
193 #endif
194 #endif
195 	.proxy_ndp		= 0,
196 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
197 	.disable_ipv6		= 0,
198 	.accept_dad		= 1,
199 	.suppress_frag_ndisc	= 1,
200 };
201 
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 	.forwarding		= 0,
204 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
205 	.mtu6			= IPV6_MIN_MTU,
206 	.accept_ra		= 1,
207 	.accept_redirects	= 1,
208 	.autoconf		= 1,
209 	.force_mld_version	= 0,
210 	.mldv1_unsolicited_report_interval = 10 * HZ,
211 	.mldv2_unsolicited_report_interval = HZ,
212 	.dad_transmits		= 1,
213 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
214 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
215 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
216 	.use_tempaddr		= 0,
217 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
218 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
219 	.regen_max_retry	= REGEN_MAX_RETRY,
220 	.max_desync_factor	= MAX_DESYNC_FACTOR,
221 	.max_addresses		= IPV6_MAX_ADDRESSES,
222 	.accept_ra_defrtr	= 1,
223 	.accept_ra_pinfo	= 1,
224 #ifdef CONFIG_IPV6_ROUTER_PREF
225 	.accept_ra_rtr_pref	= 1,
226 	.rtr_probe_interval	= 60 * HZ,
227 #ifdef CONFIG_IPV6_ROUTE_INFO
228 	.accept_ra_rt_info_max_plen = 0,
229 #endif
230 #endif
231 	.proxy_ndp		= 0,
232 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
233 	.disable_ipv6		= 0,
234 	.accept_dad		= 1,
235 	.suppress_frag_ndisc	= 1,
236 };
237 
238 /* Check if a valid qdisc is available */
239 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
240 {
241 	return !qdisc_tx_is_noop(dev);
242 }
243 
244 static void addrconf_del_rs_timer(struct inet6_dev *idev)
245 {
246 	if (del_timer(&idev->rs_timer))
247 		__in6_dev_put(idev);
248 }
249 
250 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
251 {
252 	if (del_timer(&ifp->dad_timer))
253 		__in6_ifa_put(ifp);
254 }
255 
256 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
257 				  unsigned long when)
258 {
259 	if (!timer_pending(&idev->rs_timer))
260 		in6_dev_hold(idev);
261 	mod_timer(&idev->rs_timer, jiffies + when);
262 }
263 
264 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
265 				   unsigned long when)
266 {
267 	if (!timer_pending(&ifp->dad_timer))
268 		in6_ifa_hold(ifp);
269 	mod_timer(&ifp->dad_timer, jiffies + when);
270 }
271 
272 static int snmp6_alloc_dev(struct inet6_dev *idev)
273 {
274 	int i;
275 
276 	if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
277 			  sizeof(struct ipstats_mib),
278 			  __alignof__(struct ipstats_mib)) < 0)
279 		goto err_ip;
280 
281 	for_each_possible_cpu(i) {
282 		struct ipstats_mib *addrconf_stats;
283 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
284 		u64_stats_init(&addrconf_stats->syncp);
285 #if SNMP_ARRAY_SZ == 2
286 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
287 		u64_stats_init(&addrconf_stats->syncp);
288 #endif
289 	}
290 
291 
292 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
293 					GFP_KERNEL);
294 	if (!idev->stats.icmpv6dev)
295 		goto err_icmp;
296 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
297 					   GFP_KERNEL);
298 	if (!idev->stats.icmpv6msgdev)
299 		goto err_icmpmsg;
300 
301 	return 0;
302 
303 err_icmpmsg:
304 	kfree(idev->stats.icmpv6dev);
305 err_icmp:
306 	snmp_mib_free((void __percpu **)idev->stats.ipv6);
307 err_ip:
308 	return -ENOMEM;
309 }
310 
311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
312 {
313 	struct inet6_dev *ndev;
314 
315 	ASSERT_RTNL();
316 
317 	if (dev->mtu < IPV6_MIN_MTU)
318 		return NULL;
319 
320 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
321 
322 	if (ndev == NULL)
323 		return NULL;
324 
325 	rwlock_init(&ndev->lock);
326 	ndev->dev = dev;
327 	INIT_LIST_HEAD(&ndev->addr_list);
328 	setup_timer(&ndev->rs_timer, addrconf_rs_timer,
329 		    (unsigned long)ndev);
330 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
331 	ndev->cnf.mtu6 = dev->mtu;
332 	ndev->cnf.sysctl = NULL;
333 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
334 	if (ndev->nd_parms == NULL) {
335 		kfree(ndev);
336 		return NULL;
337 	}
338 	if (ndev->cnf.forwarding)
339 		dev_disable_lro(dev);
340 	/* We refer to the device */
341 	dev_hold(dev);
342 
343 	if (snmp6_alloc_dev(ndev) < 0) {
344 		ADBG(KERN_WARNING
345 			"%s: cannot allocate memory for statistics; dev=%s.\n",
346 			__func__, dev->name);
347 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
348 		dev_put(dev);
349 		kfree(ndev);
350 		return NULL;
351 	}
352 
353 	if (snmp6_register_dev(ndev) < 0) {
354 		ADBG(KERN_WARNING
355 			"%s: cannot create /proc/net/dev_snmp6/%s\n",
356 			__func__, dev->name);
357 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
358 		ndev->dead = 1;
359 		in6_dev_finish_destroy(ndev);
360 		return NULL;
361 	}
362 
363 	/* One reference from device.  We must do this before
364 	 * we invoke __ipv6_regen_rndid().
365 	 */
366 	in6_dev_hold(ndev);
367 
368 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
369 		ndev->cnf.accept_dad = -1;
370 
371 #if IS_ENABLED(CONFIG_IPV6_SIT)
372 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
373 		pr_info("%s: Disabled Multicast RS\n", dev->name);
374 		ndev->cnf.rtr_solicits = 0;
375 	}
376 #endif
377 
378 	INIT_LIST_HEAD(&ndev->tempaddr_list);
379 	setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
380 	if ((dev->flags&IFF_LOOPBACK) ||
381 	    dev->type == ARPHRD_TUNNEL ||
382 	    dev->type == ARPHRD_TUNNEL6 ||
383 	    dev->type == ARPHRD_SIT ||
384 	    dev->type == ARPHRD_NONE) {
385 		ndev->cnf.use_tempaddr = -1;
386 	} else {
387 		in6_dev_hold(ndev);
388 		ipv6_regen_rndid((unsigned long) ndev);
389 	}
390 
391 	ndev->token = in6addr_any;
392 
393 	if (netif_running(dev) && addrconf_qdisc_ok(dev))
394 		ndev->if_flags |= IF_READY;
395 
396 	ipv6_mc_init_dev(ndev);
397 	ndev->tstamp = jiffies;
398 	addrconf_sysctl_register(ndev);
399 	/* protected by rtnl_lock */
400 	rcu_assign_pointer(dev->ip6_ptr, ndev);
401 
402 	/* Join interface-local all-node multicast group */
403 	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
404 
405 	/* Join all-node multicast group */
406 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
407 
408 	/* Join all-router multicast group if forwarding is set */
409 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
410 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
411 
412 	return ndev;
413 }
414 
415 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
416 {
417 	struct inet6_dev *idev;
418 
419 	ASSERT_RTNL();
420 
421 	idev = __in6_dev_get(dev);
422 	if (!idev) {
423 		idev = ipv6_add_dev(dev);
424 		if (!idev)
425 			return NULL;
426 	}
427 
428 	if (dev->flags&IFF_UP)
429 		ipv6_mc_up(idev);
430 	return idev;
431 }
432 
433 static int inet6_netconf_msgsize_devconf(int type)
434 {
435 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
436 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
437 
438 	/* type -1 is used for ALL */
439 	if (type == -1 || type == NETCONFA_FORWARDING)
440 		size += nla_total_size(4);
441 #ifdef CONFIG_IPV6_MROUTE
442 	if (type == -1 || type == NETCONFA_MC_FORWARDING)
443 		size += nla_total_size(4);
444 #endif
445 
446 	return size;
447 }
448 
449 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
450 				      struct ipv6_devconf *devconf, u32 portid,
451 				      u32 seq, int event, unsigned int flags,
452 				      int type)
453 {
454 	struct nlmsghdr  *nlh;
455 	struct netconfmsg *ncm;
456 
457 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
458 			flags);
459 	if (nlh == NULL)
460 		return -EMSGSIZE;
461 
462 	ncm = nlmsg_data(nlh);
463 	ncm->ncm_family = AF_INET6;
464 
465 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
466 		goto nla_put_failure;
467 
468 	/* type -1 is used for ALL */
469 	if ((type == -1 || type == NETCONFA_FORWARDING) &&
470 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
471 		goto nla_put_failure;
472 #ifdef CONFIG_IPV6_MROUTE
473 	if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
474 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
475 			devconf->mc_forwarding) < 0)
476 		goto nla_put_failure;
477 #endif
478 	return nlmsg_end(skb, nlh);
479 
480 nla_put_failure:
481 	nlmsg_cancel(skb, nlh);
482 	return -EMSGSIZE;
483 }
484 
485 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
486 				  struct ipv6_devconf *devconf)
487 {
488 	struct sk_buff *skb;
489 	int err = -ENOBUFS;
490 
491 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
492 	if (skb == NULL)
493 		goto errout;
494 
495 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
496 					 RTM_NEWNETCONF, 0, type);
497 	if (err < 0) {
498 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
499 		WARN_ON(err == -EMSGSIZE);
500 		kfree_skb(skb);
501 		goto errout;
502 	}
503 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
504 	return;
505 errout:
506 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
507 }
508 
509 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
510 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
511 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
512 };
513 
514 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
515 				     struct nlmsghdr *nlh)
516 {
517 	struct net *net = sock_net(in_skb->sk);
518 	struct nlattr *tb[NETCONFA_MAX+1];
519 	struct netconfmsg *ncm;
520 	struct sk_buff *skb;
521 	struct ipv6_devconf *devconf;
522 	struct inet6_dev *in6_dev;
523 	struct net_device *dev;
524 	int ifindex;
525 	int err;
526 
527 	err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
528 			  devconf_ipv6_policy);
529 	if (err < 0)
530 		goto errout;
531 
532 	err = EINVAL;
533 	if (!tb[NETCONFA_IFINDEX])
534 		goto errout;
535 
536 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
537 	switch (ifindex) {
538 	case NETCONFA_IFINDEX_ALL:
539 		devconf = net->ipv6.devconf_all;
540 		break;
541 	case NETCONFA_IFINDEX_DEFAULT:
542 		devconf = net->ipv6.devconf_dflt;
543 		break;
544 	default:
545 		dev = __dev_get_by_index(net, ifindex);
546 		if (dev == NULL)
547 			goto errout;
548 		in6_dev = __in6_dev_get(dev);
549 		if (in6_dev == NULL)
550 			goto errout;
551 		devconf = &in6_dev->cnf;
552 		break;
553 	}
554 
555 	err = -ENOBUFS;
556 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
557 	if (skb == NULL)
558 		goto errout;
559 
560 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
561 					 NETLINK_CB(in_skb).portid,
562 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
563 					 -1);
564 	if (err < 0) {
565 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
566 		WARN_ON(err == -EMSGSIZE);
567 		kfree_skb(skb);
568 		goto errout;
569 	}
570 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
571 errout:
572 	return err;
573 }
574 
575 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
576 				      struct netlink_callback *cb)
577 {
578 	struct net *net = sock_net(skb->sk);
579 	int h, s_h;
580 	int idx, s_idx;
581 	struct net_device *dev;
582 	struct inet6_dev *idev;
583 	struct hlist_head *head;
584 
585 	s_h = cb->args[0];
586 	s_idx = idx = cb->args[1];
587 
588 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
589 		idx = 0;
590 		head = &net->dev_index_head[h];
591 		rcu_read_lock();
592 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
593 			  net->dev_base_seq;
594 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
595 			if (idx < s_idx)
596 				goto cont;
597 			idev = __in6_dev_get(dev);
598 			if (!idev)
599 				goto cont;
600 
601 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
602 						       &idev->cnf,
603 						       NETLINK_CB(cb->skb).portid,
604 						       cb->nlh->nlmsg_seq,
605 						       RTM_NEWNETCONF,
606 						       NLM_F_MULTI,
607 						       -1) <= 0) {
608 				rcu_read_unlock();
609 				goto done;
610 			}
611 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
612 cont:
613 			idx++;
614 		}
615 		rcu_read_unlock();
616 	}
617 	if (h == NETDEV_HASHENTRIES) {
618 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
619 					       net->ipv6.devconf_all,
620 					       NETLINK_CB(cb->skb).portid,
621 					       cb->nlh->nlmsg_seq,
622 					       RTM_NEWNETCONF, NLM_F_MULTI,
623 					       -1) <= 0)
624 			goto done;
625 		else
626 			h++;
627 	}
628 	if (h == NETDEV_HASHENTRIES + 1) {
629 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
630 					       net->ipv6.devconf_dflt,
631 					       NETLINK_CB(cb->skb).portid,
632 					       cb->nlh->nlmsg_seq,
633 					       RTM_NEWNETCONF, NLM_F_MULTI,
634 					       -1) <= 0)
635 			goto done;
636 		else
637 			h++;
638 	}
639 done:
640 	cb->args[0] = h;
641 	cb->args[1] = idx;
642 
643 	return skb->len;
644 }
645 
646 #ifdef CONFIG_SYSCTL
647 static void dev_forward_change(struct inet6_dev *idev)
648 {
649 	struct net_device *dev;
650 	struct inet6_ifaddr *ifa;
651 
652 	if (!idev)
653 		return;
654 	dev = idev->dev;
655 	if (idev->cnf.forwarding)
656 		dev_disable_lro(dev);
657 	if (dev->flags & IFF_MULTICAST) {
658 		if (idev->cnf.forwarding) {
659 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
660 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
661 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
662 		} else {
663 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
664 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
665 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
666 		}
667 	}
668 
669 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
670 		if (ifa->flags&IFA_F_TENTATIVE)
671 			continue;
672 		if (idev->cnf.forwarding)
673 			addrconf_join_anycast(ifa);
674 		else
675 			addrconf_leave_anycast(ifa);
676 	}
677 	inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
678 				     dev->ifindex, &idev->cnf);
679 }
680 
681 
682 static void addrconf_forward_change(struct net *net, __s32 newf)
683 {
684 	struct net_device *dev;
685 	struct inet6_dev *idev;
686 
687 	for_each_netdev(net, dev) {
688 		idev = __in6_dev_get(dev);
689 		if (idev) {
690 			int changed = (!idev->cnf.forwarding) ^ (!newf);
691 			idev->cnf.forwarding = newf;
692 			if (changed)
693 				dev_forward_change(idev);
694 		}
695 	}
696 }
697 
698 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
699 {
700 	struct net *net;
701 	int old;
702 
703 	if (!rtnl_trylock())
704 		return restart_syscall();
705 
706 	net = (struct net *)table->extra2;
707 	old = *p;
708 	*p = newf;
709 
710 	if (p == &net->ipv6.devconf_dflt->forwarding) {
711 		if ((!newf) ^ (!old))
712 			inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
713 						     NETCONFA_IFINDEX_DEFAULT,
714 						     net->ipv6.devconf_dflt);
715 		rtnl_unlock();
716 		return 0;
717 	}
718 
719 	if (p == &net->ipv6.devconf_all->forwarding) {
720 		net->ipv6.devconf_dflt->forwarding = newf;
721 		addrconf_forward_change(net, newf);
722 		if ((!newf) ^ (!old))
723 			inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
724 						     NETCONFA_IFINDEX_ALL,
725 						     net->ipv6.devconf_all);
726 	} else if ((!newf) ^ (!old))
727 		dev_forward_change((struct inet6_dev *)table->extra1);
728 	rtnl_unlock();
729 
730 	if (newf)
731 		rt6_purge_dflt_routers(net);
732 	return 1;
733 }
734 #endif
735 
736 /* Nobody refers to this ifaddr, destroy it */
737 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
738 {
739 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
740 
741 #ifdef NET_REFCNT_DEBUG
742 	pr_debug("%s\n", __func__);
743 #endif
744 
745 	in6_dev_put(ifp->idev);
746 
747 	if (del_timer(&ifp->dad_timer))
748 		pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
749 
750 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
751 		pr_warn("Freeing alive inet6 address %p\n", ifp);
752 		return;
753 	}
754 	ip6_rt_put(ifp->rt);
755 
756 	kfree_rcu(ifp, rcu);
757 }
758 
759 static void
760 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
761 {
762 	struct list_head *p;
763 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
764 
765 	/*
766 	 * Each device address list is sorted in order of scope -
767 	 * global before linklocal.
768 	 */
769 	list_for_each(p, &idev->addr_list) {
770 		struct inet6_ifaddr *ifa
771 			= list_entry(p, struct inet6_ifaddr, if_list);
772 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
773 			break;
774 	}
775 
776 	list_add_tail(&ifp->if_list, p);
777 }
778 
779 static u32 inet6_addr_hash(const struct in6_addr *addr)
780 {
781 	return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
782 }
783 
784 /* On success it returns ifp with increased reference count */
785 
786 static struct inet6_ifaddr *
787 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
788 	      const struct in6_addr *peer_addr, int pfxlen,
789 	      int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
790 {
791 	struct inet6_ifaddr *ifa = NULL;
792 	struct rt6_info *rt;
793 	unsigned int hash;
794 	int err = 0;
795 	int addr_type = ipv6_addr_type(addr);
796 
797 	if (addr_type == IPV6_ADDR_ANY ||
798 	    addr_type & IPV6_ADDR_MULTICAST ||
799 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
800 	     addr_type & IPV6_ADDR_LOOPBACK))
801 		return ERR_PTR(-EADDRNOTAVAIL);
802 
803 	rcu_read_lock_bh();
804 	if (idev->dead) {
805 		err = -ENODEV;			/*XXX*/
806 		goto out2;
807 	}
808 
809 	if (idev->cnf.disable_ipv6) {
810 		err = -EACCES;
811 		goto out2;
812 	}
813 
814 	spin_lock(&addrconf_hash_lock);
815 
816 	/* Ignore adding duplicate addresses on an interface */
817 	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
818 		ADBG("ipv6_add_addr: already assigned\n");
819 		err = -EEXIST;
820 		goto out;
821 	}
822 
823 	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
824 
825 	if (ifa == NULL) {
826 		ADBG("ipv6_add_addr: malloc failed\n");
827 		err = -ENOBUFS;
828 		goto out;
829 	}
830 
831 	rt = addrconf_dst_alloc(idev, addr, false);
832 	if (IS_ERR(rt)) {
833 		err = PTR_ERR(rt);
834 		goto out;
835 	}
836 
837 	ifa->addr = *addr;
838 	if (peer_addr)
839 		ifa->peer_addr = *peer_addr;
840 
841 	spin_lock_init(&ifa->lock);
842 	spin_lock_init(&ifa->state_lock);
843 	setup_timer(&ifa->dad_timer, addrconf_dad_timer,
844 		    (unsigned long)ifa);
845 	INIT_HLIST_NODE(&ifa->addr_lst);
846 	ifa->scope = scope;
847 	ifa->prefix_len = pfxlen;
848 	ifa->flags = flags | IFA_F_TENTATIVE;
849 	ifa->valid_lft = valid_lft;
850 	ifa->prefered_lft = prefered_lft;
851 	ifa->cstamp = ifa->tstamp = jiffies;
852 	ifa->tokenized = false;
853 
854 	ifa->rt = rt;
855 
856 	ifa->idev = idev;
857 	in6_dev_hold(idev);
858 	/* For caller */
859 	in6_ifa_hold(ifa);
860 
861 	/* Add to big hash table */
862 	hash = inet6_addr_hash(addr);
863 
864 	hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
865 	spin_unlock(&addrconf_hash_lock);
866 
867 	write_lock(&idev->lock);
868 	/* Add to inet6_dev unicast addr list. */
869 	ipv6_link_dev_addr(idev, ifa);
870 
871 	if (ifa->flags&IFA_F_TEMPORARY) {
872 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
873 		in6_ifa_hold(ifa);
874 	}
875 
876 	in6_ifa_hold(ifa);
877 	write_unlock(&idev->lock);
878 out2:
879 	rcu_read_unlock_bh();
880 
881 	if (likely(err == 0))
882 		inet6addr_notifier_call_chain(NETDEV_UP, ifa);
883 	else {
884 		kfree(ifa);
885 		ifa = ERR_PTR(err);
886 	}
887 
888 	return ifa;
889 out:
890 	spin_unlock(&addrconf_hash_lock);
891 	goto out2;
892 }
893 
894 /* This function wants to get referenced ifp and releases it before return */
895 
896 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
897 {
898 	struct inet6_ifaddr *ifa, *ifn;
899 	struct inet6_dev *idev = ifp->idev;
900 	int state;
901 	int deleted = 0, onlink = 0;
902 	unsigned long expires = jiffies;
903 
904 	spin_lock_bh(&ifp->state_lock);
905 	state = ifp->state;
906 	ifp->state = INET6_IFADDR_STATE_DEAD;
907 	spin_unlock_bh(&ifp->state_lock);
908 
909 	if (state == INET6_IFADDR_STATE_DEAD)
910 		goto out;
911 
912 	spin_lock_bh(&addrconf_hash_lock);
913 	hlist_del_init_rcu(&ifp->addr_lst);
914 	spin_unlock_bh(&addrconf_hash_lock);
915 
916 	write_lock_bh(&idev->lock);
917 
918 	if (ifp->flags&IFA_F_TEMPORARY) {
919 		list_del(&ifp->tmp_list);
920 		if (ifp->ifpub) {
921 			in6_ifa_put(ifp->ifpub);
922 			ifp->ifpub = NULL;
923 		}
924 		__in6_ifa_put(ifp);
925 	}
926 
927 	list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
928 		if (ifa == ifp) {
929 			list_del_init(&ifp->if_list);
930 			__in6_ifa_put(ifp);
931 
932 			if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
933 				break;
934 			deleted = 1;
935 			continue;
936 		} else if (ifp->flags & IFA_F_PERMANENT) {
937 			if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
938 					      ifp->prefix_len)) {
939 				if (ifa->flags & IFA_F_PERMANENT) {
940 					onlink = 1;
941 					if (deleted)
942 						break;
943 				} else {
944 					unsigned long lifetime;
945 
946 					if (!onlink)
947 						onlink = -1;
948 
949 					spin_lock(&ifa->lock);
950 
951 					lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
952 					/*
953 					 * Note: Because this address is
954 					 * not permanent, lifetime <
955 					 * LONG_MAX / HZ here.
956 					 */
957 					if (time_before(expires,
958 							ifa->tstamp + lifetime * HZ))
959 						expires = ifa->tstamp + lifetime * HZ;
960 					spin_unlock(&ifa->lock);
961 				}
962 			}
963 		}
964 	}
965 	write_unlock_bh(&idev->lock);
966 
967 	addrconf_del_dad_timer(ifp);
968 
969 	ipv6_ifa_notify(RTM_DELADDR, ifp);
970 
971 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
972 
973 	/*
974 	 * Purge or update corresponding prefix
975 	 *
976 	 * 1) we don't purge prefix here if address was not permanent.
977 	 *    prefix is managed by its own lifetime.
978 	 * 2) if there're no addresses, delete prefix.
979 	 * 3) if there're still other permanent address(es),
980 	 *    corresponding prefix is still permanent.
981 	 * 4) otherwise, update prefix lifetime to the
982 	 *    longest valid lifetime among the corresponding
983 	 *    addresses on the device.
984 	 *    Note: subsequent RA will update lifetime.
985 	 *
986 	 * --yoshfuji
987 	 */
988 	if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
989 		struct in6_addr prefix;
990 		struct rt6_info *rt;
991 
992 		ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
993 
994 		rt = addrconf_get_prefix_route(&prefix,
995 					       ifp->prefix_len,
996 					       ifp->idev->dev,
997 					       0, RTF_GATEWAY | RTF_DEFAULT);
998 
999 		if (rt) {
1000 			if (onlink == 0) {
1001 				ip6_del_rt(rt);
1002 				rt = NULL;
1003 			} else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1004 				rt6_set_expires(rt, expires);
1005 			}
1006 		}
1007 		ip6_rt_put(rt);
1008 	}
1009 
1010 	/* clean up prefsrc entries */
1011 	rt6_remove_prefsrc(ifp);
1012 out:
1013 	in6_ifa_put(ifp);
1014 }
1015 
1016 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1017 {
1018 	struct inet6_dev *idev = ifp->idev;
1019 	struct in6_addr addr, *tmpaddr;
1020 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1021 	unsigned long regen_advance;
1022 	int tmp_plen;
1023 	int ret = 0;
1024 	u32 addr_flags;
1025 	unsigned long now = jiffies;
1026 
1027 	write_lock(&idev->lock);
1028 	if (ift) {
1029 		spin_lock_bh(&ift->lock);
1030 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1031 		spin_unlock_bh(&ift->lock);
1032 		tmpaddr = &addr;
1033 	} else {
1034 		tmpaddr = NULL;
1035 	}
1036 retry:
1037 	in6_dev_hold(idev);
1038 	if (idev->cnf.use_tempaddr <= 0) {
1039 		write_unlock(&idev->lock);
1040 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1041 		in6_dev_put(idev);
1042 		ret = -1;
1043 		goto out;
1044 	}
1045 	spin_lock_bh(&ifp->lock);
1046 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1047 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1048 		spin_unlock_bh(&ifp->lock);
1049 		write_unlock(&idev->lock);
1050 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1051 			__func__);
1052 		in6_dev_put(idev);
1053 		ret = -1;
1054 		goto out;
1055 	}
1056 	in6_ifa_hold(ifp);
1057 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1058 	__ipv6_try_regen_rndid(idev, tmpaddr);
1059 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
1060 	age = (now - ifp->tstamp) / HZ;
1061 	tmp_valid_lft = min_t(__u32,
1062 			      ifp->valid_lft,
1063 			      idev->cnf.temp_valid_lft + age);
1064 	tmp_prefered_lft = min_t(__u32,
1065 				 ifp->prefered_lft,
1066 				 idev->cnf.temp_prefered_lft + age -
1067 				 idev->cnf.max_desync_factor);
1068 	tmp_plen = ifp->prefix_len;
1069 	tmp_tstamp = ifp->tstamp;
1070 	spin_unlock_bh(&ifp->lock);
1071 
1072 	regen_advance = idev->cnf.regen_max_retry *
1073 	                idev->cnf.dad_transmits *
1074 	                idev->nd_parms->retrans_time / HZ;
1075 	write_unlock(&idev->lock);
1076 
1077 	/* A temporary address is created only if this calculated Preferred
1078 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1079 	 * an implementation must not create a temporary address with a zero
1080 	 * Preferred Lifetime.
1081 	 */
1082 	if (tmp_prefered_lft <= regen_advance) {
1083 		in6_ifa_put(ifp);
1084 		in6_dev_put(idev);
1085 		ret = -1;
1086 		goto out;
1087 	}
1088 
1089 	addr_flags = IFA_F_TEMPORARY;
1090 	/* set in addrconf_prefix_rcv() */
1091 	if (ifp->flags & IFA_F_OPTIMISTIC)
1092 		addr_flags |= IFA_F_OPTIMISTIC;
1093 
1094 	ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1095 			    ipv6_addr_scope(&addr), addr_flags,
1096 			    tmp_valid_lft, tmp_prefered_lft);
1097 	if (IS_ERR(ift)) {
1098 		in6_ifa_put(ifp);
1099 		in6_dev_put(idev);
1100 		pr_info("%s: retry temporary address regeneration\n", __func__);
1101 		tmpaddr = &addr;
1102 		write_lock(&idev->lock);
1103 		goto retry;
1104 	}
1105 
1106 	spin_lock_bh(&ift->lock);
1107 	ift->ifpub = ifp;
1108 	ift->cstamp = now;
1109 	ift->tstamp = tmp_tstamp;
1110 	spin_unlock_bh(&ift->lock);
1111 
1112 	addrconf_dad_start(ift);
1113 	in6_ifa_put(ift);
1114 	in6_dev_put(idev);
1115 out:
1116 	return ret;
1117 }
1118 
1119 /*
1120  *	Choose an appropriate source address (RFC3484)
1121  */
1122 enum {
1123 	IPV6_SADDR_RULE_INIT = 0,
1124 	IPV6_SADDR_RULE_LOCAL,
1125 	IPV6_SADDR_RULE_SCOPE,
1126 	IPV6_SADDR_RULE_PREFERRED,
1127 #ifdef CONFIG_IPV6_MIP6
1128 	IPV6_SADDR_RULE_HOA,
1129 #endif
1130 	IPV6_SADDR_RULE_OIF,
1131 	IPV6_SADDR_RULE_LABEL,
1132 	IPV6_SADDR_RULE_PRIVACY,
1133 	IPV6_SADDR_RULE_ORCHID,
1134 	IPV6_SADDR_RULE_PREFIX,
1135 	IPV6_SADDR_RULE_MAX
1136 };
1137 
1138 struct ipv6_saddr_score {
1139 	int			rule;
1140 	int			addr_type;
1141 	struct inet6_ifaddr	*ifa;
1142 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1143 	int			scopedist;
1144 	int			matchlen;
1145 };
1146 
1147 struct ipv6_saddr_dst {
1148 	const struct in6_addr *addr;
1149 	int ifindex;
1150 	int scope;
1151 	int label;
1152 	unsigned int prefs;
1153 };
1154 
1155 static inline int ipv6_saddr_preferred(int type)
1156 {
1157 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1158 		return 1;
1159 	return 0;
1160 }
1161 
1162 static int ipv6_get_saddr_eval(struct net *net,
1163 			       struct ipv6_saddr_score *score,
1164 			       struct ipv6_saddr_dst *dst,
1165 			       int i)
1166 {
1167 	int ret;
1168 
1169 	if (i <= score->rule) {
1170 		switch (i) {
1171 		case IPV6_SADDR_RULE_SCOPE:
1172 			ret = score->scopedist;
1173 			break;
1174 		case IPV6_SADDR_RULE_PREFIX:
1175 			ret = score->matchlen;
1176 			break;
1177 		default:
1178 			ret = !!test_bit(i, score->scorebits);
1179 		}
1180 		goto out;
1181 	}
1182 
1183 	switch (i) {
1184 	case IPV6_SADDR_RULE_INIT:
1185 		/* Rule 0: remember if hiscore is not ready yet */
1186 		ret = !!score->ifa;
1187 		break;
1188 	case IPV6_SADDR_RULE_LOCAL:
1189 		/* Rule 1: Prefer same address */
1190 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1191 		break;
1192 	case IPV6_SADDR_RULE_SCOPE:
1193 		/* Rule 2: Prefer appropriate scope
1194 		 *
1195 		 *      ret
1196 		 *       ^
1197 		 *    -1 |  d 15
1198 		 *    ---+--+-+---> scope
1199 		 *       |
1200 		 *       |             d is scope of the destination.
1201 		 *  B-d  |  \
1202 		 *       |   \      <- smaller scope is better if
1203 		 *  B-15 |    \        if scope is enough for destinaion.
1204 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1205 		 * d-C-1 | /
1206 		 *       |/         <- greater is better
1207 		 *   -C  /             if scope is not enough for destination.
1208 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1209 		 *
1210 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1211 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1212 		 * Assume B = 0 and we get C > 29.
1213 		 */
1214 		ret = __ipv6_addr_src_scope(score->addr_type);
1215 		if (ret >= dst->scope)
1216 			ret = -ret;
1217 		else
1218 			ret -= 128;	/* 30 is enough */
1219 		score->scopedist = ret;
1220 		break;
1221 	case IPV6_SADDR_RULE_PREFERRED:
1222 		/* Rule 3: Avoid deprecated and optimistic addresses */
1223 		ret = ipv6_saddr_preferred(score->addr_type) ||
1224 		      !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1225 		break;
1226 #ifdef CONFIG_IPV6_MIP6
1227 	case IPV6_SADDR_RULE_HOA:
1228 	    {
1229 		/* Rule 4: Prefer home address */
1230 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1231 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1232 		break;
1233 	    }
1234 #endif
1235 	case IPV6_SADDR_RULE_OIF:
1236 		/* Rule 5: Prefer outgoing interface */
1237 		ret = (!dst->ifindex ||
1238 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1239 		break;
1240 	case IPV6_SADDR_RULE_LABEL:
1241 		/* Rule 6: Prefer matching label */
1242 		ret = ipv6_addr_label(net,
1243 				      &score->ifa->addr, score->addr_type,
1244 				      score->ifa->idev->dev->ifindex) == dst->label;
1245 		break;
1246 	case IPV6_SADDR_RULE_PRIVACY:
1247 	    {
1248 		/* Rule 7: Prefer public address
1249 		 * Note: prefer temporary address if use_tempaddr >= 2
1250 		 */
1251 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1252 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1253 				score->ifa->idev->cnf.use_tempaddr >= 2;
1254 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1255 		break;
1256 	    }
1257 	case IPV6_SADDR_RULE_ORCHID:
1258 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1259 		 *	    non-ORCHID vs non-ORCHID
1260 		 */
1261 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1262 			ipv6_addr_orchid(dst->addr));
1263 		break;
1264 	case IPV6_SADDR_RULE_PREFIX:
1265 		/* Rule 8: Use longest matching prefix */
1266 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1267 		if (ret > score->ifa->prefix_len)
1268 			ret = score->ifa->prefix_len;
1269 		score->matchlen = ret;
1270 		break;
1271 	default:
1272 		ret = 0;
1273 	}
1274 
1275 	if (ret)
1276 		__set_bit(i, score->scorebits);
1277 	score->rule = i;
1278 out:
1279 	return ret;
1280 }
1281 
1282 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1283 		       const struct in6_addr *daddr, unsigned int prefs,
1284 		       struct in6_addr *saddr)
1285 {
1286 	struct ipv6_saddr_score scores[2],
1287 				*score = &scores[0], *hiscore = &scores[1];
1288 	struct ipv6_saddr_dst dst;
1289 	struct net_device *dev;
1290 	int dst_type;
1291 
1292 	dst_type = __ipv6_addr_type(daddr);
1293 	dst.addr = daddr;
1294 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1295 	dst.scope = __ipv6_addr_src_scope(dst_type);
1296 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1297 	dst.prefs = prefs;
1298 
1299 	hiscore->rule = -1;
1300 	hiscore->ifa = NULL;
1301 
1302 	rcu_read_lock();
1303 
1304 	for_each_netdev_rcu(net, dev) {
1305 		struct inet6_dev *idev;
1306 
1307 		/* Candidate Source Address (section 4)
1308 		 *  - multicast and link-local destination address,
1309 		 *    the set of candidate source address MUST only
1310 		 *    include addresses assigned to interfaces
1311 		 *    belonging to the same link as the outgoing
1312 		 *    interface.
1313 		 * (- For site-local destination addresses, the
1314 		 *    set of candidate source addresses MUST only
1315 		 *    include addresses assigned to interfaces
1316 		 *    belonging to the same site as the outgoing
1317 		 *    interface.)
1318 		 */
1319 		if (((dst_type & IPV6_ADDR_MULTICAST) ||
1320 		     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1321 		    dst.ifindex && dev->ifindex != dst.ifindex)
1322 			continue;
1323 
1324 		idev = __in6_dev_get(dev);
1325 		if (!idev)
1326 			continue;
1327 
1328 		read_lock_bh(&idev->lock);
1329 		list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1330 			int i;
1331 
1332 			/*
1333 			 * - Tentative Address (RFC2462 section 5.4)
1334 			 *  - A tentative address is not considered
1335 			 *    "assigned to an interface" in the traditional
1336 			 *    sense, unless it is also flagged as optimistic.
1337 			 * - Candidate Source Address (section 4)
1338 			 *  - In any case, anycast addresses, multicast
1339 			 *    addresses, and the unspecified address MUST
1340 			 *    NOT be included in a candidate set.
1341 			 */
1342 			if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1343 			    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1344 				continue;
1345 
1346 			score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1347 
1348 			if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1349 				     score->addr_type & IPV6_ADDR_MULTICAST)) {
1350 				LIMIT_NETDEBUG(KERN_DEBUG
1351 					       "ADDRCONF: unspecified / multicast address "
1352 					       "assigned as unicast address on %s",
1353 					       dev->name);
1354 				continue;
1355 			}
1356 
1357 			score->rule = -1;
1358 			bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1359 
1360 			for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1361 				int minihiscore, miniscore;
1362 
1363 				minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1364 				miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1365 
1366 				if (minihiscore > miniscore) {
1367 					if (i == IPV6_SADDR_RULE_SCOPE &&
1368 					    score->scopedist > 0) {
1369 						/*
1370 						 * special case:
1371 						 * each remaining entry
1372 						 * has too small (not enough)
1373 						 * scope, because ifa entries
1374 						 * are sorted by their scope
1375 						 * values.
1376 						 */
1377 						goto try_nextdev;
1378 					}
1379 					break;
1380 				} else if (minihiscore < miniscore) {
1381 					if (hiscore->ifa)
1382 						in6_ifa_put(hiscore->ifa);
1383 
1384 					in6_ifa_hold(score->ifa);
1385 
1386 					swap(hiscore, score);
1387 
1388 					/* restore our iterator */
1389 					score->ifa = hiscore->ifa;
1390 
1391 					break;
1392 				}
1393 			}
1394 		}
1395 try_nextdev:
1396 		read_unlock_bh(&idev->lock);
1397 	}
1398 	rcu_read_unlock();
1399 
1400 	if (!hiscore->ifa)
1401 		return -EADDRNOTAVAIL;
1402 
1403 	*saddr = hiscore->ifa->addr;
1404 	in6_ifa_put(hiscore->ifa);
1405 	return 0;
1406 }
1407 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1408 
1409 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1410 		      unsigned char banned_flags)
1411 {
1412 	struct inet6_ifaddr *ifp;
1413 	int err = -EADDRNOTAVAIL;
1414 
1415 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
1416 		if (ifp->scope == IFA_LINK &&
1417 		    !(ifp->flags & banned_flags)) {
1418 			*addr = ifp->addr;
1419 			err = 0;
1420 			break;
1421 		}
1422 	}
1423 	return err;
1424 }
1425 
1426 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1427 		    unsigned char banned_flags)
1428 {
1429 	struct inet6_dev *idev;
1430 	int err = -EADDRNOTAVAIL;
1431 
1432 	rcu_read_lock();
1433 	idev = __in6_dev_get(dev);
1434 	if (idev) {
1435 		read_lock_bh(&idev->lock);
1436 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1437 		read_unlock_bh(&idev->lock);
1438 	}
1439 	rcu_read_unlock();
1440 	return err;
1441 }
1442 
1443 static int ipv6_count_addresses(struct inet6_dev *idev)
1444 {
1445 	int cnt = 0;
1446 	struct inet6_ifaddr *ifp;
1447 
1448 	read_lock_bh(&idev->lock);
1449 	list_for_each_entry(ifp, &idev->addr_list, if_list)
1450 		cnt++;
1451 	read_unlock_bh(&idev->lock);
1452 	return cnt;
1453 }
1454 
1455 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1456 		  const struct net_device *dev, int strict)
1457 {
1458 	struct inet6_ifaddr *ifp;
1459 	unsigned int hash = inet6_addr_hash(addr);
1460 
1461 	rcu_read_lock_bh();
1462 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1463 		if (!net_eq(dev_net(ifp->idev->dev), net))
1464 			continue;
1465 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1466 		    !(ifp->flags&IFA_F_TENTATIVE) &&
1467 		    (dev == NULL || ifp->idev->dev == dev ||
1468 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1469 			rcu_read_unlock_bh();
1470 			return 1;
1471 		}
1472 	}
1473 
1474 	rcu_read_unlock_bh();
1475 	return 0;
1476 }
1477 EXPORT_SYMBOL(ipv6_chk_addr);
1478 
1479 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1480 			       struct net_device *dev)
1481 {
1482 	unsigned int hash = inet6_addr_hash(addr);
1483 	struct inet6_ifaddr *ifp;
1484 
1485 	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1486 		if (!net_eq(dev_net(ifp->idev->dev), net))
1487 			continue;
1488 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1489 			if (dev == NULL || ifp->idev->dev == dev)
1490 				return true;
1491 		}
1492 	}
1493 	return false;
1494 }
1495 
1496 /* Compares an address/prefix_len with addresses on device @dev.
1497  * If one is found it returns true.
1498  */
1499 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1500 	const unsigned int prefix_len, struct net_device *dev)
1501 {
1502 	struct inet6_dev *idev;
1503 	struct inet6_ifaddr *ifa;
1504 	bool ret = false;
1505 
1506 	rcu_read_lock();
1507 	idev = __in6_dev_get(dev);
1508 	if (idev) {
1509 		read_lock_bh(&idev->lock);
1510 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1511 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1512 			if (ret)
1513 				break;
1514 		}
1515 		read_unlock_bh(&idev->lock);
1516 	}
1517 	rcu_read_unlock();
1518 
1519 	return ret;
1520 }
1521 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1522 
1523 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1524 {
1525 	struct inet6_dev *idev;
1526 	struct inet6_ifaddr *ifa;
1527 	int	onlink;
1528 
1529 	onlink = 0;
1530 	rcu_read_lock();
1531 	idev = __in6_dev_get(dev);
1532 	if (idev) {
1533 		read_lock_bh(&idev->lock);
1534 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1535 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1536 						   ifa->prefix_len);
1537 			if (onlink)
1538 				break;
1539 		}
1540 		read_unlock_bh(&idev->lock);
1541 	}
1542 	rcu_read_unlock();
1543 	return onlink;
1544 }
1545 EXPORT_SYMBOL(ipv6_chk_prefix);
1546 
1547 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1548 				     struct net_device *dev, int strict)
1549 {
1550 	struct inet6_ifaddr *ifp, *result = NULL;
1551 	unsigned int hash = inet6_addr_hash(addr);
1552 
1553 	rcu_read_lock_bh();
1554 	hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1555 		if (!net_eq(dev_net(ifp->idev->dev), net))
1556 			continue;
1557 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1558 			if (dev == NULL || ifp->idev->dev == dev ||
1559 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1560 				result = ifp;
1561 				in6_ifa_hold(ifp);
1562 				break;
1563 			}
1564 		}
1565 	}
1566 	rcu_read_unlock_bh();
1567 
1568 	return result;
1569 }
1570 
1571 /* Gets referenced address, destroys ifaddr */
1572 
1573 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1574 {
1575 	if (ifp->flags&IFA_F_PERMANENT) {
1576 		spin_lock_bh(&ifp->lock);
1577 		addrconf_del_dad_timer(ifp);
1578 		ifp->flags |= IFA_F_TENTATIVE;
1579 		if (dad_failed)
1580 			ifp->flags |= IFA_F_DADFAILED;
1581 		spin_unlock_bh(&ifp->lock);
1582 		if (dad_failed)
1583 			ipv6_ifa_notify(0, ifp);
1584 		in6_ifa_put(ifp);
1585 	} else if (ifp->flags&IFA_F_TEMPORARY) {
1586 		struct inet6_ifaddr *ifpub;
1587 		spin_lock_bh(&ifp->lock);
1588 		ifpub = ifp->ifpub;
1589 		if (ifpub) {
1590 			in6_ifa_hold(ifpub);
1591 			spin_unlock_bh(&ifp->lock);
1592 			ipv6_create_tempaddr(ifpub, ifp);
1593 			in6_ifa_put(ifpub);
1594 		} else {
1595 			spin_unlock_bh(&ifp->lock);
1596 		}
1597 		ipv6_del_addr(ifp);
1598 	} else
1599 		ipv6_del_addr(ifp);
1600 }
1601 
1602 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1603 {
1604 	int err = -ENOENT;
1605 
1606 	spin_lock(&ifp->state_lock);
1607 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
1608 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
1609 		err = 0;
1610 	}
1611 	spin_unlock(&ifp->state_lock);
1612 
1613 	return err;
1614 }
1615 
1616 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1617 {
1618 	struct inet6_dev *idev = ifp->idev;
1619 
1620 	if (addrconf_dad_end(ifp)) {
1621 		in6_ifa_put(ifp);
1622 		return;
1623 	}
1624 
1625 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1626 			     ifp->idev->dev->name, &ifp->addr);
1627 
1628 	if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1629 		struct in6_addr addr;
1630 
1631 		addr.s6_addr32[0] = htonl(0xfe800000);
1632 		addr.s6_addr32[1] = 0;
1633 
1634 		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1635 		    ipv6_addr_equal(&ifp->addr, &addr)) {
1636 			/* DAD failed for link-local based on MAC address */
1637 			idev->cnf.disable_ipv6 = 1;
1638 
1639 			pr_info("%s: IPv6 being disabled!\n",
1640 				ifp->idev->dev->name);
1641 		}
1642 	}
1643 
1644 	addrconf_dad_stop(ifp, 1);
1645 }
1646 
1647 /* Join to solicited addr multicast group. */
1648 
1649 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1650 {
1651 	struct in6_addr maddr;
1652 
1653 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1654 		return;
1655 
1656 	addrconf_addr_solict_mult(addr, &maddr);
1657 	ipv6_dev_mc_inc(dev, &maddr);
1658 }
1659 
1660 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1661 {
1662 	struct in6_addr maddr;
1663 
1664 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1665 		return;
1666 
1667 	addrconf_addr_solict_mult(addr, &maddr);
1668 	__ipv6_dev_mc_dec(idev, &maddr);
1669 }
1670 
1671 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1672 {
1673 	struct in6_addr addr;
1674 	if (ifp->prefix_len == 127) /* RFC 6164 */
1675 		return;
1676 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1677 	if (ipv6_addr_any(&addr))
1678 		return;
1679 	ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1680 }
1681 
1682 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1683 {
1684 	struct in6_addr addr;
1685 	if (ifp->prefix_len == 127) /* RFC 6164 */
1686 		return;
1687 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1688 	if (ipv6_addr_any(&addr))
1689 		return;
1690 	__ipv6_dev_ac_dec(ifp->idev, &addr);
1691 }
1692 
1693 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1694 {
1695 	if (dev->addr_len != ETH_ALEN)
1696 		return -1;
1697 	memcpy(eui, dev->dev_addr, 3);
1698 	memcpy(eui + 5, dev->dev_addr + 3, 3);
1699 
1700 	/*
1701 	 * The zSeries OSA network cards can be shared among various
1702 	 * OS instances, but the OSA cards have only one MAC address.
1703 	 * This leads to duplicate address conflicts in conjunction
1704 	 * with IPv6 if more than one instance uses the same card.
1705 	 *
1706 	 * The driver for these cards can deliver a unique 16-bit
1707 	 * identifier for each instance sharing the same card.  It is
1708 	 * placed instead of 0xFFFE in the interface identifier.  The
1709 	 * "u" bit of the interface identifier is not inverted in this
1710 	 * case.  Hence the resulting interface identifier has local
1711 	 * scope according to RFC2373.
1712 	 */
1713 	if (dev->dev_id) {
1714 		eui[3] = (dev->dev_id >> 8) & 0xFF;
1715 		eui[4] = dev->dev_id & 0xFF;
1716 	} else {
1717 		eui[3] = 0xFF;
1718 		eui[4] = 0xFE;
1719 		eui[0] ^= 2;
1720 	}
1721 	return 0;
1722 }
1723 
1724 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1725 {
1726 	if (dev->addr_len != IEEE802154_ADDR_LEN)
1727 		return -1;
1728 	memcpy(eui, dev->dev_addr, 8);
1729 	eui[0] ^= 2;
1730 	return 0;
1731 }
1732 
1733 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1734 {
1735 	union fwnet_hwaddr *ha;
1736 
1737 	if (dev->addr_len != FWNET_ALEN)
1738 		return -1;
1739 
1740 	ha = (union fwnet_hwaddr *)dev->dev_addr;
1741 
1742 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1743 	eui[0] ^= 2;
1744 	return 0;
1745 }
1746 
1747 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1748 {
1749 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1750 	if (dev->addr_len != ARCNET_ALEN)
1751 		return -1;
1752 	memset(eui, 0, 7);
1753 	eui[7] = *(u8 *)dev->dev_addr;
1754 	return 0;
1755 }
1756 
1757 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1758 {
1759 	if (dev->addr_len != INFINIBAND_ALEN)
1760 		return -1;
1761 	memcpy(eui, dev->dev_addr + 12, 8);
1762 	eui[0] |= 2;
1763 	return 0;
1764 }
1765 
1766 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1767 {
1768 	if (addr == 0)
1769 		return -1;
1770 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1771 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1772 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1773 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1774 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1775 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1776 	eui[1] = 0;
1777 	eui[2] = 0x5E;
1778 	eui[3] = 0xFE;
1779 	memcpy(eui + 4, &addr, 4);
1780 	return 0;
1781 }
1782 
1783 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1784 {
1785 	if (dev->priv_flags & IFF_ISATAP)
1786 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1787 	return -1;
1788 }
1789 
1790 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1791 {
1792 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1793 }
1794 
1795 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1796 {
1797 	memcpy(eui, dev->perm_addr, 3);
1798 	memcpy(eui + 5, dev->perm_addr + 3, 3);
1799 	eui[3] = 0xFF;
1800 	eui[4] = 0xFE;
1801 	eui[0] ^= 2;
1802 	return 0;
1803 }
1804 
1805 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1806 {
1807 	switch (dev->type) {
1808 	case ARPHRD_ETHER:
1809 	case ARPHRD_FDDI:
1810 		return addrconf_ifid_eui48(eui, dev);
1811 	case ARPHRD_ARCNET:
1812 		return addrconf_ifid_arcnet(eui, dev);
1813 	case ARPHRD_INFINIBAND:
1814 		return addrconf_ifid_infiniband(eui, dev);
1815 	case ARPHRD_SIT:
1816 		return addrconf_ifid_sit(eui, dev);
1817 	case ARPHRD_IPGRE:
1818 		return addrconf_ifid_gre(eui, dev);
1819 	case ARPHRD_IEEE802154:
1820 		return addrconf_ifid_eui64(eui, dev);
1821 	case ARPHRD_IEEE1394:
1822 		return addrconf_ifid_ieee1394(eui, dev);
1823 	case ARPHRD_TUNNEL6:
1824 		return addrconf_ifid_ip6tnl(eui, dev);
1825 	}
1826 	return -1;
1827 }
1828 
1829 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1830 {
1831 	int err = -1;
1832 	struct inet6_ifaddr *ifp;
1833 
1834 	read_lock_bh(&idev->lock);
1835 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
1836 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1837 			memcpy(eui, ifp->addr.s6_addr+8, 8);
1838 			err = 0;
1839 			break;
1840 		}
1841 	}
1842 	read_unlock_bh(&idev->lock);
1843 	return err;
1844 }
1845 
1846 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1847 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1848 {
1849 regen:
1850 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
1851 	idev->rndid[0] &= ~0x02;
1852 
1853 	/*
1854 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1855 	 * check if generated address is not inappropriate
1856 	 *
1857 	 *  - Reserved subnet anycast (RFC 2526)
1858 	 *	11111101 11....11 1xxxxxxx
1859 	 *  - ISATAP (RFC4214) 6.1
1860 	 *	00-00-5E-FE-xx-xx-xx-xx
1861 	 *  - value 0
1862 	 *  - XXX: already assigned to an address on the device
1863 	 */
1864 	if (idev->rndid[0] == 0xfd &&
1865 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1866 	    (idev->rndid[7]&0x80))
1867 		goto regen;
1868 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1869 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1870 			goto regen;
1871 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1872 			goto regen;
1873 	}
1874 }
1875 
1876 static void ipv6_regen_rndid(unsigned long data)
1877 {
1878 	struct inet6_dev *idev = (struct inet6_dev *) data;
1879 	unsigned long expires;
1880 
1881 	rcu_read_lock_bh();
1882 	write_lock_bh(&idev->lock);
1883 
1884 	if (idev->dead)
1885 		goto out;
1886 
1887 	__ipv6_regen_rndid(idev);
1888 
1889 	expires = jiffies +
1890 		idev->cnf.temp_prefered_lft * HZ -
1891 		idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1892 		idev->cnf.max_desync_factor * HZ;
1893 	if (time_before(expires, jiffies)) {
1894 		pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1895 			__func__, idev->dev->name);
1896 		goto out;
1897 	}
1898 
1899 	if (!mod_timer(&idev->regen_timer, expires))
1900 		in6_dev_hold(idev);
1901 
1902 out:
1903 	write_unlock_bh(&idev->lock);
1904 	rcu_read_unlock_bh();
1905 	in6_dev_put(idev);
1906 }
1907 
1908 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1909 {
1910 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1911 		__ipv6_regen_rndid(idev);
1912 }
1913 
1914 /*
1915  *	Add prefix route.
1916  */
1917 
1918 static void
1919 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1920 		      unsigned long expires, u32 flags)
1921 {
1922 	struct fib6_config cfg = {
1923 		.fc_table = RT6_TABLE_PREFIX,
1924 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1925 		.fc_ifindex = dev->ifindex,
1926 		.fc_expires = expires,
1927 		.fc_dst_len = plen,
1928 		.fc_flags = RTF_UP | flags,
1929 		.fc_nlinfo.nl_net = dev_net(dev),
1930 		.fc_protocol = RTPROT_KERNEL,
1931 	};
1932 
1933 	cfg.fc_dst = *pfx;
1934 
1935 	/* Prevent useless cloning on PtP SIT.
1936 	   This thing is done here expecting that the whole
1937 	   class of non-broadcast devices need not cloning.
1938 	 */
1939 #if IS_ENABLED(CONFIG_IPV6_SIT)
1940 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1941 		cfg.fc_flags |= RTF_NONEXTHOP;
1942 #endif
1943 
1944 	ip6_route_add(&cfg);
1945 }
1946 
1947 
1948 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1949 						  int plen,
1950 						  const struct net_device *dev,
1951 						  u32 flags, u32 noflags)
1952 {
1953 	struct fib6_node *fn;
1954 	struct rt6_info *rt = NULL;
1955 	struct fib6_table *table;
1956 
1957 	table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1958 	if (table == NULL)
1959 		return NULL;
1960 
1961 	read_lock_bh(&table->tb6_lock);
1962 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1963 	if (!fn)
1964 		goto out;
1965 	for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1966 		if (rt->dst.dev->ifindex != dev->ifindex)
1967 			continue;
1968 		if ((rt->rt6i_flags & flags) != flags)
1969 			continue;
1970 		if ((rt->rt6i_flags & noflags) != 0)
1971 			continue;
1972 		dst_hold(&rt->dst);
1973 		break;
1974 	}
1975 out:
1976 	read_unlock_bh(&table->tb6_lock);
1977 	return rt;
1978 }
1979 
1980 
1981 /* Create "default" multicast route to the interface */
1982 
1983 static void addrconf_add_mroute(struct net_device *dev)
1984 {
1985 	struct fib6_config cfg = {
1986 		.fc_table = RT6_TABLE_LOCAL,
1987 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
1988 		.fc_ifindex = dev->ifindex,
1989 		.fc_dst_len = 8,
1990 		.fc_flags = RTF_UP,
1991 		.fc_nlinfo.nl_net = dev_net(dev),
1992 	};
1993 
1994 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1995 
1996 	ip6_route_add(&cfg);
1997 }
1998 
1999 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2000 {
2001 	struct inet6_dev *idev;
2002 
2003 	ASSERT_RTNL();
2004 
2005 	idev = ipv6_find_idev(dev);
2006 	if (!idev)
2007 		return ERR_PTR(-ENOBUFS);
2008 
2009 	if (idev->cnf.disable_ipv6)
2010 		return ERR_PTR(-EACCES);
2011 
2012 	/* Add default multicast route */
2013 	if (!(dev->flags & IFF_LOOPBACK))
2014 		addrconf_add_mroute(dev);
2015 
2016 	return idev;
2017 }
2018 
2019 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2020 {
2021 	struct prefix_info *pinfo;
2022 	__u32 valid_lft;
2023 	__u32 prefered_lft;
2024 	int addr_type;
2025 	struct inet6_dev *in6_dev;
2026 	struct net *net = dev_net(dev);
2027 
2028 	pinfo = (struct prefix_info *) opt;
2029 
2030 	if (len < sizeof(struct prefix_info)) {
2031 		ADBG("addrconf: prefix option too short\n");
2032 		return;
2033 	}
2034 
2035 	/*
2036 	 *	Validation checks ([ADDRCONF], page 19)
2037 	 */
2038 
2039 	addr_type = ipv6_addr_type(&pinfo->prefix);
2040 
2041 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2042 		return;
2043 
2044 	valid_lft = ntohl(pinfo->valid);
2045 	prefered_lft = ntohl(pinfo->prefered);
2046 
2047 	if (prefered_lft > valid_lft) {
2048 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2049 		return;
2050 	}
2051 
2052 	in6_dev = in6_dev_get(dev);
2053 
2054 	if (in6_dev == NULL) {
2055 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2056 				    dev->name);
2057 		return;
2058 	}
2059 
2060 	/*
2061 	 *	Two things going on here:
2062 	 *	1) Add routes for on-link prefixes
2063 	 *	2) Configure prefixes with the auto flag set
2064 	 */
2065 
2066 	if (pinfo->onlink) {
2067 		struct rt6_info *rt;
2068 		unsigned long rt_expires;
2069 
2070 		/* Avoid arithmetic overflow. Really, we could
2071 		 * save rt_expires in seconds, likely valid_lft,
2072 		 * but it would require division in fib gc, that it
2073 		 * not good.
2074 		 */
2075 		if (HZ > USER_HZ)
2076 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2077 		else
2078 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2079 
2080 		if (addrconf_finite_timeout(rt_expires))
2081 			rt_expires *= HZ;
2082 
2083 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2084 					       pinfo->prefix_len,
2085 					       dev,
2086 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2087 					       RTF_GATEWAY | RTF_DEFAULT);
2088 
2089 		if (rt) {
2090 			/* Autoconf prefix route */
2091 			if (valid_lft == 0) {
2092 				ip6_del_rt(rt);
2093 				rt = NULL;
2094 			} else if (addrconf_finite_timeout(rt_expires)) {
2095 				/* not infinity */
2096 				rt6_set_expires(rt, jiffies + rt_expires);
2097 			} else {
2098 				rt6_clean_expires(rt);
2099 			}
2100 		} else if (valid_lft) {
2101 			clock_t expires = 0;
2102 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2103 			if (addrconf_finite_timeout(rt_expires)) {
2104 				/* not infinity */
2105 				flags |= RTF_EXPIRES;
2106 				expires = jiffies_to_clock_t(rt_expires);
2107 			}
2108 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2109 					      dev, expires, flags);
2110 		}
2111 		ip6_rt_put(rt);
2112 	}
2113 
2114 	/* Try to figure out our local address for this prefix */
2115 
2116 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2117 		struct inet6_ifaddr *ifp;
2118 		struct in6_addr addr;
2119 		int create = 0, update_lft = 0;
2120 		bool tokenized = false;
2121 
2122 		if (pinfo->prefix_len == 64) {
2123 			memcpy(&addr, &pinfo->prefix, 8);
2124 
2125 			if (!ipv6_addr_any(&in6_dev->token)) {
2126 				read_lock_bh(&in6_dev->lock);
2127 				memcpy(addr.s6_addr + 8,
2128 				       in6_dev->token.s6_addr + 8, 8);
2129 				read_unlock_bh(&in6_dev->lock);
2130 				tokenized = true;
2131 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2132 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2133 				in6_dev_put(in6_dev);
2134 				return;
2135 			}
2136 			goto ok;
2137 		}
2138 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2139 				    pinfo->prefix_len);
2140 		in6_dev_put(in6_dev);
2141 		return;
2142 
2143 ok:
2144 
2145 		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2146 
2147 		if (ifp == NULL && valid_lft) {
2148 			int max_addresses = in6_dev->cnf.max_addresses;
2149 			u32 addr_flags = 0;
2150 
2151 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2152 			if (in6_dev->cnf.optimistic_dad &&
2153 			    !net->ipv6.devconf_all->forwarding && sllao)
2154 				addr_flags = IFA_F_OPTIMISTIC;
2155 #endif
2156 
2157 			/* Do not allow to create too much of autoconfigured
2158 			 * addresses; this would be too easy way to crash kernel.
2159 			 */
2160 			if (!max_addresses ||
2161 			    ipv6_count_addresses(in6_dev) < max_addresses)
2162 				ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2163 						    pinfo->prefix_len,
2164 						    addr_type&IPV6_ADDR_SCOPE_MASK,
2165 						    addr_flags, valid_lft,
2166 						    prefered_lft);
2167 
2168 			if (IS_ERR_OR_NULL(ifp)) {
2169 				in6_dev_put(in6_dev);
2170 				return;
2171 			}
2172 
2173 			update_lft = 0;
2174 			create = 1;
2175 			ifp->cstamp = jiffies;
2176 			ifp->tokenized = tokenized;
2177 			addrconf_dad_start(ifp);
2178 		}
2179 
2180 		if (ifp) {
2181 			int flags;
2182 			unsigned long now;
2183 			struct inet6_ifaddr *ift;
2184 			u32 stored_lft;
2185 
2186 			/* update lifetime (RFC2462 5.5.3 e) */
2187 			spin_lock(&ifp->lock);
2188 			now = jiffies;
2189 			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2190 				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2191 			else
2192 				stored_lft = 0;
2193 			if (!update_lft && !create && stored_lft) {
2194 				const u32 minimum_lft = min(
2195 					stored_lft, (u32)MIN_VALID_LIFETIME);
2196 				valid_lft = max(valid_lft, minimum_lft);
2197 
2198 				/* RFC4862 Section 5.5.3e:
2199 				 * "Note that the preferred lifetime of the
2200 				 *  corresponding address is always reset to
2201 				 *  the Preferred Lifetime in the received
2202 				 *  Prefix Information option, regardless of
2203 				 *  whether the valid lifetime is also reset or
2204 				 *  ignored."
2205 				 *
2206 				 * So we should always update prefered_lft here.
2207 				 */
2208 				update_lft = 1;
2209 			}
2210 
2211 			if (update_lft) {
2212 				ifp->valid_lft = valid_lft;
2213 				ifp->prefered_lft = prefered_lft;
2214 				ifp->tstamp = now;
2215 				flags = ifp->flags;
2216 				ifp->flags &= ~IFA_F_DEPRECATED;
2217 				spin_unlock(&ifp->lock);
2218 
2219 				if (!(flags&IFA_F_TENTATIVE))
2220 					ipv6_ifa_notify(0, ifp);
2221 			} else
2222 				spin_unlock(&ifp->lock);
2223 
2224 			read_lock_bh(&in6_dev->lock);
2225 			/* update all temporary addresses in the list */
2226 			list_for_each_entry(ift, &in6_dev->tempaddr_list,
2227 					    tmp_list) {
2228 				int age, max_valid, max_prefered;
2229 
2230 				if (ifp != ift->ifpub)
2231 					continue;
2232 
2233 				/*
2234 				 * RFC 4941 section 3.3:
2235 				 * If a received option will extend the lifetime
2236 				 * of a public address, the lifetimes of
2237 				 * temporary addresses should be extended,
2238 				 * subject to the overall constraint that no
2239 				 * temporary addresses should ever remain
2240 				 * "valid" or "preferred" for a time longer than
2241 				 * (TEMP_VALID_LIFETIME) or
2242 				 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2243 				 * respectively.
2244 				 */
2245 				age = (now - ift->cstamp) / HZ;
2246 				max_valid = in6_dev->cnf.temp_valid_lft - age;
2247 				if (max_valid < 0)
2248 					max_valid = 0;
2249 
2250 				max_prefered = in6_dev->cnf.temp_prefered_lft -
2251 					       in6_dev->cnf.max_desync_factor -
2252 					       age;
2253 				if (max_prefered < 0)
2254 					max_prefered = 0;
2255 
2256 				if (valid_lft > max_valid)
2257 					valid_lft = max_valid;
2258 
2259 				if (prefered_lft > max_prefered)
2260 					prefered_lft = max_prefered;
2261 
2262 				spin_lock(&ift->lock);
2263 				flags = ift->flags;
2264 				ift->valid_lft = valid_lft;
2265 				ift->prefered_lft = prefered_lft;
2266 				ift->tstamp = now;
2267 				if (prefered_lft > 0)
2268 					ift->flags &= ~IFA_F_DEPRECATED;
2269 
2270 				spin_unlock(&ift->lock);
2271 				if (!(flags&IFA_F_TENTATIVE))
2272 					ipv6_ifa_notify(0, ift);
2273 			}
2274 
2275 			if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2276 				/*
2277 				 * When a new public address is created as
2278 				 * described in [ADDRCONF], also create a new
2279 				 * temporary address. Also create a temporary
2280 				 * address if it's enabled but no temporary
2281 				 * address currently exists.
2282 				 */
2283 				read_unlock_bh(&in6_dev->lock);
2284 				ipv6_create_tempaddr(ifp, NULL);
2285 			} else {
2286 				read_unlock_bh(&in6_dev->lock);
2287 			}
2288 
2289 			in6_ifa_put(ifp);
2290 			addrconf_verify(0);
2291 		}
2292 	}
2293 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2294 	in6_dev_put(in6_dev);
2295 }
2296 
2297 /*
2298  *	Set destination address.
2299  *	Special case for SIT interfaces where we create a new "virtual"
2300  *	device.
2301  */
2302 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2303 {
2304 	struct in6_ifreq ireq;
2305 	struct net_device *dev;
2306 	int err = -EINVAL;
2307 
2308 	rtnl_lock();
2309 
2310 	err = -EFAULT;
2311 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2312 		goto err_exit;
2313 
2314 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2315 
2316 	err = -ENODEV;
2317 	if (dev == NULL)
2318 		goto err_exit;
2319 
2320 #if IS_ENABLED(CONFIG_IPV6_SIT)
2321 	if (dev->type == ARPHRD_SIT) {
2322 		const struct net_device_ops *ops = dev->netdev_ops;
2323 		struct ifreq ifr;
2324 		struct ip_tunnel_parm p;
2325 
2326 		err = -EADDRNOTAVAIL;
2327 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2328 			goto err_exit;
2329 
2330 		memset(&p, 0, sizeof(p));
2331 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2332 		p.iph.saddr = 0;
2333 		p.iph.version = 4;
2334 		p.iph.ihl = 5;
2335 		p.iph.protocol = IPPROTO_IPV6;
2336 		p.iph.ttl = 64;
2337 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2338 
2339 		if (ops->ndo_do_ioctl) {
2340 			mm_segment_t oldfs = get_fs();
2341 
2342 			set_fs(KERNEL_DS);
2343 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2344 			set_fs(oldfs);
2345 		} else
2346 			err = -EOPNOTSUPP;
2347 
2348 		if (err == 0) {
2349 			err = -ENOBUFS;
2350 			dev = __dev_get_by_name(net, p.name);
2351 			if (!dev)
2352 				goto err_exit;
2353 			err = dev_open(dev);
2354 		}
2355 	}
2356 #endif
2357 
2358 err_exit:
2359 	rtnl_unlock();
2360 	return err;
2361 }
2362 
2363 /*
2364  *	Manual configuration of address on an interface
2365  */
2366 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2367 			  const struct in6_addr *peer_pfx,
2368 			  unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2369 			  __u32 valid_lft)
2370 {
2371 	struct inet6_ifaddr *ifp;
2372 	struct inet6_dev *idev;
2373 	struct net_device *dev;
2374 	int scope;
2375 	u32 flags;
2376 	clock_t expires;
2377 	unsigned long timeout;
2378 
2379 	ASSERT_RTNL();
2380 
2381 	if (plen > 128)
2382 		return -EINVAL;
2383 
2384 	/* check the lifetime */
2385 	if (!valid_lft || prefered_lft > valid_lft)
2386 		return -EINVAL;
2387 
2388 	dev = __dev_get_by_index(net, ifindex);
2389 	if (!dev)
2390 		return -ENODEV;
2391 
2392 	idev = addrconf_add_dev(dev);
2393 	if (IS_ERR(idev))
2394 		return PTR_ERR(idev);
2395 
2396 	scope = ipv6_addr_scope(pfx);
2397 
2398 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2399 	if (addrconf_finite_timeout(timeout)) {
2400 		expires = jiffies_to_clock_t(timeout * HZ);
2401 		valid_lft = timeout;
2402 		flags = RTF_EXPIRES;
2403 	} else {
2404 		expires = 0;
2405 		flags = 0;
2406 		ifa_flags |= IFA_F_PERMANENT;
2407 	}
2408 
2409 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2410 	if (addrconf_finite_timeout(timeout)) {
2411 		if (timeout == 0)
2412 			ifa_flags |= IFA_F_DEPRECATED;
2413 		prefered_lft = timeout;
2414 	}
2415 
2416 	ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2417 			    valid_lft, prefered_lft);
2418 
2419 	if (!IS_ERR(ifp)) {
2420 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2421 				      expires, flags);
2422 		/*
2423 		 * Note that section 3.1 of RFC 4429 indicates
2424 		 * that the Optimistic flag should not be set for
2425 		 * manually configured addresses
2426 		 */
2427 		addrconf_dad_start(ifp);
2428 		in6_ifa_put(ifp);
2429 		addrconf_verify(0);
2430 		return 0;
2431 	}
2432 
2433 	return PTR_ERR(ifp);
2434 }
2435 
2436 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2437 			  unsigned int plen)
2438 {
2439 	struct inet6_ifaddr *ifp;
2440 	struct inet6_dev *idev;
2441 	struct net_device *dev;
2442 
2443 	if (plen > 128)
2444 		return -EINVAL;
2445 
2446 	dev = __dev_get_by_index(net, ifindex);
2447 	if (!dev)
2448 		return -ENODEV;
2449 
2450 	if ((idev = __in6_dev_get(dev)) == NULL)
2451 		return -ENXIO;
2452 
2453 	read_lock_bh(&idev->lock);
2454 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2455 		if (ifp->prefix_len == plen &&
2456 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2457 			in6_ifa_hold(ifp);
2458 			read_unlock_bh(&idev->lock);
2459 
2460 			ipv6_del_addr(ifp);
2461 			return 0;
2462 		}
2463 	}
2464 	read_unlock_bh(&idev->lock);
2465 	return -EADDRNOTAVAIL;
2466 }
2467 
2468 
2469 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2470 {
2471 	struct in6_ifreq ireq;
2472 	int err;
2473 
2474 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2475 		return -EPERM;
2476 
2477 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2478 		return -EFAULT;
2479 
2480 	rtnl_lock();
2481 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2482 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2483 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2484 	rtnl_unlock();
2485 	return err;
2486 }
2487 
2488 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2489 {
2490 	struct in6_ifreq ireq;
2491 	int err;
2492 
2493 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2494 		return -EPERM;
2495 
2496 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2497 		return -EFAULT;
2498 
2499 	rtnl_lock();
2500 	err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2501 			     ireq.ifr6_prefixlen);
2502 	rtnl_unlock();
2503 	return err;
2504 }
2505 
2506 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2507 		     int plen, int scope)
2508 {
2509 	struct inet6_ifaddr *ifp;
2510 
2511 	ifp = ipv6_add_addr(idev, addr, NULL, plen,
2512 			    scope, IFA_F_PERMANENT, 0, 0);
2513 	if (!IS_ERR(ifp)) {
2514 		spin_lock_bh(&ifp->lock);
2515 		ifp->flags &= ~IFA_F_TENTATIVE;
2516 		spin_unlock_bh(&ifp->lock);
2517 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2518 		in6_ifa_put(ifp);
2519 	}
2520 }
2521 
2522 #if IS_ENABLED(CONFIG_IPV6_SIT)
2523 static void sit_add_v4_addrs(struct inet6_dev *idev)
2524 {
2525 	struct in6_addr addr;
2526 	struct net_device *dev;
2527 	struct net *net = dev_net(idev->dev);
2528 	int scope, plen;
2529 	u32 pflags = 0;
2530 
2531 	ASSERT_RTNL();
2532 
2533 	memset(&addr, 0, sizeof(struct in6_addr));
2534 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2535 
2536 	if (idev->dev->flags&IFF_POINTOPOINT) {
2537 		addr.s6_addr32[0] = htonl(0xfe800000);
2538 		scope = IFA_LINK;
2539 		plen = 64;
2540 	} else {
2541 		scope = IPV6_ADDR_COMPATv4;
2542 		plen = 96;
2543 		pflags |= RTF_NONEXTHOP;
2544 	}
2545 
2546 	if (addr.s6_addr32[3]) {
2547 		add_addr(idev, &addr, plen, scope);
2548 		addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2549 		return;
2550 	}
2551 
2552 	for_each_netdev(net, dev) {
2553 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
2554 		if (in_dev && (dev->flags & IFF_UP)) {
2555 			struct in_ifaddr *ifa;
2556 
2557 			int flag = scope;
2558 
2559 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2560 
2561 				addr.s6_addr32[3] = ifa->ifa_local;
2562 
2563 				if (ifa->ifa_scope == RT_SCOPE_LINK)
2564 					continue;
2565 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2566 					if (idev->dev->flags&IFF_POINTOPOINT)
2567 						continue;
2568 					flag |= IFA_HOST;
2569 				}
2570 
2571 				add_addr(idev, &addr, plen, flag);
2572 				addrconf_prefix_route(&addr, plen, idev->dev, 0,
2573 						      pflags);
2574 			}
2575 		}
2576 	}
2577 }
2578 #endif
2579 
2580 static void init_loopback(struct net_device *dev)
2581 {
2582 	struct inet6_dev  *idev;
2583 	struct net_device *sp_dev;
2584 	struct inet6_ifaddr *sp_ifa;
2585 	struct rt6_info *sp_rt;
2586 
2587 	/* ::1 */
2588 
2589 	ASSERT_RTNL();
2590 
2591 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2592 		pr_debug("%s: add_dev failed\n", __func__);
2593 		return;
2594 	}
2595 
2596 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2597 
2598 	/* Add routes to other interface's IPv6 addresses */
2599 	for_each_netdev(dev_net(dev), sp_dev) {
2600 		if (!strcmp(sp_dev->name, dev->name))
2601 			continue;
2602 
2603 		idev = __in6_dev_get(sp_dev);
2604 		if (!idev)
2605 			continue;
2606 
2607 		read_lock_bh(&idev->lock);
2608 		list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2609 
2610 			if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2611 				continue;
2612 
2613 			if (sp_ifa->rt)
2614 				continue;
2615 
2616 			sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2617 
2618 			/* Failure cases are ignored */
2619 			if (!IS_ERR(sp_rt)) {
2620 				sp_ifa->rt = sp_rt;
2621 				ip6_ins_rt(sp_rt);
2622 			}
2623 		}
2624 		read_unlock_bh(&idev->lock);
2625 	}
2626 }
2627 
2628 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2629 {
2630 	struct inet6_ifaddr *ifp;
2631 	u32 addr_flags = IFA_F_PERMANENT;
2632 
2633 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2634 	if (idev->cnf.optimistic_dad &&
2635 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2636 		addr_flags |= IFA_F_OPTIMISTIC;
2637 #endif
2638 
2639 
2640 	ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2641 	if (!IS_ERR(ifp)) {
2642 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2643 		addrconf_dad_start(ifp);
2644 		in6_ifa_put(ifp);
2645 	}
2646 }
2647 
2648 static void addrconf_dev_config(struct net_device *dev)
2649 {
2650 	struct in6_addr addr;
2651 	struct inet6_dev *idev;
2652 
2653 	ASSERT_RTNL();
2654 
2655 	if ((dev->type != ARPHRD_ETHER) &&
2656 	    (dev->type != ARPHRD_FDDI) &&
2657 	    (dev->type != ARPHRD_ARCNET) &&
2658 	    (dev->type != ARPHRD_INFINIBAND) &&
2659 	    (dev->type != ARPHRD_IEEE802154) &&
2660 	    (dev->type != ARPHRD_IEEE1394) &&
2661 	    (dev->type != ARPHRD_TUNNEL6)) {
2662 		/* Alas, we support only Ethernet autoconfiguration. */
2663 		return;
2664 	}
2665 
2666 	idev = addrconf_add_dev(dev);
2667 	if (IS_ERR(idev))
2668 		return;
2669 
2670 	memset(&addr, 0, sizeof(struct in6_addr));
2671 	addr.s6_addr32[0] = htonl(0xFE800000);
2672 
2673 	if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2674 		addrconf_add_linklocal(idev, &addr);
2675 }
2676 
2677 #if IS_ENABLED(CONFIG_IPV6_SIT)
2678 static void addrconf_sit_config(struct net_device *dev)
2679 {
2680 	struct inet6_dev *idev;
2681 
2682 	ASSERT_RTNL();
2683 
2684 	/*
2685 	 * Configure the tunnel with one of our IPv4
2686 	 * addresses... we should configure all of
2687 	 * our v4 addrs in the tunnel
2688 	 */
2689 
2690 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2691 		pr_debug("%s: add_dev failed\n", __func__);
2692 		return;
2693 	}
2694 
2695 	if (dev->priv_flags & IFF_ISATAP) {
2696 		struct in6_addr addr;
2697 
2698 		ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2699 		if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2700 			addrconf_add_linklocal(idev, &addr);
2701 		return;
2702 	}
2703 
2704 	sit_add_v4_addrs(idev);
2705 
2706 	if (dev->flags&IFF_POINTOPOINT)
2707 		addrconf_add_mroute(dev);
2708 }
2709 #endif
2710 
2711 #if IS_ENABLED(CONFIG_NET_IPGRE)
2712 static void addrconf_gre_config(struct net_device *dev)
2713 {
2714 	struct inet6_dev *idev;
2715 	struct in6_addr addr;
2716 
2717 	ASSERT_RTNL();
2718 
2719 	if ((idev = ipv6_find_idev(dev)) == NULL) {
2720 		pr_debug("%s: add_dev failed\n", __func__);
2721 		return;
2722 	}
2723 
2724 	ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2725 	if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2726 		addrconf_add_linklocal(idev, &addr);
2727 }
2728 #endif
2729 
2730 static inline int
2731 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2732 {
2733 	struct in6_addr lladdr;
2734 
2735 	if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2736 		addrconf_add_linklocal(idev, &lladdr);
2737 		return 0;
2738 	}
2739 	return -1;
2740 }
2741 
2742 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2743 			   void *ptr)
2744 {
2745 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2746 	struct inet6_dev *idev = __in6_dev_get(dev);
2747 	int run_pending = 0;
2748 	int err;
2749 
2750 	switch (event) {
2751 	case NETDEV_REGISTER:
2752 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2753 			idev = ipv6_add_dev(dev);
2754 			if (!idev)
2755 				return notifier_from_errno(-ENOMEM);
2756 		}
2757 		break;
2758 
2759 	case NETDEV_UP:
2760 	case NETDEV_CHANGE:
2761 		if (dev->flags & IFF_SLAVE)
2762 			break;
2763 
2764 		if (event == NETDEV_UP) {
2765 			if (!addrconf_qdisc_ok(dev)) {
2766 				/* device is not ready yet. */
2767 				pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2768 					dev->name);
2769 				break;
2770 			}
2771 
2772 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
2773 				idev = ipv6_add_dev(dev);
2774 
2775 			if (idev) {
2776 				idev->if_flags |= IF_READY;
2777 				run_pending = 1;
2778 			}
2779 		} else {
2780 			if (!addrconf_qdisc_ok(dev)) {
2781 				/* device is still not ready. */
2782 				break;
2783 			}
2784 
2785 			if (idev) {
2786 				if (idev->if_flags & IF_READY)
2787 					/* device is already configured. */
2788 					break;
2789 				idev->if_flags |= IF_READY;
2790 			}
2791 
2792 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2793 				dev->name);
2794 
2795 			run_pending = 1;
2796 		}
2797 
2798 		switch (dev->type) {
2799 #if IS_ENABLED(CONFIG_IPV6_SIT)
2800 		case ARPHRD_SIT:
2801 			addrconf_sit_config(dev);
2802 			break;
2803 #endif
2804 #if IS_ENABLED(CONFIG_NET_IPGRE)
2805 		case ARPHRD_IPGRE:
2806 			addrconf_gre_config(dev);
2807 			break;
2808 #endif
2809 		case ARPHRD_LOOPBACK:
2810 			init_loopback(dev);
2811 			break;
2812 
2813 		default:
2814 			addrconf_dev_config(dev);
2815 			break;
2816 		}
2817 
2818 		if (idev) {
2819 			if (run_pending)
2820 				addrconf_dad_run(idev);
2821 
2822 			/*
2823 			 * If the MTU changed during the interface down,
2824 			 * when the interface up, the changed MTU must be
2825 			 * reflected in the idev as well as routers.
2826 			 */
2827 			if (idev->cnf.mtu6 != dev->mtu &&
2828 			    dev->mtu >= IPV6_MIN_MTU) {
2829 				rt6_mtu_change(dev, dev->mtu);
2830 				idev->cnf.mtu6 = dev->mtu;
2831 			}
2832 			idev->tstamp = jiffies;
2833 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
2834 
2835 			/*
2836 			 * If the changed mtu during down is lower than
2837 			 * IPV6_MIN_MTU stop IPv6 on this interface.
2838 			 */
2839 			if (dev->mtu < IPV6_MIN_MTU)
2840 				addrconf_ifdown(dev, 1);
2841 		}
2842 		break;
2843 
2844 	case NETDEV_CHANGEMTU:
2845 		if (idev && dev->mtu >= IPV6_MIN_MTU) {
2846 			rt6_mtu_change(dev, dev->mtu);
2847 			idev->cnf.mtu6 = dev->mtu;
2848 			break;
2849 		}
2850 
2851 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2852 			idev = ipv6_add_dev(dev);
2853 			if (idev)
2854 				break;
2855 		}
2856 
2857 		/*
2858 		 * MTU falled under IPV6_MIN_MTU.
2859 		 * Stop IPv6 on this interface.
2860 		 */
2861 
2862 	case NETDEV_DOWN:
2863 	case NETDEV_UNREGISTER:
2864 		/*
2865 		 *	Remove all addresses from this interface.
2866 		 */
2867 		addrconf_ifdown(dev, event != NETDEV_DOWN);
2868 		break;
2869 
2870 	case NETDEV_CHANGENAME:
2871 		if (idev) {
2872 			snmp6_unregister_dev(idev);
2873 			addrconf_sysctl_unregister(idev);
2874 			addrconf_sysctl_register(idev);
2875 			err = snmp6_register_dev(idev);
2876 			if (err)
2877 				return notifier_from_errno(err);
2878 		}
2879 		break;
2880 
2881 	case NETDEV_PRE_TYPE_CHANGE:
2882 	case NETDEV_POST_TYPE_CHANGE:
2883 		addrconf_type_change(dev, event);
2884 		break;
2885 	}
2886 
2887 	return NOTIFY_OK;
2888 }
2889 
2890 /*
2891  *	addrconf module should be notified of a device going up
2892  */
2893 static struct notifier_block ipv6_dev_notf = {
2894 	.notifier_call = addrconf_notify,
2895 };
2896 
2897 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2898 {
2899 	struct inet6_dev *idev;
2900 	ASSERT_RTNL();
2901 
2902 	idev = __in6_dev_get(dev);
2903 
2904 	if (event == NETDEV_POST_TYPE_CHANGE)
2905 		ipv6_mc_remap(idev);
2906 	else if (event == NETDEV_PRE_TYPE_CHANGE)
2907 		ipv6_mc_unmap(idev);
2908 }
2909 
2910 static int addrconf_ifdown(struct net_device *dev, int how)
2911 {
2912 	struct net *net = dev_net(dev);
2913 	struct inet6_dev *idev;
2914 	struct inet6_ifaddr *ifa;
2915 	int state, i;
2916 
2917 	ASSERT_RTNL();
2918 
2919 	rt6_ifdown(net, dev);
2920 	neigh_ifdown(&nd_tbl, dev);
2921 
2922 	idev = __in6_dev_get(dev);
2923 	if (idev == NULL)
2924 		return -ENODEV;
2925 
2926 	/*
2927 	 * Step 1: remove reference to ipv6 device from parent device.
2928 	 *	   Do not dev_put!
2929 	 */
2930 	if (how) {
2931 		idev->dead = 1;
2932 
2933 		/* protected by rtnl_lock */
2934 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2935 
2936 		/* Step 1.5: remove snmp6 entry */
2937 		snmp6_unregister_dev(idev);
2938 
2939 	}
2940 
2941 	/* Step 2: clear hash table */
2942 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2943 		struct hlist_head *h = &inet6_addr_lst[i];
2944 
2945 		spin_lock_bh(&addrconf_hash_lock);
2946 	restart:
2947 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2948 			if (ifa->idev == idev) {
2949 				hlist_del_init_rcu(&ifa->addr_lst);
2950 				addrconf_del_dad_timer(ifa);
2951 				goto restart;
2952 			}
2953 		}
2954 		spin_unlock_bh(&addrconf_hash_lock);
2955 	}
2956 
2957 	write_lock_bh(&idev->lock);
2958 
2959 	addrconf_del_rs_timer(idev);
2960 
2961 	/* Step 2: clear flags for stateless addrconf */
2962 	if (!how)
2963 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2964 
2965 	if (how && del_timer(&idev->regen_timer))
2966 		in6_dev_put(idev);
2967 
2968 	/* Step 3: clear tempaddr list */
2969 	while (!list_empty(&idev->tempaddr_list)) {
2970 		ifa = list_first_entry(&idev->tempaddr_list,
2971 				       struct inet6_ifaddr, tmp_list);
2972 		list_del(&ifa->tmp_list);
2973 		write_unlock_bh(&idev->lock);
2974 		spin_lock_bh(&ifa->lock);
2975 
2976 		if (ifa->ifpub) {
2977 			in6_ifa_put(ifa->ifpub);
2978 			ifa->ifpub = NULL;
2979 		}
2980 		spin_unlock_bh(&ifa->lock);
2981 		in6_ifa_put(ifa);
2982 		write_lock_bh(&idev->lock);
2983 	}
2984 
2985 	while (!list_empty(&idev->addr_list)) {
2986 		ifa = list_first_entry(&idev->addr_list,
2987 				       struct inet6_ifaddr, if_list);
2988 		addrconf_del_dad_timer(ifa);
2989 
2990 		list_del(&ifa->if_list);
2991 
2992 		write_unlock_bh(&idev->lock);
2993 
2994 		spin_lock_bh(&ifa->state_lock);
2995 		state = ifa->state;
2996 		ifa->state = INET6_IFADDR_STATE_DEAD;
2997 		spin_unlock_bh(&ifa->state_lock);
2998 
2999 		if (state != INET6_IFADDR_STATE_DEAD) {
3000 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3001 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3002 		}
3003 		in6_ifa_put(ifa);
3004 
3005 		write_lock_bh(&idev->lock);
3006 	}
3007 
3008 	write_unlock_bh(&idev->lock);
3009 
3010 	/* Step 5: Discard multicast list */
3011 	if (how)
3012 		ipv6_mc_destroy_dev(idev);
3013 	else
3014 		ipv6_mc_down(idev);
3015 
3016 	idev->tstamp = jiffies;
3017 
3018 	/* Last: Shot the device (if unregistered) */
3019 	if (how) {
3020 		addrconf_sysctl_unregister(idev);
3021 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3022 		neigh_ifdown(&nd_tbl, dev);
3023 		in6_dev_put(idev);
3024 	}
3025 	return 0;
3026 }
3027 
3028 static void addrconf_rs_timer(unsigned long data)
3029 {
3030 	struct inet6_dev *idev = (struct inet6_dev *)data;
3031 	struct net_device *dev = idev->dev;
3032 	struct in6_addr lladdr;
3033 
3034 	write_lock(&idev->lock);
3035 	if (idev->dead || !(idev->if_flags & IF_READY))
3036 		goto out;
3037 
3038 	if (!ipv6_accept_ra(idev))
3039 		goto out;
3040 
3041 	/* Announcement received after solicitation was sent */
3042 	if (idev->if_flags & IF_RA_RCVD)
3043 		goto out;
3044 
3045 	if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3046 		write_unlock(&idev->lock);
3047 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3048 			ndisc_send_rs(dev, &lladdr,
3049 				      &in6addr_linklocal_allrouters);
3050 		else
3051 			goto put;
3052 
3053 		write_lock(&idev->lock);
3054 		/* The wait after the last probe can be shorter */
3055 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3056 					     idev->cnf.rtr_solicits) ?
3057 				      idev->cnf.rtr_solicit_delay :
3058 				      idev->cnf.rtr_solicit_interval);
3059 	} else {
3060 		/*
3061 		 * Note: we do not support deprecated "all on-link"
3062 		 * assumption any longer.
3063 		 */
3064 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3065 	}
3066 
3067 out:
3068 	write_unlock(&idev->lock);
3069 put:
3070 	in6_dev_put(idev);
3071 }
3072 
3073 /*
3074  *	Duplicate Address Detection
3075  */
3076 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3077 {
3078 	unsigned long rand_num;
3079 	struct inet6_dev *idev = ifp->idev;
3080 
3081 	if (ifp->flags & IFA_F_OPTIMISTIC)
3082 		rand_num = 0;
3083 	else
3084 		rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3085 
3086 	ifp->dad_probes = idev->cnf.dad_transmits;
3087 	addrconf_mod_dad_timer(ifp, rand_num);
3088 }
3089 
3090 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3091 {
3092 	struct inet6_dev *idev = ifp->idev;
3093 	struct net_device *dev = idev->dev;
3094 
3095 	addrconf_join_solict(dev, &ifp->addr);
3096 
3097 	net_srandom(ifp->addr.s6_addr32[3]);
3098 
3099 	read_lock_bh(&idev->lock);
3100 	spin_lock(&ifp->lock);
3101 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3102 		goto out;
3103 
3104 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3105 	    idev->cnf.accept_dad < 1 ||
3106 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3107 	    ifp->flags & IFA_F_NODAD) {
3108 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3109 		spin_unlock(&ifp->lock);
3110 		read_unlock_bh(&idev->lock);
3111 
3112 		addrconf_dad_completed(ifp);
3113 		return;
3114 	}
3115 
3116 	if (!(idev->if_flags & IF_READY)) {
3117 		spin_unlock(&ifp->lock);
3118 		read_unlock_bh(&idev->lock);
3119 		/*
3120 		 * If the device is not ready:
3121 		 * - keep it tentative if it is a permanent address.
3122 		 * - otherwise, kill it.
3123 		 */
3124 		in6_ifa_hold(ifp);
3125 		addrconf_dad_stop(ifp, 0);
3126 		return;
3127 	}
3128 
3129 	/*
3130 	 * Optimistic nodes can start receiving
3131 	 * Frames right away
3132 	 */
3133 	if (ifp->flags & IFA_F_OPTIMISTIC)
3134 		ip6_ins_rt(ifp->rt);
3135 
3136 	addrconf_dad_kick(ifp);
3137 out:
3138 	spin_unlock(&ifp->lock);
3139 	read_unlock_bh(&idev->lock);
3140 }
3141 
3142 static void addrconf_dad_timer(unsigned long data)
3143 {
3144 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3145 	struct inet6_dev *idev = ifp->idev;
3146 	struct in6_addr mcaddr;
3147 
3148 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
3149 		goto out;
3150 
3151 	write_lock(&idev->lock);
3152 	if (idev->dead || !(idev->if_flags & IF_READY)) {
3153 		write_unlock(&idev->lock);
3154 		goto out;
3155 	}
3156 
3157 	spin_lock(&ifp->lock);
3158 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3159 		spin_unlock(&ifp->lock);
3160 		write_unlock(&idev->lock);
3161 		goto out;
3162 	}
3163 
3164 	if (ifp->dad_probes == 0) {
3165 		/*
3166 		 * DAD was successful
3167 		 */
3168 
3169 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3170 		spin_unlock(&ifp->lock);
3171 		write_unlock(&idev->lock);
3172 
3173 		addrconf_dad_completed(ifp);
3174 
3175 		goto out;
3176 	}
3177 
3178 	ifp->dad_probes--;
3179 	addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3180 	spin_unlock(&ifp->lock);
3181 	write_unlock(&idev->lock);
3182 
3183 	/* send a neighbour solicitation for our addr */
3184 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3185 	ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3186 out:
3187 	in6_ifa_put(ifp);
3188 }
3189 
3190 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3191 {
3192 	struct net_device *dev = ifp->idev->dev;
3193 	struct in6_addr lladdr;
3194 	bool send_rs, send_mld;
3195 
3196 	addrconf_del_dad_timer(ifp);
3197 
3198 	/*
3199 	 *	Configure the address for reception. Now it is valid.
3200 	 */
3201 
3202 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
3203 
3204 	/* If added prefix is link local and we are prepared to process
3205 	   router advertisements, start sending router solicitations.
3206 	 */
3207 
3208 	read_lock_bh(&ifp->idev->lock);
3209 	spin_lock(&ifp->lock);
3210 	send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3211 		   ifp->idev->valid_ll_addr_cnt == 1;
3212 	send_rs = send_mld &&
3213 		  ipv6_accept_ra(ifp->idev) &&
3214 		  ifp->idev->cnf.rtr_solicits > 0 &&
3215 		  (dev->flags&IFF_LOOPBACK) == 0;
3216 	spin_unlock(&ifp->lock);
3217 	read_unlock_bh(&ifp->idev->lock);
3218 
3219 	/* While dad is in progress mld report's source address is in6_addrany.
3220 	 * Resend with proper ll now.
3221 	 */
3222 	if (send_mld)
3223 		ipv6_mc_dad_complete(ifp->idev);
3224 
3225 	if (send_rs) {
3226 		/*
3227 		 *	If a host as already performed a random delay
3228 		 *	[...] as part of DAD [...] there is no need
3229 		 *	to delay again before sending the first RS
3230 		 */
3231 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3232 			return;
3233 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3234 
3235 		write_lock_bh(&ifp->idev->lock);
3236 		spin_lock(&ifp->lock);
3237 		ifp->idev->rs_probes = 1;
3238 		ifp->idev->if_flags |= IF_RS_SENT;
3239 		addrconf_mod_rs_timer(ifp->idev,
3240 				      ifp->idev->cnf.rtr_solicit_interval);
3241 		spin_unlock(&ifp->lock);
3242 		write_unlock_bh(&ifp->idev->lock);
3243 	}
3244 }
3245 
3246 static void addrconf_dad_run(struct inet6_dev *idev)
3247 {
3248 	struct inet6_ifaddr *ifp;
3249 
3250 	read_lock_bh(&idev->lock);
3251 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3252 		spin_lock(&ifp->lock);
3253 		if (ifp->flags & IFA_F_TENTATIVE &&
3254 		    ifp->state == INET6_IFADDR_STATE_DAD)
3255 			addrconf_dad_kick(ifp);
3256 		spin_unlock(&ifp->lock);
3257 	}
3258 	read_unlock_bh(&idev->lock);
3259 }
3260 
3261 #ifdef CONFIG_PROC_FS
3262 struct if6_iter_state {
3263 	struct seq_net_private p;
3264 	int bucket;
3265 	int offset;
3266 };
3267 
3268 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3269 {
3270 	struct inet6_ifaddr *ifa = NULL;
3271 	struct if6_iter_state *state = seq->private;
3272 	struct net *net = seq_file_net(seq);
3273 	int p = 0;
3274 
3275 	/* initial bucket if pos is 0 */
3276 	if (pos == 0) {
3277 		state->bucket = 0;
3278 		state->offset = 0;
3279 	}
3280 
3281 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3282 		hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3283 					 addr_lst) {
3284 			if (!net_eq(dev_net(ifa->idev->dev), net))
3285 				continue;
3286 			/* sync with offset */
3287 			if (p < state->offset) {
3288 				p++;
3289 				continue;
3290 			}
3291 			state->offset++;
3292 			return ifa;
3293 		}
3294 
3295 		/* prepare for next bucket */
3296 		state->offset = 0;
3297 		p = 0;
3298 	}
3299 	return NULL;
3300 }
3301 
3302 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3303 					 struct inet6_ifaddr *ifa)
3304 {
3305 	struct if6_iter_state *state = seq->private;
3306 	struct net *net = seq_file_net(seq);
3307 
3308 	hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3309 		if (!net_eq(dev_net(ifa->idev->dev), net))
3310 			continue;
3311 		state->offset++;
3312 		return ifa;
3313 	}
3314 
3315 	while (++state->bucket < IN6_ADDR_HSIZE) {
3316 		state->offset = 0;
3317 		hlist_for_each_entry_rcu_bh(ifa,
3318 				     &inet6_addr_lst[state->bucket], addr_lst) {
3319 			if (!net_eq(dev_net(ifa->idev->dev), net))
3320 				continue;
3321 			state->offset++;
3322 			return ifa;
3323 		}
3324 	}
3325 
3326 	return NULL;
3327 }
3328 
3329 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3330 	__acquires(rcu_bh)
3331 {
3332 	rcu_read_lock_bh();
3333 	return if6_get_first(seq, *pos);
3334 }
3335 
3336 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3337 {
3338 	struct inet6_ifaddr *ifa;
3339 
3340 	ifa = if6_get_next(seq, v);
3341 	++*pos;
3342 	return ifa;
3343 }
3344 
3345 static void if6_seq_stop(struct seq_file *seq, void *v)
3346 	__releases(rcu_bh)
3347 {
3348 	rcu_read_unlock_bh();
3349 }
3350 
3351 static int if6_seq_show(struct seq_file *seq, void *v)
3352 {
3353 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3354 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3355 		   &ifp->addr,
3356 		   ifp->idev->dev->ifindex,
3357 		   ifp->prefix_len,
3358 		   ifp->scope,
3359 		   ifp->flags,
3360 		   ifp->idev->dev->name);
3361 	return 0;
3362 }
3363 
3364 static const struct seq_operations if6_seq_ops = {
3365 	.start	= if6_seq_start,
3366 	.next	= if6_seq_next,
3367 	.show	= if6_seq_show,
3368 	.stop	= if6_seq_stop,
3369 };
3370 
3371 static int if6_seq_open(struct inode *inode, struct file *file)
3372 {
3373 	return seq_open_net(inode, file, &if6_seq_ops,
3374 			    sizeof(struct if6_iter_state));
3375 }
3376 
3377 static const struct file_operations if6_fops = {
3378 	.owner		= THIS_MODULE,
3379 	.open		= if6_seq_open,
3380 	.read		= seq_read,
3381 	.llseek		= seq_lseek,
3382 	.release	= seq_release_net,
3383 };
3384 
3385 static int __net_init if6_proc_net_init(struct net *net)
3386 {
3387 	if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3388 		return -ENOMEM;
3389 	return 0;
3390 }
3391 
3392 static void __net_exit if6_proc_net_exit(struct net *net)
3393 {
3394 	remove_proc_entry("if_inet6", net->proc_net);
3395 }
3396 
3397 static struct pernet_operations if6_proc_net_ops = {
3398        .init = if6_proc_net_init,
3399        .exit = if6_proc_net_exit,
3400 };
3401 
3402 int __init if6_proc_init(void)
3403 {
3404 	return register_pernet_subsys(&if6_proc_net_ops);
3405 }
3406 
3407 void if6_proc_exit(void)
3408 {
3409 	unregister_pernet_subsys(&if6_proc_net_ops);
3410 }
3411 #endif	/* CONFIG_PROC_FS */
3412 
3413 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3414 /* Check if address is a home address configured on any interface. */
3415 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3416 {
3417 	int ret = 0;
3418 	struct inet6_ifaddr *ifp = NULL;
3419 	unsigned int hash = inet6_addr_hash(addr);
3420 
3421 	rcu_read_lock_bh();
3422 	hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3423 		if (!net_eq(dev_net(ifp->idev->dev), net))
3424 			continue;
3425 		if (ipv6_addr_equal(&ifp->addr, addr) &&
3426 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3427 			ret = 1;
3428 			break;
3429 		}
3430 	}
3431 	rcu_read_unlock_bh();
3432 	return ret;
3433 }
3434 #endif
3435 
3436 /*
3437  *	Periodic address status verification
3438  */
3439 
3440 static void addrconf_verify(unsigned long foo)
3441 {
3442 	unsigned long now, next, next_sec, next_sched;
3443 	struct inet6_ifaddr *ifp;
3444 	int i;
3445 
3446 	rcu_read_lock_bh();
3447 	spin_lock(&addrconf_verify_lock);
3448 	now = jiffies;
3449 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3450 
3451 	del_timer(&addr_chk_timer);
3452 
3453 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3454 restart:
3455 		hlist_for_each_entry_rcu_bh(ifp,
3456 					 &inet6_addr_lst[i], addr_lst) {
3457 			unsigned long age;
3458 
3459 			if (ifp->flags & IFA_F_PERMANENT)
3460 				continue;
3461 
3462 			spin_lock(&ifp->lock);
3463 			/* We try to batch several events at once. */
3464 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3465 
3466 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3467 			    age >= ifp->valid_lft) {
3468 				spin_unlock(&ifp->lock);
3469 				in6_ifa_hold(ifp);
3470 				ipv6_del_addr(ifp);
3471 				goto restart;
3472 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3473 				spin_unlock(&ifp->lock);
3474 				continue;
3475 			} else if (age >= ifp->prefered_lft) {
3476 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3477 				int deprecate = 0;
3478 
3479 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3480 					deprecate = 1;
3481 					ifp->flags |= IFA_F_DEPRECATED;
3482 				}
3483 
3484 				if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3485 					next = ifp->tstamp + ifp->valid_lft * HZ;
3486 
3487 				spin_unlock(&ifp->lock);
3488 
3489 				if (deprecate) {
3490 					in6_ifa_hold(ifp);
3491 
3492 					ipv6_ifa_notify(0, ifp);
3493 					in6_ifa_put(ifp);
3494 					goto restart;
3495 				}
3496 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3497 				   !(ifp->flags&IFA_F_TENTATIVE)) {
3498 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3499 					ifp->idev->cnf.dad_transmits *
3500 					ifp->idev->nd_parms->retrans_time / HZ;
3501 
3502 				if (age >= ifp->prefered_lft - regen_advance) {
3503 					struct inet6_ifaddr *ifpub = ifp->ifpub;
3504 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3505 						next = ifp->tstamp + ifp->prefered_lft * HZ;
3506 					if (!ifp->regen_count && ifpub) {
3507 						ifp->regen_count++;
3508 						in6_ifa_hold(ifp);
3509 						in6_ifa_hold(ifpub);
3510 						spin_unlock(&ifp->lock);
3511 
3512 						spin_lock(&ifpub->lock);
3513 						ifpub->regen_count = 0;
3514 						spin_unlock(&ifpub->lock);
3515 						ipv6_create_tempaddr(ifpub, ifp);
3516 						in6_ifa_put(ifpub);
3517 						in6_ifa_put(ifp);
3518 						goto restart;
3519 					}
3520 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3521 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3522 				spin_unlock(&ifp->lock);
3523 			} else {
3524 				/* ifp->prefered_lft <= ifp->valid_lft */
3525 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3526 					next = ifp->tstamp + ifp->prefered_lft * HZ;
3527 				spin_unlock(&ifp->lock);
3528 			}
3529 		}
3530 	}
3531 
3532 	next_sec = round_jiffies_up(next);
3533 	next_sched = next;
3534 
3535 	/* If rounded timeout is accurate enough, accept it. */
3536 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3537 		next_sched = next_sec;
3538 
3539 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3540 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3541 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3542 
3543 	ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3544 	      now, next, next_sec, next_sched);
3545 
3546 	addr_chk_timer.expires = next_sched;
3547 	add_timer(&addr_chk_timer);
3548 	spin_unlock(&addrconf_verify_lock);
3549 	rcu_read_unlock_bh();
3550 }
3551 
3552 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3553 				     struct in6_addr **peer_pfx)
3554 {
3555 	struct in6_addr *pfx = NULL;
3556 
3557 	*peer_pfx = NULL;
3558 
3559 	if (addr)
3560 		pfx = nla_data(addr);
3561 
3562 	if (local) {
3563 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3564 			*peer_pfx = pfx;
3565 		pfx = nla_data(local);
3566 	}
3567 
3568 	return pfx;
3569 }
3570 
3571 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3572 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
3573 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
3574 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
3575 };
3576 
3577 static int
3578 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3579 {
3580 	struct net *net = sock_net(skb->sk);
3581 	struct ifaddrmsg *ifm;
3582 	struct nlattr *tb[IFA_MAX+1];
3583 	struct in6_addr *pfx, *peer_pfx;
3584 	int err;
3585 
3586 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3587 	if (err < 0)
3588 		return err;
3589 
3590 	ifm = nlmsg_data(nlh);
3591 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3592 	if (pfx == NULL)
3593 		return -EINVAL;
3594 
3595 	return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3596 }
3597 
3598 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3599 			     u32 prefered_lft, u32 valid_lft)
3600 {
3601 	u32 flags;
3602 	clock_t expires;
3603 	unsigned long timeout;
3604 
3605 	if (!valid_lft || (prefered_lft > valid_lft))
3606 		return -EINVAL;
3607 
3608 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
3609 	if (addrconf_finite_timeout(timeout)) {
3610 		expires = jiffies_to_clock_t(timeout * HZ);
3611 		valid_lft = timeout;
3612 		flags = RTF_EXPIRES;
3613 	} else {
3614 		expires = 0;
3615 		flags = 0;
3616 		ifa_flags |= IFA_F_PERMANENT;
3617 	}
3618 
3619 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3620 	if (addrconf_finite_timeout(timeout)) {
3621 		if (timeout == 0)
3622 			ifa_flags |= IFA_F_DEPRECATED;
3623 		prefered_lft = timeout;
3624 	}
3625 
3626 	spin_lock_bh(&ifp->lock);
3627 	ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3628 	ifp->tstamp = jiffies;
3629 	ifp->valid_lft = valid_lft;
3630 	ifp->prefered_lft = prefered_lft;
3631 
3632 	spin_unlock_bh(&ifp->lock);
3633 	if (!(ifp->flags&IFA_F_TENTATIVE))
3634 		ipv6_ifa_notify(0, ifp);
3635 
3636 	addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3637 			      expires, flags);
3638 	addrconf_verify(0);
3639 
3640 	return 0;
3641 }
3642 
3643 static int
3644 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3645 {
3646 	struct net *net = sock_net(skb->sk);
3647 	struct ifaddrmsg *ifm;
3648 	struct nlattr *tb[IFA_MAX+1];
3649 	struct in6_addr *pfx, *peer_pfx;
3650 	struct inet6_ifaddr *ifa;
3651 	struct net_device *dev;
3652 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3653 	u8 ifa_flags;
3654 	int err;
3655 
3656 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3657 	if (err < 0)
3658 		return err;
3659 
3660 	ifm = nlmsg_data(nlh);
3661 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3662 	if (pfx == NULL)
3663 		return -EINVAL;
3664 
3665 	if (tb[IFA_CACHEINFO]) {
3666 		struct ifa_cacheinfo *ci;
3667 
3668 		ci = nla_data(tb[IFA_CACHEINFO]);
3669 		valid_lft = ci->ifa_valid;
3670 		preferred_lft = ci->ifa_prefered;
3671 	} else {
3672 		preferred_lft = INFINITY_LIFE_TIME;
3673 		valid_lft = INFINITY_LIFE_TIME;
3674 	}
3675 
3676 	dev =  __dev_get_by_index(net, ifm->ifa_index);
3677 	if (dev == NULL)
3678 		return -ENODEV;
3679 
3680 	/* We ignore other flags so far. */
3681 	ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3682 
3683 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3684 	if (ifa == NULL) {
3685 		/*
3686 		 * It would be best to check for !NLM_F_CREATE here but
3687 		 * userspace alreay relies on not having to provide this.
3688 		 */
3689 		return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3690 				      ifm->ifa_prefixlen, ifa_flags,
3691 				      preferred_lft, valid_lft);
3692 	}
3693 
3694 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
3695 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
3696 		err = -EEXIST;
3697 	else
3698 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3699 
3700 	in6_ifa_put(ifa);
3701 
3702 	return err;
3703 }
3704 
3705 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3706 			  u8 scope, int ifindex)
3707 {
3708 	struct ifaddrmsg *ifm;
3709 
3710 	ifm = nlmsg_data(nlh);
3711 	ifm->ifa_family = AF_INET6;
3712 	ifm->ifa_prefixlen = prefixlen;
3713 	ifm->ifa_flags = flags;
3714 	ifm->ifa_scope = scope;
3715 	ifm->ifa_index = ifindex;
3716 }
3717 
3718 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3719 			 unsigned long tstamp, u32 preferred, u32 valid)
3720 {
3721 	struct ifa_cacheinfo ci;
3722 
3723 	ci.cstamp = cstamp_delta(cstamp);
3724 	ci.tstamp = cstamp_delta(tstamp);
3725 	ci.ifa_prefered = preferred;
3726 	ci.ifa_valid = valid;
3727 
3728 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3729 }
3730 
3731 static inline int rt_scope(int ifa_scope)
3732 {
3733 	if (ifa_scope & IFA_HOST)
3734 		return RT_SCOPE_HOST;
3735 	else if (ifa_scope & IFA_LINK)
3736 		return RT_SCOPE_LINK;
3737 	else if (ifa_scope & IFA_SITE)
3738 		return RT_SCOPE_SITE;
3739 	else
3740 		return RT_SCOPE_UNIVERSE;
3741 }
3742 
3743 static inline int inet6_ifaddr_msgsize(void)
3744 {
3745 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3746 	       + nla_total_size(16) /* IFA_LOCAL */
3747 	       + nla_total_size(16) /* IFA_ADDRESS */
3748 	       + nla_total_size(sizeof(struct ifa_cacheinfo));
3749 }
3750 
3751 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3752 			     u32 portid, u32 seq, int event, unsigned int flags)
3753 {
3754 	struct nlmsghdr  *nlh;
3755 	u32 preferred, valid;
3756 
3757 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3758 	if (nlh == NULL)
3759 		return -EMSGSIZE;
3760 
3761 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3762 		      ifa->idev->dev->ifindex);
3763 
3764 	if (!(ifa->flags&IFA_F_PERMANENT)) {
3765 		preferred = ifa->prefered_lft;
3766 		valid = ifa->valid_lft;
3767 		if (preferred != INFINITY_LIFE_TIME) {
3768 			long tval = (jiffies - ifa->tstamp)/HZ;
3769 			if (preferred > tval)
3770 				preferred -= tval;
3771 			else
3772 				preferred = 0;
3773 			if (valid != INFINITY_LIFE_TIME) {
3774 				if (valid > tval)
3775 					valid -= tval;
3776 				else
3777 					valid = 0;
3778 			}
3779 		}
3780 	} else {
3781 		preferred = INFINITY_LIFE_TIME;
3782 		valid = INFINITY_LIFE_TIME;
3783 	}
3784 
3785 	if (!ipv6_addr_any(&ifa->peer_addr)) {
3786 		if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3787 		    nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3788 			goto error;
3789 	} else
3790 		if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3791 			goto error;
3792 
3793 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3794 		goto error;
3795 
3796 	return nlmsg_end(skb, nlh);
3797 
3798 error:
3799 	nlmsg_cancel(skb, nlh);
3800 	return -EMSGSIZE;
3801 }
3802 
3803 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3804 				u32 portid, u32 seq, int event, u16 flags)
3805 {
3806 	struct nlmsghdr  *nlh;
3807 	u8 scope = RT_SCOPE_UNIVERSE;
3808 	int ifindex = ifmca->idev->dev->ifindex;
3809 
3810 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3811 		scope = RT_SCOPE_SITE;
3812 
3813 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3814 	if (nlh == NULL)
3815 		return -EMSGSIZE;
3816 
3817 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3818 	if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3819 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3820 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3821 		nlmsg_cancel(skb, nlh);
3822 		return -EMSGSIZE;
3823 	}
3824 
3825 	return nlmsg_end(skb, nlh);
3826 }
3827 
3828 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3829 				u32 portid, u32 seq, int event, unsigned int flags)
3830 {
3831 	struct nlmsghdr  *nlh;
3832 	u8 scope = RT_SCOPE_UNIVERSE;
3833 	int ifindex = ifaca->aca_idev->dev->ifindex;
3834 
3835 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3836 		scope = RT_SCOPE_SITE;
3837 
3838 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3839 	if (nlh == NULL)
3840 		return -EMSGSIZE;
3841 
3842 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3843 	if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3844 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3845 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3846 		nlmsg_cancel(skb, nlh);
3847 		return -EMSGSIZE;
3848 	}
3849 
3850 	return nlmsg_end(skb, nlh);
3851 }
3852 
3853 enum addr_type_t {
3854 	UNICAST_ADDR,
3855 	MULTICAST_ADDR,
3856 	ANYCAST_ADDR,
3857 };
3858 
3859 /* called with rcu_read_lock() */
3860 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3861 			  struct netlink_callback *cb, enum addr_type_t type,
3862 			  int s_ip_idx, int *p_ip_idx)
3863 {
3864 	struct ifmcaddr6 *ifmca;
3865 	struct ifacaddr6 *ifaca;
3866 	int err = 1;
3867 	int ip_idx = *p_ip_idx;
3868 
3869 	read_lock_bh(&idev->lock);
3870 	switch (type) {
3871 	case UNICAST_ADDR: {
3872 		struct inet6_ifaddr *ifa;
3873 
3874 		/* unicast address incl. temp addr */
3875 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
3876 			if (++ip_idx < s_ip_idx)
3877 				continue;
3878 			err = inet6_fill_ifaddr(skb, ifa,
3879 						NETLINK_CB(cb->skb).portid,
3880 						cb->nlh->nlmsg_seq,
3881 						RTM_NEWADDR,
3882 						NLM_F_MULTI);
3883 			if (err <= 0)
3884 				break;
3885 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3886 		}
3887 		break;
3888 	}
3889 	case MULTICAST_ADDR:
3890 		/* multicast address */
3891 		for (ifmca = idev->mc_list; ifmca;
3892 		     ifmca = ifmca->next, ip_idx++) {
3893 			if (ip_idx < s_ip_idx)
3894 				continue;
3895 			err = inet6_fill_ifmcaddr(skb, ifmca,
3896 						  NETLINK_CB(cb->skb).portid,
3897 						  cb->nlh->nlmsg_seq,
3898 						  RTM_GETMULTICAST,
3899 						  NLM_F_MULTI);
3900 			if (err <= 0)
3901 				break;
3902 		}
3903 		break;
3904 	case ANYCAST_ADDR:
3905 		/* anycast address */
3906 		for (ifaca = idev->ac_list; ifaca;
3907 		     ifaca = ifaca->aca_next, ip_idx++) {
3908 			if (ip_idx < s_ip_idx)
3909 				continue;
3910 			err = inet6_fill_ifacaddr(skb, ifaca,
3911 						  NETLINK_CB(cb->skb).portid,
3912 						  cb->nlh->nlmsg_seq,
3913 						  RTM_GETANYCAST,
3914 						  NLM_F_MULTI);
3915 			if (err <= 0)
3916 				break;
3917 		}
3918 		break;
3919 	default:
3920 		break;
3921 	}
3922 	read_unlock_bh(&idev->lock);
3923 	*p_ip_idx = ip_idx;
3924 	return err;
3925 }
3926 
3927 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3928 			   enum addr_type_t type)
3929 {
3930 	struct net *net = sock_net(skb->sk);
3931 	int h, s_h;
3932 	int idx, ip_idx;
3933 	int s_idx, s_ip_idx;
3934 	struct net_device *dev;
3935 	struct inet6_dev *idev;
3936 	struct hlist_head *head;
3937 
3938 	s_h = cb->args[0];
3939 	s_idx = idx = cb->args[1];
3940 	s_ip_idx = ip_idx = cb->args[2];
3941 
3942 	rcu_read_lock();
3943 	cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3944 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3945 		idx = 0;
3946 		head = &net->dev_index_head[h];
3947 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
3948 			if (idx < s_idx)
3949 				goto cont;
3950 			if (h > s_h || idx > s_idx)
3951 				s_ip_idx = 0;
3952 			ip_idx = 0;
3953 			idev = __in6_dev_get(dev);
3954 			if (!idev)
3955 				goto cont;
3956 
3957 			if (in6_dump_addrs(idev, skb, cb, type,
3958 					   s_ip_idx, &ip_idx) <= 0)
3959 				goto done;
3960 cont:
3961 			idx++;
3962 		}
3963 	}
3964 done:
3965 	rcu_read_unlock();
3966 	cb->args[0] = h;
3967 	cb->args[1] = idx;
3968 	cb->args[2] = ip_idx;
3969 
3970 	return skb->len;
3971 }
3972 
3973 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3974 {
3975 	enum addr_type_t type = UNICAST_ADDR;
3976 
3977 	return inet6_dump_addr(skb, cb, type);
3978 }
3979 
3980 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3981 {
3982 	enum addr_type_t type = MULTICAST_ADDR;
3983 
3984 	return inet6_dump_addr(skb, cb, type);
3985 }
3986 
3987 
3988 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3989 {
3990 	enum addr_type_t type = ANYCAST_ADDR;
3991 
3992 	return inet6_dump_addr(skb, cb, type);
3993 }
3994 
3995 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3996 {
3997 	struct net *net = sock_net(in_skb->sk);
3998 	struct ifaddrmsg *ifm;
3999 	struct nlattr *tb[IFA_MAX+1];
4000 	struct in6_addr *addr = NULL, *peer;
4001 	struct net_device *dev = NULL;
4002 	struct inet6_ifaddr *ifa;
4003 	struct sk_buff *skb;
4004 	int err;
4005 
4006 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4007 	if (err < 0)
4008 		goto errout;
4009 
4010 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4011 	if (addr == NULL) {
4012 		err = -EINVAL;
4013 		goto errout;
4014 	}
4015 
4016 	ifm = nlmsg_data(nlh);
4017 	if (ifm->ifa_index)
4018 		dev = __dev_get_by_index(net, ifm->ifa_index);
4019 
4020 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4021 	if (!ifa) {
4022 		err = -EADDRNOTAVAIL;
4023 		goto errout;
4024 	}
4025 
4026 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4027 	if (!skb) {
4028 		err = -ENOBUFS;
4029 		goto errout_ifa;
4030 	}
4031 
4032 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4033 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
4034 	if (err < 0) {
4035 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4036 		WARN_ON(err == -EMSGSIZE);
4037 		kfree_skb(skb);
4038 		goto errout_ifa;
4039 	}
4040 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4041 errout_ifa:
4042 	in6_ifa_put(ifa);
4043 errout:
4044 	return err;
4045 }
4046 
4047 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4048 {
4049 	struct sk_buff *skb;
4050 	struct net *net = dev_net(ifa->idev->dev);
4051 	int err = -ENOBUFS;
4052 
4053 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4054 	if (skb == NULL)
4055 		goto errout;
4056 
4057 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4058 	if (err < 0) {
4059 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4060 		WARN_ON(err == -EMSGSIZE);
4061 		kfree_skb(skb);
4062 		goto errout;
4063 	}
4064 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4065 	return;
4066 errout:
4067 	if (err < 0)
4068 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4069 }
4070 
4071 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4072 				__s32 *array, int bytes)
4073 {
4074 	BUG_ON(bytes < (DEVCONF_MAX * 4));
4075 
4076 	memset(array, 0, bytes);
4077 	array[DEVCONF_FORWARDING] = cnf->forwarding;
4078 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4079 	array[DEVCONF_MTU6] = cnf->mtu6;
4080 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4081 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4082 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
4083 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4084 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4085 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4086 		jiffies_to_msecs(cnf->rtr_solicit_interval);
4087 	array[DEVCONF_RTR_SOLICIT_DELAY] =
4088 		jiffies_to_msecs(cnf->rtr_solicit_delay);
4089 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4090 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4091 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4092 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4093 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4094 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4095 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4096 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4097 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4098 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4099 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4100 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4101 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4102 #ifdef CONFIG_IPV6_ROUTER_PREF
4103 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4104 	array[DEVCONF_RTR_PROBE_INTERVAL] =
4105 		jiffies_to_msecs(cnf->rtr_probe_interval);
4106 #ifdef CONFIG_IPV6_ROUTE_INFO
4107 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4108 #endif
4109 #endif
4110 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4111 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4112 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4113 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4114 #endif
4115 #ifdef CONFIG_IPV6_MROUTE
4116 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4117 #endif
4118 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4119 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4120 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4121 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4122 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4123 }
4124 
4125 static inline size_t inet6_ifla6_size(void)
4126 {
4127 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
4128 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
4129 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4130 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4131 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4132 	     + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4133 }
4134 
4135 static inline size_t inet6_if_nlmsg_size(void)
4136 {
4137 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4138 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4139 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4140 	       + nla_total_size(4) /* IFLA_MTU */
4141 	       + nla_total_size(4) /* IFLA_LINK */
4142 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4143 }
4144 
4145 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4146 				      int items, int bytes)
4147 {
4148 	int i;
4149 	int pad = bytes - sizeof(u64) * items;
4150 	BUG_ON(pad < 0);
4151 
4152 	/* Use put_unaligned() because stats may not be aligned for u64. */
4153 	put_unaligned(items, &stats[0]);
4154 	for (i = 1; i < items; i++)
4155 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4156 
4157 	memset(&stats[items], 0, pad);
4158 }
4159 
4160 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4161 				      int items, int bytes, size_t syncpoff)
4162 {
4163 	int i;
4164 	int pad = bytes - sizeof(u64) * items;
4165 	BUG_ON(pad < 0);
4166 
4167 	/* Use put_unaligned() because stats may not be aligned for u64. */
4168 	put_unaligned(items, &stats[0]);
4169 	for (i = 1; i < items; i++)
4170 		put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4171 
4172 	memset(&stats[items], 0, pad);
4173 }
4174 
4175 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4176 			     int bytes)
4177 {
4178 	switch (attrtype) {
4179 	case IFLA_INET6_STATS:
4180 		__snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4181 				     IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4182 		break;
4183 	case IFLA_INET6_ICMP6STATS:
4184 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4185 		break;
4186 	}
4187 }
4188 
4189 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4190 {
4191 	struct nlattr *nla;
4192 	struct ifla_cacheinfo ci;
4193 
4194 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4195 		goto nla_put_failure;
4196 	ci.max_reasm_len = IPV6_MAXPLEN;
4197 	ci.tstamp = cstamp_delta(idev->tstamp);
4198 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4199 	ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4200 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4201 		goto nla_put_failure;
4202 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4203 	if (nla == NULL)
4204 		goto nla_put_failure;
4205 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4206 
4207 	/* XXX - MC not implemented */
4208 
4209 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4210 	if (nla == NULL)
4211 		goto nla_put_failure;
4212 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4213 
4214 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4215 	if (nla == NULL)
4216 		goto nla_put_failure;
4217 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4218 
4219 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4220 	if (nla == NULL)
4221 		goto nla_put_failure;
4222 	read_lock_bh(&idev->lock);
4223 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4224 	read_unlock_bh(&idev->lock);
4225 
4226 	return 0;
4227 
4228 nla_put_failure:
4229 	return -EMSGSIZE;
4230 }
4231 
4232 static size_t inet6_get_link_af_size(const struct net_device *dev)
4233 {
4234 	if (!__in6_dev_get(dev))
4235 		return 0;
4236 
4237 	return inet6_ifla6_size();
4238 }
4239 
4240 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4241 {
4242 	struct inet6_dev *idev = __in6_dev_get(dev);
4243 
4244 	if (!idev)
4245 		return -ENODATA;
4246 
4247 	if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4248 		return -EMSGSIZE;
4249 
4250 	return 0;
4251 }
4252 
4253 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4254 {
4255 	struct inet6_ifaddr *ifp;
4256 	struct net_device *dev = idev->dev;
4257 	bool update_rs = false;
4258 	struct in6_addr ll_addr;
4259 
4260 	if (token == NULL)
4261 		return -EINVAL;
4262 	if (ipv6_addr_any(token))
4263 		return -EINVAL;
4264 	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4265 		return -EINVAL;
4266 	if (!ipv6_accept_ra(idev))
4267 		return -EINVAL;
4268 	if (idev->cnf.rtr_solicits <= 0)
4269 		return -EINVAL;
4270 
4271 	write_lock_bh(&idev->lock);
4272 
4273 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4274 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4275 
4276 	write_unlock_bh(&idev->lock);
4277 
4278 	if (!idev->dead && (idev->if_flags & IF_READY) &&
4279 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4280 			     IFA_F_OPTIMISTIC)) {
4281 
4282 		/* If we're not ready, then normal ifup will take care
4283 		 * of this. Otherwise, we need to request our rs here.
4284 		 */
4285 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4286 		update_rs = true;
4287 	}
4288 
4289 	write_lock_bh(&idev->lock);
4290 
4291 	if (update_rs) {
4292 		idev->if_flags |= IF_RS_SENT;
4293 		idev->rs_probes = 1;
4294 		addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4295 	}
4296 
4297 	/* Well, that's kinda nasty ... */
4298 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4299 		spin_lock(&ifp->lock);
4300 		if (ifp->tokenized) {
4301 			ifp->valid_lft = 0;
4302 			ifp->prefered_lft = 0;
4303 		}
4304 		spin_unlock(&ifp->lock);
4305 	}
4306 
4307 	write_unlock_bh(&idev->lock);
4308 	addrconf_verify(0);
4309 	return 0;
4310 }
4311 
4312 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4313 {
4314 	int err = -EINVAL;
4315 	struct inet6_dev *idev = __in6_dev_get(dev);
4316 	struct nlattr *tb[IFLA_INET6_MAX + 1];
4317 
4318 	if (!idev)
4319 		return -EAFNOSUPPORT;
4320 
4321 	if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4322 		BUG();
4323 
4324 	if (tb[IFLA_INET6_TOKEN])
4325 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4326 
4327 	return err;
4328 }
4329 
4330 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4331 			     u32 portid, u32 seq, int event, unsigned int flags)
4332 {
4333 	struct net_device *dev = idev->dev;
4334 	struct ifinfomsg *hdr;
4335 	struct nlmsghdr *nlh;
4336 	void *protoinfo;
4337 
4338 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4339 	if (nlh == NULL)
4340 		return -EMSGSIZE;
4341 
4342 	hdr = nlmsg_data(nlh);
4343 	hdr->ifi_family = AF_INET6;
4344 	hdr->__ifi_pad = 0;
4345 	hdr->ifi_type = dev->type;
4346 	hdr->ifi_index = dev->ifindex;
4347 	hdr->ifi_flags = dev_get_flags(dev);
4348 	hdr->ifi_change = 0;
4349 
4350 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4351 	    (dev->addr_len &&
4352 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4353 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4354 	    (dev->ifindex != dev->iflink &&
4355 	     nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4356 		goto nla_put_failure;
4357 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4358 	if (protoinfo == NULL)
4359 		goto nla_put_failure;
4360 
4361 	if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4362 		goto nla_put_failure;
4363 
4364 	nla_nest_end(skb, protoinfo);
4365 	return nlmsg_end(skb, nlh);
4366 
4367 nla_put_failure:
4368 	nlmsg_cancel(skb, nlh);
4369 	return -EMSGSIZE;
4370 }
4371 
4372 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4373 {
4374 	struct net *net = sock_net(skb->sk);
4375 	int h, s_h;
4376 	int idx = 0, s_idx;
4377 	struct net_device *dev;
4378 	struct inet6_dev *idev;
4379 	struct hlist_head *head;
4380 
4381 	s_h = cb->args[0];
4382 	s_idx = cb->args[1];
4383 
4384 	rcu_read_lock();
4385 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4386 		idx = 0;
4387 		head = &net->dev_index_head[h];
4388 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
4389 			if (idx < s_idx)
4390 				goto cont;
4391 			idev = __in6_dev_get(dev);
4392 			if (!idev)
4393 				goto cont;
4394 			if (inet6_fill_ifinfo(skb, idev,
4395 					      NETLINK_CB(cb->skb).portid,
4396 					      cb->nlh->nlmsg_seq,
4397 					      RTM_NEWLINK, NLM_F_MULTI) <= 0)
4398 				goto out;
4399 cont:
4400 			idx++;
4401 		}
4402 	}
4403 out:
4404 	rcu_read_unlock();
4405 	cb->args[1] = idx;
4406 	cb->args[0] = h;
4407 
4408 	return skb->len;
4409 }
4410 
4411 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4412 {
4413 	struct sk_buff *skb;
4414 	struct net *net = dev_net(idev->dev);
4415 	int err = -ENOBUFS;
4416 
4417 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4418 	if (skb == NULL)
4419 		goto errout;
4420 
4421 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4422 	if (err < 0) {
4423 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4424 		WARN_ON(err == -EMSGSIZE);
4425 		kfree_skb(skb);
4426 		goto errout;
4427 	}
4428 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4429 	return;
4430 errout:
4431 	if (err < 0)
4432 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4433 }
4434 
4435 static inline size_t inet6_prefix_nlmsg_size(void)
4436 {
4437 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
4438 	       + nla_total_size(sizeof(struct in6_addr))
4439 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
4440 }
4441 
4442 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4443 			     struct prefix_info *pinfo, u32 portid, u32 seq,
4444 			     int event, unsigned int flags)
4445 {
4446 	struct prefixmsg *pmsg;
4447 	struct nlmsghdr *nlh;
4448 	struct prefix_cacheinfo	ci;
4449 
4450 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4451 	if (nlh == NULL)
4452 		return -EMSGSIZE;
4453 
4454 	pmsg = nlmsg_data(nlh);
4455 	pmsg->prefix_family = AF_INET6;
4456 	pmsg->prefix_pad1 = 0;
4457 	pmsg->prefix_pad2 = 0;
4458 	pmsg->prefix_ifindex = idev->dev->ifindex;
4459 	pmsg->prefix_len = pinfo->prefix_len;
4460 	pmsg->prefix_type = pinfo->type;
4461 	pmsg->prefix_pad3 = 0;
4462 	pmsg->prefix_flags = 0;
4463 	if (pinfo->onlink)
4464 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4465 	if (pinfo->autoconf)
4466 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4467 
4468 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4469 		goto nla_put_failure;
4470 	ci.preferred_time = ntohl(pinfo->prefered);
4471 	ci.valid_time = ntohl(pinfo->valid);
4472 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4473 		goto nla_put_failure;
4474 	return nlmsg_end(skb, nlh);
4475 
4476 nla_put_failure:
4477 	nlmsg_cancel(skb, nlh);
4478 	return -EMSGSIZE;
4479 }
4480 
4481 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4482 			 struct prefix_info *pinfo)
4483 {
4484 	struct sk_buff *skb;
4485 	struct net *net = dev_net(idev->dev);
4486 	int err = -ENOBUFS;
4487 
4488 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4489 	if (skb == NULL)
4490 		goto errout;
4491 
4492 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4493 	if (err < 0) {
4494 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4495 		WARN_ON(err == -EMSGSIZE);
4496 		kfree_skb(skb);
4497 		goto errout;
4498 	}
4499 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4500 	return;
4501 errout:
4502 	if (err < 0)
4503 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4504 }
4505 
4506 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4507 {
4508 	write_lock_bh(&ifp->idev->lock);
4509 	spin_lock(&ifp->lock);
4510 	if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4511 			    IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4512 	    (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4513 		ifp->idev->valid_ll_addr_cnt += count;
4514 	WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4515 	spin_unlock(&ifp->lock);
4516 	write_unlock_bh(&ifp->idev->lock);
4517 }
4518 
4519 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4520 {
4521 	struct net *net = dev_net(ifp->idev->dev);
4522 
4523 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4524 
4525 	switch (event) {
4526 	case RTM_NEWADDR:
4527 		update_valid_ll_addr_cnt(ifp, 1);
4528 
4529 		/*
4530 		 * If the address was optimistic
4531 		 * we inserted the route at the start of
4532 		 * our DAD process, so we don't need
4533 		 * to do it again
4534 		 */
4535 		if (!(ifp->rt->rt6i_node))
4536 			ip6_ins_rt(ifp->rt);
4537 		if (ifp->idev->cnf.forwarding)
4538 			addrconf_join_anycast(ifp);
4539 		if (!ipv6_addr_any(&ifp->peer_addr))
4540 			addrconf_prefix_route(&ifp->peer_addr, 128,
4541 					      ifp->idev->dev, 0, 0);
4542 		break;
4543 	case RTM_DELADDR:
4544 		update_valid_ll_addr_cnt(ifp, -1);
4545 
4546 		if (ifp->idev->cnf.forwarding)
4547 			addrconf_leave_anycast(ifp);
4548 		addrconf_leave_solict(ifp->idev, &ifp->addr);
4549 		if (!ipv6_addr_any(&ifp->peer_addr)) {
4550 			struct rt6_info *rt;
4551 			struct net_device *dev = ifp->idev->dev;
4552 
4553 			rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4554 					dev->ifindex, 1);
4555 			if (rt) {
4556 				dst_hold(&rt->dst);
4557 				if (ip6_del_rt(rt))
4558 					dst_free(&rt->dst);
4559 			}
4560 		}
4561 		dst_hold(&ifp->rt->dst);
4562 
4563 		if (ip6_del_rt(ifp->rt))
4564 			dst_free(&ifp->rt->dst);
4565 		break;
4566 	}
4567 	atomic_inc(&net->ipv6.dev_addr_genid);
4568 	rt_genid_bump_ipv6(net);
4569 }
4570 
4571 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4572 {
4573 	rcu_read_lock_bh();
4574 	if (likely(ifp->idev->dead == 0))
4575 		__ipv6_ifa_notify(event, ifp);
4576 	rcu_read_unlock_bh();
4577 }
4578 
4579 #ifdef CONFIG_SYSCTL
4580 
4581 static
4582 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4583 			   void __user *buffer, size_t *lenp, loff_t *ppos)
4584 {
4585 	int *valp = ctl->data;
4586 	int val = *valp;
4587 	loff_t pos = *ppos;
4588 	struct ctl_table lctl;
4589 	int ret;
4590 
4591 	/*
4592 	 * ctl->data points to idev->cnf.forwarding, we should
4593 	 * not modify it until we get the rtnl lock.
4594 	 */
4595 	lctl = *ctl;
4596 	lctl.data = &val;
4597 
4598 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4599 
4600 	if (write)
4601 		ret = addrconf_fixup_forwarding(ctl, valp, val);
4602 	if (ret)
4603 		*ppos = pos;
4604 	return ret;
4605 }
4606 
4607 static void dev_disable_change(struct inet6_dev *idev)
4608 {
4609 	struct netdev_notifier_info info;
4610 
4611 	if (!idev || !idev->dev)
4612 		return;
4613 
4614 	netdev_notifier_info_init(&info, idev->dev);
4615 	if (idev->cnf.disable_ipv6)
4616 		addrconf_notify(NULL, NETDEV_DOWN, &info);
4617 	else
4618 		addrconf_notify(NULL, NETDEV_UP, &info);
4619 }
4620 
4621 static void addrconf_disable_change(struct net *net, __s32 newf)
4622 {
4623 	struct net_device *dev;
4624 	struct inet6_dev *idev;
4625 
4626 	rcu_read_lock();
4627 	for_each_netdev_rcu(net, dev) {
4628 		idev = __in6_dev_get(dev);
4629 		if (idev) {
4630 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4631 			idev->cnf.disable_ipv6 = newf;
4632 			if (changed)
4633 				dev_disable_change(idev);
4634 		}
4635 	}
4636 	rcu_read_unlock();
4637 }
4638 
4639 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4640 {
4641 	struct net *net;
4642 	int old;
4643 
4644 	if (!rtnl_trylock())
4645 		return restart_syscall();
4646 
4647 	net = (struct net *)table->extra2;
4648 	old = *p;
4649 	*p = newf;
4650 
4651 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4652 		rtnl_unlock();
4653 		return 0;
4654 	}
4655 
4656 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
4657 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
4658 		addrconf_disable_change(net, newf);
4659 	} else if ((!newf) ^ (!old))
4660 		dev_disable_change((struct inet6_dev *)table->extra1);
4661 
4662 	rtnl_unlock();
4663 	return 0;
4664 }
4665 
4666 static
4667 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4668 			    void __user *buffer, size_t *lenp, loff_t *ppos)
4669 {
4670 	int *valp = ctl->data;
4671 	int val = *valp;
4672 	loff_t pos = *ppos;
4673 	struct ctl_table lctl;
4674 	int ret;
4675 
4676 	/*
4677 	 * ctl->data points to idev->cnf.disable_ipv6, we should
4678 	 * not modify it until we get the rtnl lock.
4679 	 */
4680 	lctl = *ctl;
4681 	lctl.data = &val;
4682 
4683 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4684 
4685 	if (write)
4686 		ret = addrconf_disable_ipv6(ctl, valp, val);
4687 	if (ret)
4688 		*ppos = pos;
4689 	return ret;
4690 }
4691 
4692 static struct addrconf_sysctl_table
4693 {
4694 	struct ctl_table_header *sysctl_header;
4695 	struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4696 } addrconf_sysctl __read_mostly = {
4697 	.sysctl_header = NULL,
4698 	.addrconf_vars = {
4699 		{
4700 			.procname	= "forwarding",
4701 			.data		= &ipv6_devconf.forwarding,
4702 			.maxlen		= sizeof(int),
4703 			.mode		= 0644,
4704 			.proc_handler	= addrconf_sysctl_forward,
4705 		},
4706 		{
4707 			.procname	= "hop_limit",
4708 			.data		= &ipv6_devconf.hop_limit,
4709 			.maxlen		= sizeof(int),
4710 			.mode		= 0644,
4711 			.proc_handler	= proc_dointvec,
4712 		},
4713 		{
4714 			.procname	= "mtu",
4715 			.data		= &ipv6_devconf.mtu6,
4716 			.maxlen		= sizeof(int),
4717 			.mode		= 0644,
4718 			.proc_handler	= proc_dointvec,
4719 		},
4720 		{
4721 			.procname	= "accept_ra",
4722 			.data		= &ipv6_devconf.accept_ra,
4723 			.maxlen		= sizeof(int),
4724 			.mode		= 0644,
4725 			.proc_handler	= proc_dointvec,
4726 		},
4727 		{
4728 			.procname	= "accept_redirects",
4729 			.data		= &ipv6_devconf.accept_redirects,
4730 			.maxlen		= sizeof(int),
4731 			.mode		= 0644,
4732 			.proc_handler	= proc_dointvec,
4733 		},
4734 		{
4735 			.procname	= "autoconf",
4736 			.data		= &ipv6_devconf.autoconf,
4737 			.maxlen		= sizeof(int),
4738 			.mode		= 0644,
4739 			.proc_handler	= proc_dointvec,
4740 		},
4741 		{
4742 			.procname	= "dad_transmits",
4743 			.data		= &ipv6_devconf.dad_transmits,
4744 			.maxlen		= sizeof(int),
4745 			.mode		= 0644,
4746 			.proc_handler	= proc_dointvec,
4747 		},
4748 		{
4749 			.procname	= "router_solicitations",
4750 			.data		= &ipv6_devconf.rtr_solicits,
4751 			.maxlen		= sizeof(int),
4752 			.mode		= 0644,
4753 			.proc_handler	= proc_dointvec,
4754 		},
4755 		{
4756 			.procname	= "router_solicitation_interval",
4757 			.data		= &ipv6_devconf.rtr_solicit_interval,
4758 			.maxlen		= sizeof(int),
4759 			.mode		= 0644,
4760 			.proc_handler	= proc_dointvec_jiffies,
4761 		},
4762 		{
4763 			.procname	= "router_solicitation_delay",
4764 			.data		= &ipv6_devconf.rtr_solicit_delay,
4765 			.maxlen		= sizeof(int),
4766 			.mode		= 0644,
4767 			.proc_handler	= proc_dointvec_jiffies,
4768 		},
4769 		{
4770 			.procname	= "force_mld_version",
4771 			.data		= &ipv6_devconf.force_mld_version,
4772 			.maxlen		= sizeof(int),
4773 			.mode		= 0644,
4774 			.proc_handler	= proc_dointvec,
4775 		},
4776 		{
4777 			.procname	= "mldv1_unsolicited_report_interval",
4778 			.data		=
4779 				&ipv6_devconf.mldv1_unsolicited_report_interval,
4780 			.maxlen		= sizeof(int),
4781 			.mode		= 0644,
4782 			.proc_handler	= proc_dointvec_ms_jiffies,
4783 		},
4784 		{
4785 			.procname	= "mldv2_unsolicited_report_interval",
4786 			.data		=
4787 				&ipv6_devconf.mldv2_unsolicited_report_interval,
4788 			.maxlen		= sizeof(int),
4789 			.mode		= 0644,
4790 			.proc_handler	= proc_dointvec_ms_jiffies,
4791 		},
4792 		{
4793 			.procname	= "use_tempaddr",
4794 			.data		= &ipv6_devconf.use_tempaddr,
4795 			.maxlen		= sizeof(int),
4796 			.mode		= 0644,
4797 			.proc_handler	= proc_dointvec,
4798 		},
4799 		{
4800 			.procname	= "temp_valid_lft",
4801 			.data		= &ipv6_devconf.temp_valid_lft,
4802 			.maxlen		= sizeof(int),
4803 			.mode		= 0644,
4804 			.proc_handler	= proc_dointvec,
4805 		},
4806 		{
4807 			.procname	= "temp_prefered_lft",
4808 			.data		= &ipv6_devconf.temp_prefered_lft,
4809 			.maxlen		= sizeof(int),
4810 			.mode		= 0644,
4811 			.proc_handler	= proc_dointvec,
4812 		},
4813 		{
4814 			.procname	= "regen_max_retry",
4815 			.data		= &ipv6_devconf.regen_max_retry,
4816 			.maxlen		= sizeof(int),
4817 			.mode		= 0644,
4818 			.proc_handler	= proc_dointvec,
4819 		},
4820 		{
4821 			.procname	= "max_desync_factor",
4822 			.data		= &ipv6_devconf.max_desync_factor,
4823 			.maxlen		= sizeof(int),
4824 			.mode		= 0644,
4825 			.proc_handler	= proc_dointvec,
4826 		},
4827 		{
4828 			.procname	= "max_addresses",
4829 			.data		= &ipv6_devconf.max_addresses,
4830 			.maxlen		= sizeof(int),
4831 			.mode		= 0644,
4832 			.proc_handler	= proc_dointvec,
4833 		},
4834 		{
4835 			.procname	= "accept_ra_defrtr",
4836 			.data		= &ipv6_devconf.accept_ra_defrtr,
4837 			.maxlen		= sizeof(int),
4838 			.mode		= 0644,
4839 			.proc_handler	= proc_dointvec,
4840 		},
4841 		{
4842 			.procname	= "accept_ra_pinfo",
4843 			.data		= &ipv6_devconf.accept_ra_pinfo,
4844 			.maxlen		= sizeof(int),
4845 			.mode		= 0644,
4846 			.proc_handler	= proc_dointvec,
4847 		},
4848 #ifdef CONFIG_IPV6_ROUTER_PREF
4849 		{
4850 			.procname	= "accept_ra_rtr_pref",
4851 			.data		= &ipv6_devconf.accept_ra_rtr_pref,
4852 			.maxlen		= sizeof(int),
4853 			.mode		= 0644,
4854 			.proc_handler	= proc_dointvec,
4855 		},
4856 		{
4857 			.procname	= "router_probe_interval",
4858 			.data		= &ipv6_devconf.rtr_probe_interval,
4859 			.maxlen		= sizeof(int),
4860 			.mode		= 0644,
4861 			.proc_handler	= proc_dointvec_jiffies,
4862 		},
4863 #ifdef CONFIG_IPV6_ROUTE_INFO
4864 		{
4865 			.procname	= "accept_ra_rt_info_max_plen",
4866 			.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
4867 			.maxlen		= sizeof(int),
4868 			.mode		= 0644,
4869 			.proc_handler	= proc_dointvec,
4870 		},
4871 #endif
4872 #endif
4873 		{
4874 			.procname	= "proxy_ndp",
4875 			.data		= &ipv6_devconf.proxy_ndp,
4876 			.maxlen		= sizeof(int),
4877 			.mode		= 0644,
4878 			.proc_handler	= proc_dointvec,
4879 		},
4880 		{
4881 			.procname	= "accept_source_route",
4882 			.data		= &ipv6_devconf.accept_source_route,
4883 			.maxlen		= sizeof(int),
4884 			.mode		= 0644,
4885 			.proc_handler	= proc_dointvec,
4886 		},
4887 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4888 		{
4889 			.procname       = "optimistic_dad",
4890 			.data           = &ipv6_devconf.optimistic_dad,
4891 			.maxlen         = sizeof(int),
4892 			.mode           = 0644,
4893 			.proc_handler   = proc_dointvec,
4894 
4895 		},
4896 #endif
4897 #ifdef CONFIG_IPV6_MROUTE
4898 		{
4899 			.procname	= "mc_forwarding",
4900 			.data		= &ipv6_devconf.mc_forwarding,
4901 			.maxlen		= sizeof(int),
4902 			.mode		= 0444,
4903 			.proc_handler	= proc_dointvec,
4904 		},
4905 #endif
4906 		{
4907 			.procname	= "disable_ipv6",
4908 			.data		= &ipv6_devconf.disable_ipv6,
4909 			.maxlen		= sizeof(int),
4910 			.mode		= 0644,
4911 			.proc_handler	= addrconf_sysctl_disable,
4912 		},
4913 		{
4914 			.procname	= "accept_dad",
4915 			.data		= &ipv6_devconf.accept_dad,
4916 			.maxlen		= sizeof(int),
4917 			.mode		= 0644,
4918 			.proc_handler	= proc_dointvec,
4919 		},
4920 		{
4921 			.procname       = "force_tllao",
4922 			.data           = &ipv6_devconf.force_tllao,
4923 			.maxlen         = sizeof(int),
4924 			.mode           = 0644,
4925 			.proc_handler   = proc_dointvec
4926 		},
4927 		{
4928 			.procname       = "ndisc_notify",
4929 			.data           = &ipv6_devconf.ndisc_notify,
4930 			.maxlen         = sizeof(int),
4931 			.mode           = 0644,
4932 			.proc_handler   = proc_dointvec
4933 		},
4934 		{
4935 			.procname	= "suppress_frag_ndisc",
4936 			.data		= &ipv6_devconf.suppress_frag_ndisc,
4937 			.maxlen		= sizeof(int),
4938 			.mode		= 0644,
4939 			.proc_handler	= proc_dointvec
4940 		},
4941 		{
4942 			/* sentinel */
4943 		}
4944 	},
4945 };
4946 
4947 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4948 		struct inet6_dev *idev, struct ipv6_devconf *p)
4949 {
4950 	int i;
4951 	struct addrconf_sysctl_table *t;
4952 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4953 
4954 	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4955 	if (t == NULL)
4956 		goto out;
4957 
4958 	for (i = 0; t->addrconf_vars[i].data; i++) {
4959 		t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4960 		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4961 		t->addrconf_vars[i].extra2 = net;
4962 	}
4963 
4964 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4965 
4966 	t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4967 	if (t->sysctl_header == NULL)
4968 		goto free;
4969 
4970 	p->sysctl = t;
4971 	return 0;
4972 
4973 free:
4974 	kfree(t);
4975 out:
4976 	return -ENOBUFS;
4977 }
4978 
4979 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4980 {
4981 	struct addrconf_sysctl_table *t;
4982 
4983 	if (p->sysctl == NULL)
4984 		return;
4985 
4986 	t = p->sysctl;
4987 	p->sysctl = NULL;
4988 	unregister_net_sysctl_table(t->sysctl_header);
4989 	kfree(t);
4990 }
4991 
4992 static void addrconf_sysctl_register(struct inet6_dev *idev)
4993 {
4994 	neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4995 			      &ndisc_ifinfo_sysctl_change);
4996 	__addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4997 					idev, &idev->cnf);
4998 }
4999 
5000 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5001 {
5002 	__addrconf_sysctl_unregister(&idev->cnf);
5003 	neigh_sysctl_unregister(idev->nd_parms);
5004 }
5005 
5006 
5007 #endif
5008 
5009 static int __net_init addrconf_init_net(struct net *net)
5010 {
5011 	int err = -ENOMEM;
5012 	struct ipv6_devconf *all, *dflt;
5013 
5014 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5015 	if (all == NULL)
5016 		goto err_alloc_all;
5017 
5018 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5019 	if (dflt == NULL)
5020 		goto err_alloc_dflt;
5021 
5022 	/* these will be inherited by all namespaces */
5023 	dflt->autoconf = ipv6_defaults.autoconf;
5024 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5025 
5026 	net->ipv6.devconf_all = all;
5027 	net->ipv6.devconf_dflt = dflt;
5028 
5029 #ifdef CONFIG_SYSCTL
5030 	err = __addrconf_sysctl_register(net, "all", NULL, all);
5031 	if (err < 0)
5032 		goto err_reg_all;
5033 
5034 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5035 	if (err < 0)
5036 		goto err_reg_dflt;
5037 #endif
5038 	return 0;
5039 
5040 #ifdef CONFIG_SYSCTL
5041 err_reg_dflt:
5042 	__addrconf_sysctl_unregister(all);
5043 err_reg_all:
5044 	kfree(dflt);
5045 #endif
5046 err_alloc_dflt:
5047 	kfree(all);
5048 err_alloc_all:
5049 	return err;
5050 }
5051 
5052 static void __net_exit addrconf_exit_net(struct net *net)
5053 {
5054 #ifdef CONFIG_SYSCTL
5055 	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5056 	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
5057 #endif
5058 	if (!net_eq(net, &init_net)) {
5059 		kfree(net->ipv6.devconf_dflt);
5060 		kfree(net->ipv6.devconf_all);
5061 	}
5062 }
5063 
5064 static struct pernet_operations addrconf_ops = {
5065 	.init = addrconf_init_net,
5066 	.exit = addrconf_exit_net,
5067 };
5068 
5069 static struct rtnl_af_ops inet6_ops = {
5070 	.family		  = AF_INET6,
5071 	.fill_link_af	  = inet6_fill_link_af,
5072 	.get_link_af_size = inet6_get_link_af_size,
5073 	.set_link_af	  = inet6_set_link_af,
5074 };
5075 
5076 /*
5077  *	Init / cleanup code
5078  */
5079 
5080 int __init addrconf_init(void)
5081 {
5082 	int i, err;
5083 
5084 	err = ipv6_addr_label_init();
5085 	if (err < 0) {
5086 		pr_crit("%s: cannot initialize default policy table: %d\n",
5087 			__func__, err);
5088 		goto out;
5089 	}
5090 
5091 	err = register_pernet_subsys(&addrconf_ops);
5092 	if (err < 0)
5093 		goto out_addrlabel;
5094 
5095 	/* The addrconf netdev notifier requires that loopback_dev
5096 	 * has it's ipv6 private information allocated and setup
5097 	 * before it can bring up and give link-local addresses
5098 	 * to other devices which are up.
5099 	 *
5100 	 * Unfortunately, loopback_dev is not necessarily the first
5101 	 * entry in the global dev_base list of net devices.  In fact,
5102 	 * it is likely to be the very last entry on that list.
5103 	 * So this causes the notifier registry below to try and
5104 	 * give link-local addresses to all devices besides loopback_dev
5105 	 * first, then loopback_dev, which cases all the non-loopback_dev
5106 	 * devices to fail to get a link-local address.
5107 	 *
5108 	 * So, as a temporary fix, allocate the ipv6 structure for
5109 	 * loopback_dev first by hand.
5110 	 * Longer term, all of the dependencies ipv6 has upon the loopback
5111 	 * device and it being up should be removed.
5112 	 */
5113 	rtnl_lock();
5114 	if (!ipv6_add_dev(init_net.loopback_dev))
5115 		err = -ENOMEM;
5116 	rtnl_unlock();
5117 	if (err)
5118 		goto errlo;
5119 
5120 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
5121 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5122 
5123 	register_netdevice_notifier(&ipv6_dev_notf);
5124 
5125 	addrconf_verify(0);
5126 
5127 	err = rtnl_af_register(&inet6_ops);
5128 	if (err < 0)
5129 		goto errout_af;
5130 
5131 	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5132 			      NULL);
5133 	if (err < 0)
5134 		goto errout;
5135 
5136 	/* Only the first call to __rtnl_register can fail */
5137 	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5138 	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5139 	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5140 			inet6_dump_ifaddr, NULL);
5141 	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5142 			inet6_dump_ifmcaddr, NULL);
5143 	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5144 			inet6_dump_ifacaddr, NULL);
5145 	__rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5146 			inet6_netconf_dump_devconf, NULL);
5147 
5148 	ipv6_addr_label_rtnl_register();
5149 
5150 	return 0;
5151 errout:
5152 	rtnl_af_unregister(&inet6_ops);
5153 errout_af:
5154 	unregister_netdevice_notifier(&ipv6_dev_notf);
5155 errlo:
5156 	unregister_pernet_subsys(&addrconf_ops);
5157 out_addrlabel:
5158 	ipv6_addr_label_cleanup();
5159 out:
5160 	return err;
5161 }
5162 
5163 void addrconf_cleanup(void)
5164 {
5165 	struct net_device *dev;
5166 	int i;
5167 
5168 	unregister_netdevice_notifier(&ipv6_dev_notf);
5169 	unregister_pernet_subsys(&addrconf_ops);
5170 	ipv6_addr_label_cleanup();
5171 
5172 	rtnl_lock();
5173 
5174 	__rtnl_af_unregister(&inet6_ops);
5175 
5176 	/* clean dev list */
5177 	for_each_netdev(&init_net, dev) {
5178 		if (__in6_dev_get(dev) == NULL)
5179 			continue;
5180 		addrconf_ifdown(dev, 1);
5181 	}
5182 	addrconf_ifdown(init_net.loopback_dev, 2);
5183 
5184 	/*
5185 	 *	Check hash table.
5186 	 */
5187 	spin_lock_bh(&addrconf_hash_lock);
5188 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
5189 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5190 	spin_unlock_bh(&addrconf_hash_lock);
5191 
5192 	del_timer(&addr_chk_timer);
5193 	rtnl_unlock();
5194 }
5195