xref: /linux/net/ipv6/addrconf.c (revision 2cd1f0ddbb5667f61e69089964209e8f716e9009)
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/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
61 #ifdef CONFIG_SYSCTL
62 #include <linux/sysctl.h>
63 #endif
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
69 
70 #include <net/net_namespace.h>
71 #include <net/sock.h>
72 #include <net/snmp.h>
73 
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
76 #include <net/ipv6.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
81 #include <net/tcp.h>
82 #include <net/ip.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92 
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96 
97 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
98 
99 #define IPV6_MAX_STRLEN \
100 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
101 
102 static inline u32 cstamp_delta(unsigned long cstamp)
103 {
104 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
105 }
106 
107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
108 {
109 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
110 	u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
111 	do_div(tmp, 1000000);
112 	return (s32)tmp;
113 }
114 
115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
116 {
117 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
118 	u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
119 	do_div(tmp, 1000000);
120 	if ((s32)tmp > mrt) {
121 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
122 		tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
123 		do_div(tmp, 1000000);
124 	}
125 	return (s32)tmp;
126 }
127 
128 #ifdef CONFIG_SYSCTL
129 static int addrconf_sysctl_register(struct inet6_dev *idev);
130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
131 #else
132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
133 {
134 	return 0;
135 }
136 
137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
138 {
139 }
140 #endif
141 
142 static void ipv6_regen_rndid(struct inet6_dev *idev);
143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
144 
145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
146 static int ipv6_count_addresses(const struct inet6_dev *idev);
147 static int ipv6_generate_stable_address(struct in6_addr *addr,
148 					u8 dad_count,
149 					const struct inet6_dev *idev);
150 
151 #define IN6_ADDR_HSIZE_SHIFT	8
152 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
153 /*
154  *	Configured unicast address hash table
155  */
156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
157 static DEFINE_SPINLOCK(addrconf_hash_lock);
158 
159 static void addrconf_verify(void);
160 static void addrconf_verify_rtnl(void);
161 static void addrconf_verify_work(struct work_struct *);
162 
163 static struct workqueue_struct *addrconf_wq;
164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
165 
166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
168 
169 static void addrconf_type_change(struct net_device *dev,
170 				 unsigned long event);
171 static int addrconf_ifdown(struct net_device *dev, int how);
172 
173 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
174 						  int plen,
175 						  const struct net_device *dev,
176 						  u32 flags, u32 noflags);
177 
178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
179 static void addrconf_dad_work(struct work_struct *w);
180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
181 				   bool send_na);
182 static void addrconf_dad_run(struct inet6_dev *idev);
183 static void addrconf_rs_timer(struct timer_list *t);
184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
186 
187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
188 				struct prefix_info *pinfo);
189 
190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
191 	.forwarding		= 0,
192 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
193 	.mtu6			= IPV6_MIN_MTU,
194 	.accept_ra		= 1,
195 	.accept_redirects	= 1,
196 	.autoconf		= 1,
197 	.force_mld_version	= 0,
198 	.mldv1_unsolicited_report_interval = 10 * HZ,
199 	.mldv2_unsolicited_report_interval = HZ,
200 	.dad_transmits		= 1,
201 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
202 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
203 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
204 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
205 	.use_tempaddr		= 0,
206 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
207 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
208 	.regen_max_retry	= REGEN_MAX_RETRY,
209 	.max_desync_factor	= MAX_DESYNC_FACTOR,
210 	.max_addresses		= IPV6_MAX_ADDRESSES,
211 	.accept_ra_defrtr	= 1,
212 	.accept_ra_from_local	= 0,
213 	.accept_ra_min_hop_limit= 1,
214 	.accept_ra_pinfo	= 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216 	.accept_ra_rtr_pref	= 1,
217 	.rtr_probe_interval	= 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219 	.accept_ra_rt_info_min_plen = 0,
220 	.accept_ra_rt_info_max_plen = 0,
221 #endif
222 #endif
223 	.proxy_ndp		= 0,
224 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
225 	.disable_ipv6		= 0,
226 	.accept_dad		= 0,
227 	.suppress_frag_ndisc	= 1,
228 	.accept_ra_mtu		= 1,
229 	.stable_secret		= {
230 		.initialized = false,
231 	},
232 	.use_oif_addrs_only	= 0,
233 	.ignore_routes_with_linkdown = 0,
234 	.keep_addr_on_down	= 0,
235 	.seg6_enabled		= 0,
236 #ifdef CONFIG_IPV6_SEG6_HMAC
237 	.seg6_require_hmac	= 0,
238 #endif
239 	.enhanced_dad           = 1,
240 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
241 	.disable_policy		= 0,
242 };
243 
244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
245 	.forwarding		= 0,
246 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
247 	.mtu6			= IPV6_MIN_MTU,
248 	.accept_ra		= 1,
249 	.accept_redirects	= 1,
250 	.autoconf		= 1,
251 	.force_mld_version	= 0,
252 	.mldv1_unsolicited_report_interval = 10 * HZ,
253 	.mldv2_unsolicited_report_interval = HZ,
254 	.dad_transmits		= 1,
255 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
256 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
257 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
258 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
259 	.use_tempaddr		= 0,
260 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
261 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
262 	.regen_max_retry	= REGEN_MAX_RETRY,
263 	.max_desync_factor	= MAX_DESYNC_FACTOR,
264 	.max_addresses		= IPV6_MAX_ADDRESSES,
265 	.accept_ra_defrtr	= 1,
266 	.accept_ra_from_local	= 0,
267 	.accept_ra_min_hop_limit= 1,
268 	.accept_ra_pinfo	= 1,
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270 	.accept_ra_rtr_pref	= 1,
271 	.rtr_probe_interval	= 60 * HZ,
272 #ifdef CONFIG_IPV6_ROUTE_INFO
273 	.accept_ra_rt_info_min_plen = 0,
274 	.accept_ra_rt_info_max_plen = 0,
275 #endif
276 #endif
277 	.proxy_ndp		= 0,
278 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
279 	.disable_ipv6		= 0,
280 	.accept_dad		= 1,
281 	.suppress_frag_ndisc	= 1,
282 	.accept_ra_mtu		= 1,
283 	.stable_secret		= {
284 		.initialized = false,
285 	},
286 	.use_oif_addrs_only	= 0,
287 	.ignore_routes_with_linkdown = 0,
288 	.keep_addr_on_down	= 0,
289 	.seg6_enabled		= 0,
290 #ifdef CONFIG_IPV6_SEG6_HMAC
291 	.seg6_require_hmac	= 0,
292 #endif
293 	.enhanced_dad           = 1,
294 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
295 	.disable_policy		= 0,
296 };
297 
298 /* Check if link is ready: is it up and is a valid qdisc available */
299 static inline bool addrconf_link_ready(const struct net_device *dev)
300 {
301 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
302 }
303 
304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
305 {
306 	if (del_timer(&idev->rs_timer))
307 		__in6_dev_put(idev);
308 }
309 
310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
311 {
312 	if (cancel_delayed_work(&ifp->dad_work))
313 		__in6_ifa_put(ifp);
314 }
315 
316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
317 				  unsigned long when)
318 {
319 	if (!timer_pending(&idev->rs_timer))
320 		in6_dev_hold(idev);
321 	mod_timer(&idev->rs_timer, jiffies + when);
322 }
323 
324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
325 				   unsigned long delay)
326 {
327 	in6_ifa_hold(ifp);
328 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
329 		in6_ifa_put(ifp);
330 }
331 
332 static int snmp6_alloc_dev(struct inet6_dev *idev)
333 {
334 	int i;
335 
336 	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
337 	if (!idev->stats.ipv6)
338 		goto err_ip;
339 
340 	for_each_possible_cpu(i) {
341 		struct ipstats_mib *addrconf_stats;
342 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
343 		u64_stats_init(&addrconf_stats->syncp);
344 	}
345 
346 
347 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
348 					GFP_KERNEL);
349 	if (!idev->stats.icmpv6dev)
350 		goto err_icmp;
351 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
352 					   GFP_KERNEL);
353 	if (!idev->stats.icmpv6msgdev)
354 		goto err_icmpmsg;
355 
356 	return 0;
357 
358 err_icmpmsg:
359 	kfree(idev->stats.icmpv6dev);
360 err_icmp:
361 	free_percpu(idev->stats.ipv6);
362 err_ip:
363 	return -ENOMEM;
364 }
365 
366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
367 {
368 	struct inet6_dev *ndev;
369 	int err = -ENOMEM;
370 
371 	ASSERT_RTNL();
372 
373 	if (dev->mtu < IPV6_MIN_MTU)
374 		return ERR_PTR(-EINVAL);
375 
376 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
377 	if (!ndev)
378 		return ERR_PTR(err);
379 
380 	rwlock_init(&ndev->lock);
381 	ndev->dev = dev;
382 	INIT_LIST_HEAD(&ndev->addr_list);
383 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
384 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
385 
386 	if (ndev->cnf.stable_secret.initialized)
387 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
388 	else
389 		ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
390 
391 	ndev->cnf.mtu6 = dev->mtu;
392 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
393 	if (!ndev->nd_parms) {
394 		kfree(ndev);
395 		return ERR_PTR(err);
396 	}
397 	if (ndev->cnf.forwarding)
398 		dev_disable_lro(dev);
399 	/* We refer to the device */
400 	dev_hold(dev);
401 
402 	if (snmp6_alloc_dev(ndev) < 0) {
403 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
404 			   __func__);
405 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
406 		dev_put(dev);
407 		kfree(ndev);
408 		return ERR_PTR(err);
409 	}
410 
411 	if (snmp6_register_dev(ndev) < 0) {
412 		netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
413 			   __func__, dev->name);
414 		goto err_release;
415 	}
416 
417 	/* One reference from device. */
418 	refcount_set(&ndev->refcnt, 1);
419 
420 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
421 		ndev->cnf.accept_dad = -1;
422 
423 #if IS_ENABLED(CONFIG_IPV6_SIT)
424 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
425 		pr_info("%s: Disabled Multicast RS\n", dev->name);
426 		ndev->cnf.rtr_solicits = 0;
427 	}
428 #endif
429 
430 	INIT_LIST_HEAD(&ndev->tempaddr_list);
431 	ndev->desync_factor = U32_MAX;
432 	if ((dev->flags&IFF_LOOPBACK) ||
433 	    dev->type == ARPHRD_TUNNEL ||
434 	    dev->type == ARPHRD_TUNNEL6 ||
435 	    dev->type == ARPHRD_SIT ||
436 	    dev->type == ARPHRD_NONE) {
437 		ndev->cnf.use_tempaddr = -1;
438 	} else
439 		ipv6_regen_rndid(ndev);
440 
441 	ndev->token = in6addr_any;
442 
443 	if (netif_running(dev) && addrconf_link_ready(dev))
444 		ndev->if_flags |= IF_READY;
445 
446 	ipv6_mc_init_dev(ndev);
447 	ndev->tstamp = jiffies;
448 	err = addrconf_sysctl_register(ndev);
449 	if (err) {
450 		ipv6_mc_destroy_dev(ndev);
451 		snmp6_unregister_dev(ndev);
452 		goto err_release;
453 	}
454 	/* protected by rtnl_lock */
455 	rcu_assign_pointer(dev->ip6_ptr, ndev);
456 
457 	/* Join interface-local all-node multicast group */
458 	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
459 
460 	/* Join all-node multicast group */
461 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
462 
463 	/* Join all-router multicast group if forwarding is set */
464 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
465 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
466 
467 	return ndev;
468 
469 err_release:
470 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
471 	ndev->dead = 1;
472 	in6_dev_finish_destroy(ndev);
473 	return ERR_PTR(err);
474 }
475 
476 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
477 {
478 	struct inet6_dev *idev;
479 
480 	ASSERT_RTNL();
481 
482 	idev = __in6_dev_get(dev);
483 	if (!idev) {
484 		idev = ipv6_add_dev(dev);
485 		if (IS_ERR(idev))
486 			return NULL;
487 	}
488 
489 	if (dev->flags&IFF_UP)
490 		ipv6_mc_up(idev);
491 	return idev;
492 }
493 
494 static int inet6_netconf_msgsize_devconf(int type)
495 {
496 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
497 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
498 	bool all = false;
499 
500 	if (type == NETCONFA_ALL)
501 		all = true;
502 
503 	if (all || type == NETCONFA_FORWARDING)
504 		size += nla_total_size(4);
505 #ifdef CONFIG_IPV6_MROUTE
506 	if (all || type == NETCONFA_MC_FORWARDING)
507 		size += nla_total_size(4);
508 #endif
509 	if (all || type == NETCONFA_PROXY_NEIGH)
510 		size += nla_total_size(4);
511 
512 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
513 		size += nla_total_size(4);
514 
515 	return size;
516 }
517 
518 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
519 				      struct ipv6_devconf *devconf, u32 portid,
520 				      u32 seq, int event, unsigned int flags,
521 				      int type)
522 {
523 	struct nlmsghdr  *nlh;
524 	struct netconfmsg *ncm;
525 	bool all = false;
526 
527 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
528 			flags);
529 	if (!nlh)
530 		return -EMSGSIZE;
531 
532 	if (type == NETCONFA_ALL)
533 		all = true;
534 
535 	ncm = nlmsg_data(nlh);
536 	ncm->ncm_family = AF_INET6;
537 
538 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
539 		goto nla_put_failure;
540 
541 	if (!devconf)
542 		goto out;
543 
544 	if ((all || type == NETCONFA_FORWARDING) &&
545 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
546 		goto nla_put_failure;
547 #ifdef CONFIG_IPV6_MROUTE
548 	if ((all || type == NETCONFA_MC_FORWARDING) &&
549 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
550 			devconf->mc_forwarding) < 0)
551 		goto nla_put_failure;
552 #endif
553 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
554 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
555 		goto nla_put_failure;
556 
557 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
558 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
559 			devconf->ignore_routes_with_linkdown) < 0)
560 		goto nla_put_failure;
561 
562 out:
563 	nlmsg_end(skb, nlh);
564 	return 0;
565 
566 nla_put_failure:
567 	nlmsg_cancel(skb, nlh);
568 	return -EMSGSIZE;
569 }
570 
571 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
572 				  int ifindex, struct ipv6_devconf *devconf)
573 {
574 	struct sk_buff *skb;
575 	int err = -ENOBUFS;
576 
577 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
578 	if (!skb)
579 		goto errout;
580 
581 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
582 					 event, 0, type);
583 	if (err < 0) {
584 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
585 		WARN_ON(err == -EMSGSIZE);
586 		kfree_skb(skb);
587 		goto errout;
588 	}
589 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
590 	return;
591 errout:
592 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
593 }
594 
595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
596 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
597 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
598 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
599 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
600 };
601 
602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
603 				     struct nlmsghdr *nlh,
604 				     struct netlink_ext_ack *extack)
605 {
606 	struct net *net = sock_net(in_skb->sk);
607 	struct nlattr *tb[NETCONFA_MAX+1];
608 	struct inet6_dev *in6_dev = NULL;
609 	struct net_device *dev = NULL;
610 	struct netconfmsg *ncm;
611 	struct sk_buff *skb;
612 	struct ipv6_devconf *devconf;
613 	int ifindex;
614 	int err;
615 
616 	err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
617 			  devconf_ipv6_policy, extack);
618 	if (err < 0)
619 		return err;
620 
621 	if (!tb[NETCONFA_IFINDEX])
622 		return -EINVAL;
623 
624 	err = -EINVAL;
625 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
626 	switch (ifindex) {
627 	case NETCONFA_IFINDEX_ALL:
628 		devconf = net->ipv6.devconf_all;
629 		break;
630 	case NETCONFA_IFINDEX_DEFAULT:
631 		devconf = net->ipv6.devconf_dflt;
632 		break;
633 	default:
634 		dev = dev_get_by_index(net, ifindex);
635 		if (!dev)
636 			return -EINVAL;
637 		in6_dev = in6_dev_get(dev);
638 		if (!in6_dev)
639 			goto errout;
640 		devconf = &in6_dev->cnf;
641 		break;
642 	}
643 
644 	err = -ENOBUFS;
645 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
646 	if (!skb)
647 		goto errout;
648 
649 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
650 					 NETLINK_CB(in_skb).portid,
651 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
652 					 NETCONFA_ALL);
653 	if (err < 0) {
654 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
655 		WARN_ON(err == -EMSGSIZE);
656 		kfree_skb(skb);
657 		goto errout;
658 	}
659 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
660 errout:
661 	if (in6_dev)
662 		in6_dev_put(in6_dev);
663 	if (dev)
664 		dev_put(dev);
665 	return err;
666 }
667 
668 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
669 				      struct netlink_callback *cb)
670 {
671 	struct net *net = sock_net(skb->sk);
672 	int h, s_h;
673 	int idx, s_idx;
674 	struct net_device *dev;
675 	struct inet6_dev *idev;
676 	struct hlist_head *head;
677 
678 	s_h = cb->args[0];
679 	s_idx = idx = cb->args[1];
680 
681 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
682 		idx = 0;
683 		head = &net->dev_index_head[h];
684 		rcu_read_lock();
685 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
686 			  net->dev_base_seq;
687 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
688 			if (idx < s_idx)
689 				goto cont;
690 			idev = __in6_dev_get(dev);
691 			if (!idev)
692 				goto cont;
693 
694 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
695 						       &idev->cnf,
696 						       NETLINK_CB(cb->skb).portid,
697 						       cb->nlh->nlmsg_seq,
698 						       RTM_NEWNETCONF,
699 						       NLM_F_MULTI,
700 						       NETCONFA_ALL) < 0) {
701 				rcu_read_unlock();
702 				goto done;
703 			}
704 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
705 cont:
706 			idx++;
707 		}
708 		rcu_read_unlock();
709 	}
710 	if (h == NETDEV_HASHENTRIES) {
711 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
712 					       net->ipv6.devconf_all,
713 					       NETLINK_CB(cb->skb).portid,
714 					       cb->nlh->nlmsg_seq,
715 					       RTM_NEWNETCONF, NLM_F_MULTI,
716 					       NETCONFA_ALL) < 0)
717 			goto done;
718 		else
719 			h++;
720 	}
721 	if (h == NETDEV_HASHENTRIES + 1) {
722 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
723 					       net->ipv6.devconf_dflt,
724 					       NETLINK_CB(cb->skb).portid,
725 					       cb->nlh->nlmsg_seq,
726 					       RTM_NEWNETCONF, NLM_F_MULTI,
727 					       NETCONFA_ALL) < 0)
728 			goto done;
729 		else
730 			h++;
731 	}
732 done:
733 	cb->args[0] = h;
734 	cb->args[1] = idx;
735 
736 	return skb->len;
737 }
738 
739 #ifdef CONFIG_SYSCTL
740 static void dev_forward_change(struct inet6_dev *idev)
741 {
742 	struct net_device *dev;
743 	struct inet6_ifaddr *ifa;
744 
745 	if (!idev)
746 		return;
747 	dev = idev->dev;
748 	if (idev->cnf.forwarding)
749 		dev_disable_lro(dev);
750 	if (dev->flags & IFF_MULTICAST) {
751 		if (idev->cnf.forwarding) {
752 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
753 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
754 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
755 		} else {
756 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
757 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
758 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
759 		}
760 	}
761 
762 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
763 		if (ifa->flags&IFA_F_TENTATIVE)
764 			continue;
765 		if (idev->cnf.forwarding)
766 			addrconf_join_anycast(ifa);
767 		else
768 			addrconf_leave_anycast(ifa);
769 	}
770 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
771 				     NETCONFA_FORWARDING,
772 				     dev->ifindex, &idev->cnf);
773 }
774 
775 
776 static void addrconf_forward_change(struct net *net, __s32 newf)
777 {
778 	struct net_device *dev;
779 	struct inet6_dev *idev;
780 
781 	for_each_netdev(net, dev) {
782 		idev = __in6_dev_get(dev);
783 		if (idev) {
784 			int changed = (!idev->cnf.forwarding) ^ (!newf);
785 			idev->cnf.forwarding = newf;
786 			if (changed)
787 				dev_forward_change(idev);
788 		}
789 	}
790 }
791 
792 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
793 {
794 	struct net *net;
795 	int old;
796 
797 	if (!rtnl_trylock())
798 		return restart_syscall();
799 
800 	net = (struct net *)table->extra2;
801 	old = *p;
802 	*p = newf;
803 
804 	if (p == &net->ipv6.devconf_dflt->forwarding) {
805 		if ((!newf) ^ (!old))
806 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
807 						     NETCONFA_FORWARDING,
808 						     NETCONFA_IFINDEX_DEFAULT,
809 						     net->ipv6.devconf_dflt);
810 		rtnl_unlock();
811 		return 0;
812 	}
813 
814 	if (p == &net->ipv6.devconf_all->forwarding) {
815 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
816 
817 		net->ipv6.devconf_dflt->forwarding = newf;
818 		if ((!newf) ^ (!old_dflt))
819 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
820 						     NETCONFA_FORWARDING,
821 						     NETCONFA_IFINDEX_DEFAULT,
822 						     net->ipv6.devconf_dflt);
823 
824 		addrconf_forward_change(net, newf);
825 		if ((!newf) ^ (!old))
826 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
827 						     NETCONFA_FORWARDING,
828 						     NETCONFA_IFINDEX_ALL,
829 						     net->ipv6.devconf_all);
830 	} else if ((!newf) ^ (!old))
831 		dev_forward_change((struct inet6_dev *)table->extra1);
832 	rtnl_unlock();
833 
834 	if (newf)
835 		rt6_purge_dflt_routers(net);
836 	return 1;
837 }
838 
839 static void addrconf_linkdown_change(struct net *net, __s32 newf)
840 {
841 	struct net_device *dev;
842 	struct inet6_dev *idev;
843 
844 	for_each_netdev(net, dev) {
845 		idev = __in6_dev_get(dev);
846 		if (idev) {
847 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
848 
849 			idev->cnf.ignore_routes_with_linkdown = newf;
850 			if (changed)
851 				inet6_netconf_notify_devconf(dev_net(dev),
852 							     RTM_NEWNETCONF,
853 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
854 							     dev->ifindex,
855 							     &idev->cnf);
856 		}
857 	}
858 }
859 
860 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
861 {
862 	struct net *net;
863 	int old;
864 
865 	if (!rtnl_trylock())
866 		return restart_syscall();
867 
868 	net = (struct net *)table->extra2;
869 	old = *p;
870 	*p = newf;
871 
872 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
873 		if ((!newf) ^ (!old))
874 			inet6_netconf_notify_devconf(net,
875 						     RTM_NEWNETCONF,
876 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
877 						     NETCONFA_IFINDEX_DEFAULT,
878 						     net->ipv6.devconf_dflt);
879 		rtnl_unlock();
880 		return 0;
881 	}
882 
883 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
884 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
885 		addrconf_linkdown_change(net, newf);
886 		if ((!newf) ^ (!old))
887 			inet6_netconf_notify_devconf(net,
888 						     RTM_NEWNETCONF,
889 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
890 						     NETCONFA_IFINDEX_ALL,
891 						     net->ipv6.devconf_all);
892 	}
893 	rtnl_unlock();
894 
895 	return 1;
896 }
897 
898 #endif
899 
900 /* Nobody refers to this ifaddr, destroy it */
901 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
902 {
903 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
904 
905 #ifdef NET_REFCNT_DEBUG
906 	pr_debug("%s\n", __func__);
907 #endif
908 
909 	in6_dev_put(ifp->idev);
910 
911 	if (cancel_delayed_work(&ifp->dad_work))
912 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
913 			  ifp);
914 
915 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
916 		pr_warn("Freeing alive inet6 address %p\n", ifp);
917 		return;
918 	}
919 	ip6_rt_put(ifp->rt);
920 
921 	kfree_rcu(ifp, rcu);
922 }
923 
924 static void
925 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
926 {
927 	struct list_head *p;
928 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
929 
930 	/*
931 	 * Each device address list is sorted in order of scope -
932 	 * global before linklocal.
933 	 */
934 	list_for_each(p, &idev->addr_list) {
935 		struct inet6_ifaddr *ifa
936 			= list_entry(p, struct inet6_ifaddr, if_list);
937 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
938 			break;
939 	}
940 
941 	list_add_tail_rcu(&ifp->if_list, p);
942 }
943 
944 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
945 {
946 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
947 
948 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
949 }
950 
951 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
952 			       struct net_device *dev, unsigned int hash)
953 {
954 	struct inet6_ifaddr *ifp;
955 
956 	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
957 		if (!net_eq(dev_net(ifp->idev->dev), net))
958 			continue;
959 		if (ipv6_addr_equal(&ifp->addr, addr)) {
960 			if (!dev || ifp->idev->dev == dev)
961 				return true;
962 		}
963 	}
964 	return false;
965 }
966 
967 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
968 {
969 	unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
970 	int err = 0;
971 
972 	spin_lock(&addrconf_hash_lock);
973 
974 	/* Ignore adding duplicate addresses on an interface */
975 	if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
976 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
977 		err = -EEXIST;
978 	} else {
979 		hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
980 	}
981 
982 	spin_unlock(&addrconf_hash_lock);
983 
984 	return err;
985 }
986 
987 /* On success it returns ifp with increased reference count */
988 
989 static struct inet6_ifaddr *
990 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
991 	      const struct in6_addr *peer_addr, int pfxlen,
992 	      int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
993 	      bool can_block, struct netlink_ext_ack *extack)
994 {
995 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
996 	struct net *net = dev_net(idev->dev);
997 	struct inet6_ifaddr *ifa = NULL;
998 	struct rt6_info *rt = NULL;
999 	int err = 0;
1000 	int addr_type = ipv6_addr_type(addr);
1001 
1002 	if (addr_type == IPV6_ADDR_ANY ||
1003 	    addr_type & IPV6_ADDR_MULTICAST ||
1004 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1005 	     addr_type & IPV6_ADDR_LOOPBACK))
1006 		return ERR_PTR(-EADDRNOTAVAIL);
1007 
1008 	if (idev->dead) {
1009 		err = -ENODEV;			/*XXX*/
1010 		goto out;
1011 	}
1012 
1013 	if (idev->cnf.disable_ipv6) {
1014 		err = -EACCES;
1015 		goto out;
1016 	}
1017 
1018 	/* validator notifier needs to be blocking;
1019 	 * do not call in atomic context
1020 	 */
1021 	if (can_block) {
1022 		struct in6_validator_info i6vi = {
1023 			.i6vi_addr = *addr,
1024 			.i6vi_dev = idev,
1025 			.extack = extack,
1026 		};
1027 
1028 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1029 		err = notifier_to_errno(err);
1030 		if (err < 0)
1031 			goto out;
1032 	}
1033 
1034 	ifa = kzalloc(sizeof(*ifa), gfp_flags);
1035 	if (!ifa) {
1036 		err = -ENOBUFS;
1037 		goto out;
1038 	}
1039 
1040 	rt = addrconf_dst_alloc(idev, addr, false);
1041 	if (IS_ERR(rt)) {
1042 		err = PTR_ERR(rt);
1043 		rt = NULL;
1044 		goto out;
1045 	}
1046 
1047 	if (net->ipv6.devconf_all->disable_policy ||
1048 	    idev->cnf.disable_policy)
1049 		rt->dst.flags |= DST_NOPOLICY;
1050 
1051 	neigh_parms_data_state_setall(idev->nd_parms);
1052 
1053 	ifa->addr = *addr;
1054 	if (peer_addr)
1055 		ifa->peer_addr = *peer_addr;
1056 
1057 	spin_lock_init(&ifa->lock);
1058 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1059 	INIT_HLIST_NODE(&ifa->addr_lst);
1060 	ifa->scope = scope;
1061 	ifa->prefix_len = pfxlen;
1062 	ifa->flags = flags;
1063 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1064 	if (!(flags & IFA_F_NODAD))
1065 		ifa->flags |= IFA_F_TENTATIVE;
1066 	ifa->valid_lft = valid_lft;
1067 	ifa->prefered_lft = prefered_lft;
1068 	ifa->cstamp = ifa->tstamp = jiffies;
1069 	ifa->tokenized = false;
1070 
1071 	ifa->rt = rt;
1072 
1073 	ifa->idev = idev;
1074 	in6_dev_hold(idev);
1075 
1076 	/* For caller */
1077 	refcount_set(&ifa->refcnt, 1);
1078 
1079 	rcu_read_lock_bh();
1080 
1081 	err = ipv6_add_addr_hash(idev->dev, ifa);
1082 	if (err < 0) {
1083 		rcu_read_unlock_bh();
1084 		goto out;
1085 	}
1086 
1087 	write_lock(&idev->lock);
1088 
1089 	/* Add to inet6_dev unicast addr list. */
1090 	ipv6_link_dev_addr(idev, ifa);
1091 
1092 	if (ifa->flags&IFA_F_TEMPORARY) {
1093 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1094 		in6_ifa_hold(ifa);
1095 	}
1096 
1097 	in6_ifa_hold(ifa);
1098 	write_unlock(&idev->lock);
1099 
1100 	rcu_read_unlock_bh();
1101 
1102 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1103 out:
1104 	if (unlikely(err < 0)) {
1105 		if (rt)
1106 			ip6_rt_put(rt);
1107 		if (ifa) {
1108 			if (ifa->idev)
1109 				in6_dev_put(ifa->idev);
1110 			kfree(ifa);
1111 		}
1112 		ifa = ERR_PTR(err);
1113 	}
1114 
1115 	return ifa;
1116 }
1117 
1118 enum cleanup_prefix_rt_t {
1119 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1120 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1121 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1122 };
1123 
1124 /*
1125  * Check, whether the prefix for ifp would still need a prefix route
1126  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1127  * constants.
1128  *
1129  * 1) we don't purge prefix if address was not permanent.
1130  *    prefix is managed by its own lifetime.
1131  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1132  * 3) if there are no addresses, delete prefix.
1133  * 4) if there are still other permanent address(es),
1134  *    corresponding prefix is still permanent.
1135  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1136  *    don't purge the prefix, assume user space is managing it.
1137  * 6) otherwise, update prefix lifetime to the
1138  *    longest valid lifetime among the corresponding
1139  *    addresses on the device.
1140  *    Note: subsequent RA will update lifetime.
1141  **/
1142 static enum cleanup_prefix_rt_t
1143 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1144 {
1145 	struct inet6_ifaddr *ifa;
1146 	struct inet6_dev *idev = ifp->idev;
1147 	unsigned long lifetime;
1148 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1149 
1150 	*expires = jiffies;
1151 
1152 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1153 		if (ifa == ifp)
1154 			continue;
1155 		if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1156 				       ifp->prefix_len))
1157 			continue;
1158 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1159 			return CLEANUP_PREFIX_RT_NOP;
1160 
1161 		action = CLEANUP_PREFIX_RT_EXPIRE;
1162 
1163 		spin_lock(&ifa->lock);
1164 
1165 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1166 		/*
1167 		 * Note: Because this address is
1168 		 * not permanent, lifetime <
1169 		 * LONG_MAX / HZ here.
1170 		 */
1171 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1172 			*expires = ifa->tstamp + lifetime * HZ;
1173 		spin_unlock(&ifa->lock);
1174 	}
1175 
1176 	return action;
1177 }
1178 
1179 static void
1180 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1181 {
1182 	struct rt6_info *rt;
1183 
1184 	rt = addrconf_get_prefix_route(&ifp->addr,
1185 				       ifp->prefix_len,
1186 				       ifp->idev->dev,
1187 				       0, RTF_GATEWAY | RTF_DEFAULT);
1188 	if (rt) {
1189 		if (del_rt)
1190 			ip6_del_rt(rt);
1191 		else {
1192 			if (!(rt->rt6i_flags & RTF_EXPIRES))
1193 				rt6_set_expires(rt, expires);
1194 			ip6_rt_put(rt);
1195 		}
1196 	}
1197 }
1198 
1199 
1200 /* This function wants to get referenced ifp and releases it before return */
1201 
1202 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1203 {
1204 	int state;
1205 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1206 	unsigned long expires;
1207 
1208 	ASSERT_RTNL();
1209 
1210 	spin_lock_bh(&ifp->lock);
1211 	state = ifp->state;
1212 	ifp->state = INET6_IFADDR_STATE_DEAD;
1213 	spin_unlock_bh(&ifp->lock);
1214 
1215 	if (state == INET6_IFADDR_STATE_DEAD)
1216 		goto out;
1217 
1218 	spin_lock_bh(&addrconf_hash_lock);
1219 	hlist_del_init_rcu(&ifp->addr_lst);
1220 	spin_unlock_bh(&addrconf_hash_lock);
1221 
1222 	write_lock_bh(&ifp->idev->lock);
1223 
1224 	if (ifp->flags&IFA_F_TEMPORARY) {
1225 		list_del(&ifp->tmp_list);
1226 		if (ifp->ifpub) {
1227 			in6_ifa_put(ifp->ifpub);
1228 			ifp->ifpub = NULL;
1229 		}
1230 		__in6_ifa_put(ifp);
1231 	}
1232 
1233 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1234 		action = check_cleanup_prefix_route(ifp, &expires);
1235 
1236 	list_del_rcu(&ifp->if_list);
1237 	__in6_ifa_put(ifp);
1238 
1239 	write_unlock_bh(&ifp->idev->lock);
1240 
1241 	addrconf_del_dad_work(ifp);
1242 
1243 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1244 
1245 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1246 
1247 	if (action != CLEANUP_PREFIX_RT_NOP) {
1248 		cleanup_prefix_route(ifp, expires,
1249 			action == CLEANUP_PREFIX_RT_DEL);
1250 	}
1251 
1252 	/* clean up prefsrc entries */
1253 	rt6_remove_prefsrc(ifp);
1254 out:
1255 	in6_ifa_put(ifp);
1256 }
1257 
1258 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1259 				struct inet6_ifaddr *ift,
1260 				bool block)
1261 {
1262 	struct inet6_dev *idev = ifp->idev;
1263 	struct in6_addr addr, *tmpaddr;
1264 	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1265 	unsigned long regen_advance;
1266 	int tmp_plen;
1267 	int ret = 0;
1268 	u32 addr_flags;
1269 	unsigned long now = jiffies;
1270 	long max_desync_factor;
1271 	s32 cnf_temp_preferred_lft;
1272 
1273 	write_lock_bh(&idev->lock);
1274 	if (ift) {
1275 		spin_lock_bh(&ift->lock);
1276 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1277 		spin_unlock_bh(&ift->lock);
1278 		tmpaddr = &addr;
1279 	} else {
1280 		tmpaddr = NULL;
1281 	}
1282 retry:
1283 	in6_dev_hold(idev);
1284 	if (idev->cnf.use_tempaddr <= 0) {
1285 		write_unlock_bh(&idev->lock);
1286 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1287 		in6_dev_put(idev);
1288 		ret = -1;
1289 		goto out;
1290 	}
1291 	spin_lock_bh(&ifp->lock);
1292 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1293 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1294 		spin_unlock_bh(&ifp->lock);
1295 		write_unlock_bh(&idev->lock);
1296 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1297 			__func__);
1298 		in6_dev_put(idev);
1299 		ret = -1;
1300 		goto out;
1301 	}
1302 	in6_ifa_hold(ifp);
1303 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1304 	ipv6_try_regen_rndid(idev, tmpaddr);
1305 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
1306 	age = (now - ifp->tstamp) / HZ;
1307 
1308 	regen_advance = idev->cnf.regen_max_retry *
1309 			idev->cnf.dad_transmits *
1310 			NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1311 
1312 	/* recalculate max_desync_factor each time and update
1313 	 * idev->desync_factor if it's larger
1314 	 */
1315 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1316 	max_desync_factor = min_t(__u32,
1317 				  idev->cnf.max_desync_factor,
1318 				  cnf_temp_preferred_lft - regen_advance);
1319 
1320 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1321 		if (max_desync_factor > 0) {
1322 			get_random_bytes(&idev->desync_factor,
1323 					 sizeof(idev->desync_factor));
1324 			idev->desync_factor %= max_desync_factor;
1325 		} else {
1326 			idev->desync_factor = 0;
1327 		}
1328 	}
1329 
1330 	tmp_valid_lft = min_t(__u32,
1331 			      ifp->valid_lft,
1332 			      idev->cnf.temp_valid_lft + age);
1333 	tmp_prefered_lft = cnf_temp_preferred_lft + age -
1334 			    idev->desync_factor;
1335 	tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1336 	tmp_plen = ifp->prefix_len;
1337 	tmp_tstamp = ifp->tstamp;
1338 	spin_unlock_bh(&ifp->lock);
1339 
1340 	write_unlock_bh(&idev->lock);
1341 
1342 	/* A temporary address is created only if this calculated Preferred
1343 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1344 	 * an implementation must not create a temporary address with a zero
1345 	 * Preferred Lifetime.
1346 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1347 	 * temporary addresses being generated.
1348 	 */
1349 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1350 	if (tmp_prefered_lft <= regen_advance + age) {
1351 		in6_ifa_put(ifp);
1352 		in6_dev_put(idev);
1353 		ret = -1;
1354 		goto out;
1355 	}
1356 
1357 	addr_flags = IFA_F_TEMPORARY;
1358 	/* set in addrconf_prefix_rcv() */
1359 	if (ifp->flags & IFA_F_OPTIMISTIC)
1360 		addr_flags |= IFA_F_OPTIMISTIC;
1361 
1362 	ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1363 			    ipv6_addr_scope(&addr), addr_flags,
1364 			    tmp_valid_lft, tmp_prefered_lft, block, NULL);
1365 	if (IS_ERR(ift)) {
1366 		in6_ifa_put(ifp);
1367 		in6_dev_put(idev);
1368 		pr_info("%s: retry temporary address regeneration\n", __func__);
1369 		tmpaddr = &addr;
1370 		write_lock_bh(&idev->lock);
1371 		goto retry;
1372 	}
1373 
1374 	spin_lock_bh(&ift->lock);
1375 	ift->ifpub = ifp;
1376 	ift->cstamp = now;
1377 	ift->tstamp = tmp_tstamp;
1378 	spin_unlock_bh(&ift->lock);
1379 
1380 	addrconf_dad_start(ift);
1381 	in6_ifa_put(ift);
1382 	in6_dev_put(idev);
1383 out:
1384 	return ret;
1385 }
1386 
1387 /*
1388  *	Choose an appropriate source address (RFC3484)
1389  */
1390 enum {
1391 	IPV6_SADDR_RULE_INIT = 0,
1392 	IPV6_SADDR_RULE_LOCAL,
1393 	IPV6_SADDR_RULE_SCOPE,
1394 	IPV6_SADDR_RULE_PREFERRED,
1395 #ifdef CONFIG_IPV6_MIP6
1396 	IPV6_SADDR_RULE_HOA,
1397 #endif
1398 	IPV6_SADDR_RULE_OIF,
1399 	IPV6_SADDR_RULE_LABEL,
1400 	IPV6_SADDR_RULE_PRIVACY,
1401 	IPV6_SADDR_RULE_ORCHID,
1402 	IPV6_SADDR_RULE_PREFIX,
1403 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1404 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1405 #endif
1406 	IPV6_SADDR_RULE_MAX
1407 };
1408 
1409 struct ipv6_saddr_score {
1410 	int			rule;
1411 	int			addr_type;
1412 	struct inet6_ifaddr	*ifa;
1413 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1414 	int			scopedist;
1415 	int			matchlen;
1416 };
1417 
1418 struct ipv6_saddr_dst {
1419 	const struct in6_addr *addr;
1420 	int ifindex;
1421 	int scope;
1422 	int label;
1423 	unsigned int prefs;
1424 };
1425 
1426 static inline int ipv6_saddr_preferred(int type)
1427 {
1428 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1429 		return 1;
1430 	return 0;
1431 }
1432 
1433 static bool ipv6_use_optimistic_addr(struct net *net,
1434 				     struct inet6_dev *idev)
1435 {
1436 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1437 	if (!idev)
1438 		return false;
1439 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1440 		return false;
1441 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1442 		return false;
1443 
1444 	return true;
1445 #else
1446 	return false;
1447 #endif
1448 }
1449 
1450 static bool ipv6_allow_optimistic_dad(struct net *net,
1451 				      struct inet6_dev *idev)
1452 {
1453 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1454 	if (!idev)
1455 		return false;
1456 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1457 		return false;
1458 
1459 	return true;
1460 #else
1461 	return false;
1462 #endif
1463 }
1464 
1465 static int ipv6_get_saddr_eval(struct net *net,
1466 			       struct ipv6_saddr_score *score,
1467 			       struct ipv6_saddr_dst *dst,
1468 			       int i)
1469 {
1470 	int ret;
1471 
1472 	if (i <= score->rule) {
1473 		switch (i) {
1474 		case IPV6_SADDR_RULE_SCOPE:
1475 			ret = score->scopedist;
1476 			break;
1477 		case IPV6_SADDR_RULE_PREFIX:
1478 			ret = score->matchlen;
1479 			break;
1480 		default:
1481 			ret = !!test_bit(i, score->scorebits);
1482 		}
1483 		goto out;
1484 	}
1485 
1486 	switch (i) {
1487 	case IPV6_SADDR_RULE_INIT:
1488 		/* Rule 0: remember if hiscore is not ready yet */
1489 		ret = !!score->ifa;
1490 		break;
1491 	case IPV6_SADDR_RULE_LOCAL:
1492 		/* Rule 1: Prefer same address */
1493 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1494 		break;
1495 	case IPV6_SADDR_RULE_SCOPE:
1496 		/* Rule 2: Prefer appropriate scope
1497 		 *
1498 		 *      ret
1499 		 *       ^
1500 		 *    -1 |  d 15
1501 		 *    ---+--+-+---> scope
1502 		 *       |
1503 		 *       |             d is scope of the destination.
1504 		 *  B-d  |  \
1505 		 *       |   \      <- smaller scope is better if
1506 		 *  B-15 |    \        if scope is enough for destination.
1507 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1508 		 * d-C-1 | /
1509 		 *       |/         <- greater is better
1510 		 *   -C  /             if scope is not enough for destination.
1511 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1512 		 *
1513 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1514 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1515 		 * Assume B = 0 and we get C > 29.
1516 		 */
1517 		ret = __ipv6_addr_src_scope(score->addr_type);
1518 		if (ret >= dst->scope)
1519 			ret = -ret;
1520 		else
1521 			ret -= 128;	/* 30 is enough */
1522 		score->scopedist = ret;
1523 		break;
1524 	case IPV6_SADDR_RULE_PREFERRED:
1525 	    {
1526 		/* Rule 3: Avoid deprecated and optimistic addresses */
1527 		u8 avoid = IFA_F_DEPRECATED;
1528 
1529 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1530 			avoid |= IFA_F_OPTIMISTIC;
1531 		ret = ipv6_saddr_preferred(score->addr_type) ||
1532 		      !(score->ifa->flags & avoid);
1533 		break;
1534 	    }
1535 #ifdef CONFIG_IPV6_MIP6
1536 	case IPV6_SADDR_RULE_HOA:
1537 	    {
1538 		/* Rule 4: Prefer home address */
1539 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1540 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1541 		break;
1542 	    }
1543 #endif
1544 	case IPV6_SADDR_RULE_OIF:
1545 		/* Rule 5: Prefer outgoing interface */
1546 		ret = (!dst->ifindex ||
1547 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1548 		break;
1549 	case IPV6_SADDR_RULE_LABEL:
1550 		/* Rule 6: Prefer matching label */
1551 		ret = ipv6_addr_label(net,
1552 				      &score->ifa->addr, score->addr_type,
1553 				      score->ifa->idev->dev->ifindex) == dst->label;
1554 		break;
1555 	case IPV6_SADDR_RULE_PRIVACY:
1556 	    {
1557 		/* Rule 7: Prefer public address
1558 		 * Note: prefer temporary address if use_tempaddr >= 2
1559 		 */
1560 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1561 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1562 				score->ifa->idev->cnf.use_tempaddr >= 2;
1563 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1564 		break;
1565 	    }
1566 	case IPV6_SADDR_RULE_ORCHID:
1567 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1568 		 *	    non-ORCHID vs non-ORCHID
1569 		 */
1570 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1571 			ipv6_addr_orchid(dst->addr));
1572 		break;
1573 	case IPV6_SADDR_RULE_PREFIX:
1574 		/* Rule 8: Use longest matching prefix */
1575 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1576 		if (ret > score->ifa->prefix_len)
1577 			ret = score->ifa->prefix_len;
1578 		score->matchlen = ret;
1579 		break;
1580 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1581 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1582 		/* Optimistic addresses still have lower precedence than other
1583 		 * preferred addresses.
1584 		 */
1585 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1586 		break;
1587 #endif
1588 	default:
1589 		ret = 0;
1590 	}
1591 
1592 	if (ret)
1593 		__set_bit(i, score->scorebits);
1594 	score->rule = i;
1595 out:
1596 	return ret;
1597 }
1598 
1599 static int __ipv6_dev_get_saddr(struct net *net,
1600 				struct ipv6_saddr_dst *dst,
1601 				struct inet6_dev *idev,
1602 				struct ipv6_saddr_score *scores,
1603 				int hiscore_idx)
1604 {
1605 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1606 
1607 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1608 		int i;
1609 
1610 		/*
1611 		 * - Tentative Address (RFC2462 section 5.4)
1612 		 *  - A tentative address is not considered
1613 		 *    "assigned to an interface" in the traditional
1614 		 *    sense, unless it is also flagged as optimistic.
1615 		 * - Candidate Source Address (section 4)
1616 		 *  - In any case, anycast addresses, multicast
1617 		 *    addresses, and the unspecified address MUST
1618 		 *    NOT be included in a candidate set.
1619 		 */
1620 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1621 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1622 			continue;
1623 
1624 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1625 
1626 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1627 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1628 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1629 					    idev->dev->name);
1630 			continue;
1631 		}
1632 
1633 		score->rule = -1;
1634 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1635 
1636 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1637 			int minihiscore, miniscore;
1638 
1639 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1640 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1641 
1642 			if (minihiscore > miniscore) {
1643 				if (i == IPV6_SADDR_RULE_SCOPE &&
1644 				    score->scopedist > 0) {
1645 					/*
1646 					 * special case:
1647 					 * each remaining entry
1648 					 * has too small (not enough)
1649 					 * scope, because ifa entries
1650 					 * are sorted by their scope
1651 					 * values.
1652 					 */
1653 					goto out;
1654 				}
1655 				break;
1656 			} else if (minihiscore < miniscore) {
1657 				swap(hiscore, score);
1658 				hiscore_idx = 1 - hiscore_idx;
1659 
1660 				/* restore our iterator */
1661 				score->ifa = hiscore->ifa;
1662 
1663 				break;
1664 			}
1665 		}
1666 	}
1667 out:
1668 	return hiscore_idx;
1669 }
1670 
1671 static int ipv6_get_saddr_master(struct net *net,
1672 				 const struct net_device *dst_dev,
1673 				 const struct net_device *master,
1674 				 struct ipv6_saddr_dst *dst,
1675 				 struct ipv6_saddr_score *scores,
1676 				 int hiscore_idx)
1677 {
1678 	struct inet6_dev *idev;
1679 
1680 	idev = __in6_dev_get(dst_dev);
1681 	if (idev)
1682 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1683 						   scores, hiscore_idx);
1684 
1685 	idev = __in6_dev_get(master);
1686 	if (idev)
1687 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1688 						   scores, hiscore_idx);
1689 
1690 	return hiscore_idx;
1691 }
1692 
1693 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1694 		       const struct in6_addr *daddr, unsigned int prefs,
1695 		       struct in6_addr *saddr)
1696 {
1697 	struct ipv6_saddr_score scores[2], *hiscore;
1698 	struct ipv6_saddr_dst dst;
1699 	struct inet6_dev *idev;
1700 	struct net_device *dev;
1701 	int dst_type;
1702 	bool use_oif_addr = false;
1703 	int hiscore_idx = 0;
1704 	int ret = 0;
1705 
1706 	dst_type = __ipv6_addr_type(daddr);
1707 	dst.addr = daddr;
1708 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1709 	dst.scope = __ipv6_addr_src_scope(dst_type);
1710 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1711 	dst.prefs = prefs;
1712 
1713 	scores[hiscore_idx].rule = -1;
1714 	scores[hiscore_idx].ifa = NULL;
1715 
1716 	rcu_read_lock();
1717 
1718 	/* Candidate Source Address (section 4)
1719 	 *  - multicast and link-local destination address,
1720 	 *    the set of candidate source address MUST only
1721 	 *    include addresses assigned to interfaces
1722 	 *    belonging to the same link as the outgoing
1723 	 *    interface.
1724 	 * (- For site-local destination addresses, the
1725 	 *    set of candidate source addresses MUST only
1726 	 *    include addresses assigned to interfaces
1727 	 *    belonging to the same site as the outgoing
1728 	 *    interface.)
1729 	 *  - "It is RECOMMENDED that the candidate source addresses
1730 	 *    be the set of unicast addresses assigned to the
1731 	 *    interface that will be used to send to the destination
1732 	 *    (the 'outgoing' interface)." (RFC 6724)
1733 	 */
1734 	if (dst_dev) {
1735 		idev = __in6_dev_get(dst_dev);
1736 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1737 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1738 		    (idev && idev->cnf.use_oif_addrs_only)) {
1739 			use_oif_addr = true;
1740 		}
1741 	}
1742 
1743 	if (use_oif_addr) {
1744 		if (idev)
1745 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1746 	} else {
1747 		const struct net_device *master;
1748 		int master_idx = 0;
1749 
1750 		/* if dst_dev exists and is enslaved to an L3 device, then
1751 		 * prefer addresses from dst_dev and then the master over
1752 		 * any other enslaved devices in the L3 domain.
1753 		 */
1754 		master = l3mdev_master_dev_rcu(dst_dev);
1755 		if (master) {
1756 			master_idx = master->ifindex;
1757 
1758 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1759 							    master, &dst,
1760 							    scores, hiscore_idx);
1761 
1762 			if (scores[hiscore_idx].ifa)
1763 				goto out;
1764 		}
1765 
1766 		for_each_netdev_rcu(net, dev) {
1767 			/* only consider addresses on devices in the
1768 			 * same L3 domain
1769 			 */
1770 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1771 				continue;
1772 			idev = __in6_dev_get(dev);
1773 			if (!idev)
1774 				continue;
1775 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1776 		}
1777 	}
1778 
1779 out:
1780 	hiscore = &scores[hiscore_idx];
1781 	if (!hiscore->ifa)
1782 		ret = -EADDRNOTAVAIL;
1783 	else
1784 		*saddr = hiscore->ifa->addr;
1785 
1786 	rcu_read_unlock();
1787 	return ret;
1788 }
1789 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1790 
1791 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1792 		      u32 banned_flags)
1793 {
1794 	struct inet6_ifaddr *ifp;
1795 	int err = -EADDRNOTAVAIL;
1796 
1797 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1798 		if (ifp->scope > IFA_LINK)
1799 			break;
1800 		if (ifp->scope == IFA_LINK &&
1801 		    !(ifp->flags & banned_flags)) {
1802 			*addr = ifp->addr;
1803 			err = 0;
1804 			break;
1805 		}
1806 	}
1807 	return err;
1808 }
1809 
1810 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1811 		    u32 banned_flags)
1812 {
1813 	struct inet6_dev *idev;
1814 	int err = -EADDRNOTAVAIL;
1815 
1816 	rcu_read_lock();
1817 	idev = __in6_dev_get(dev);
1818 	if (idev) {
1819 		read_lock_bh(&idev->lock);
1820 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1821 		read_unlock_bh(&idev->lock);
1822 	}
1823 	rcu_read_unlock();
1824 	return err;
1825 }
1826 
1827 static int ipv6_count_addresses(const struct inet6_dev *idev)
1828 {
1829 	const struct inet6_ifaddr *ifp;
1830 	int cnt = 0;
1831 
1832 	rcu_read_lock();
1833 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1834 		cnt++;
1835 	rcu_read_unlock();
1836 	return cnt;
1837 }
1838 
1839 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1840 		  const struct net_device *dev, int strict)
1841 {
1842 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1843 				       strict, IFA_F_TENTATIVE);
1844 }
1845 EXPORT_SYMBOL(ipv6_chk_addr);
1846 
1847 /* device argument is used to find the L3 domain of interest. If
1848  * skip_dev_check is set, then the ifp device is not checked against
1849  * the passed in dev argument. So the 2 cases for addresses checks are:
1850  *   1. does the address exist in the L3 domain that dev is part of
1851  *      (skip_dev_check = true), or
1852  *
1853  *   2. does the address exist on the specific device
1854  *      (skip_dev_check = false)
1855  */
1856 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1857 			    const struct net_device *dev, bool skip_dev_check,
1858 			    int strict, u32 banned_flags)
1859 {
1860 	unsigned int hash = inet6_addr_hash(net, addr);
1861 	const struct net_device *l3mdev;
1862 	struct inet6_ifaddr *ifp;
1863 	u32 ifp_flags;
1864 
1865 	rcu_read_lock();
1866 
1867 	l3mdev = l3mdev_master_dev_rcu(dev);
1868 	if (skip_dev_check)
1869 		dev = NULL;
1870 
1871 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1872 		if (!net_eq(dev_net(ifp->idev->dev), net))
1873 			continue;
1874 
1875 		if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1876 			continue;
1877 
1878 		/* Decouple optimistic from tentative for evaluation here.
1879 		 * Ban optimistic addresses explicitly, when required.
1880 		 */
1881 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1882 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1883 			    : ifp->flags;
1884 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1885 		    !(ifp_flags&banned_flags) &&
1886 		    (!dev || ifp->idev->dev == dev ||
1887 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1888 			rcu_read_unlock();
1889 			return 1;
1890 		}
1891 	}
1892 
1893 	rcu_read_unlock();
1894 	return 0;
1895 }
1896 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1897 
1898 
1899 /* Compares an address/prefix_len with addresses on device @dev.
1900  * If one is found it returns true.
1901  */
1902 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1903 	const unsigned int prefix_len, struct net_device *dev)
1904 {
1905 	const struct inet6_ifaddr *ifa;
1906 	const struct inet6_dev *idev;
1907 	bool ret = false;
1908 
1909 	rcu_read_lock();
1910 	idev = __in6_dev_get(dev);
1911 	if (idev) {
1912 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1913 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1914 			if (ret)
1915 				break;
1916 		}
1917 	}
1918 	rcu_read_unlock();
1919 
1920 	return ret;
1921 }
1922 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1923 
1924 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1925 {
1926 	const struct inet6_ifaddr *ifa;
1927 	const struct inet6_dev *idev;
1928 	int	onlink;
1929 
1930 	onlink = 0;
1931 	rcu_read_lock();
1932 	idev = __in6_dev_get(dev);
1933 	if (idev) {
1934 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1935 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1936 						   ifa->prefix_len);
1937 			if (onlink)
1938 				break;
1939 		}
1940 	}
1941 	rcu_read_unlock();
1942 	return onlink;
1943 }
1944 EXPORT_SYMBOL(ipv6_chk_prefix);
1945 
1946 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1947 				     struct net_device *dev, int strict)
1948 {
1949 	unsigned int hash = inet6_addr_hash(net, addr);
1950 	struct inet6_ifaddr *ifp, *result = NULL;
1951 
1952 	rcu_read_lock();
1953 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1954 		if (!net_eq(dev_net(ifp->idev->dev), net))
1955 			continue;
1956 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1957 			if (!dev || ifp->idev->dev == dev ||
1958 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1959 				result = ifp;
1960 				in6_ifa_hold(ifp);
1961 				break;
1962 			}
1963 		}
1964 	}
1965 	rcu_read_unlock();
1966 
1967 	return result;
1968 }
1969 
1970 /* Gets referenced address, destroys ifaddr */
1971 
1972 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1973 {
1974 	if (dad_failed)
1975 		ifp->flags |= IFA_F_DADFAILED;
1976 
1977 	if (ifp->flags&IFA_F_TEMPORARY) {
1978 		struct inet6_ifaddr *ifpub;
1979 		spin_lock_bh(&ifp->lock);
1980 		ifpub = ifp->ifpub;
1981 		if (ifpub) {
1982 			in6_ifa_hold(ifpub);
1983 			spin_unlock_bh(&ifp->lock);
1984 			ipv6_create_tempaddr(ifpub, ifp, true);
1985 			in6_ifa_put(ifpub);
1986 		} else {
1987 			spin_unlock_bh(&ifp->lock);
1988 		}
1989 		ipv6_del_addr(ifp);
1990 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1991 		spin_lock_bh(&ifp->lock);
1992 		addrconf_del_dad_work(ifp);
1993 		ifp->flags |= IFA_F_TENTATIVE;
1994 		if (dad_failed)
1995 			ifp->flags &= ~IFA_F_OPTIMISTIC;
1996 		spin_unlock_bh(&ifp->lock);
1997 		if (dad_failed)
1998 			ipv6_ifa_notify(0, ifp);
1999 		in6_ifa_put(ifp);
2000 	} else {
2001 		ipv6_del_addr(ifp);
2002 	}
2003 }
2004 
2005 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2006 {
2007 	int err = -ENOENT;
2008 
2009 	spin_lock_bh(&ifp->lock);
2010 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2011 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2012 		err = 0;
2013 	}
2014 	spin_unlock_bh(&ifp->lock);
2015 
2016 	return err;
2017 }
2018 
2019 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2020 {
2021 	struct inet6_dev *idev = ifp->idev;
2022 	struct net *net = dev_net(ifp->idev->dev);
2023 
2024 	if (addrconf_dad_end(ifp)) {
2025 		in6_ifa_put(ifp);
2026 		return;
2027 	}
2028 
2029 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2030 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2031 
2032 	spin_lock_bh(&ifp->lock);
2033 
2034 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2035 		int scope = ifp->scope;
2036 		u32 flags = ifp->flags;
2037 		struct in6_addr new_addr;
2038 		struct inet6_ifaddr *ifp2;
2039 		u32 valid_lft, preferred_lft;
2040 		int pfxlen = ifp->prefix_len;
2041 		int retries = ifp->stable_privacy_retry + 1;
2042 
2043 		if (retries > net->ipv6.sysctl.idgen_retries) {
2044 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2045 					     ifp->idev->dev->name);
2046 			goto errdad;
2047 		}
2048 
2049 		new_addr = ifp->addr;
2050 		if (ipv6_generate_stable_address(&new_addr, retries,
2051 						 idev))
2052 			goto errdad;
2053 
2054 		valid_lft = ifp->valid_lft;
2055 		preferred_lft = ifp->prefered_lft;
2056 
2057 		spin_unlock_bh(&ifp->lock);
2058 
2059 		if (idev->cnf.max_addresses &&
2060 		    ipv6_count_addresses(idev) >=
2061 		    idev->cnf.max_addresses)
2062 			goto lock_errdad;
2063 
2064 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2065 				     ifp->idev->dev->name);
2066 
2067 		ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2068 				     scope, flags, valid_lft,
2069 				     preferred_lft, false, NULL);
2070 		if (IS_ERR(ifp2))
2071 			goto lock_errdad;
2072 
2073 		spin_lock_bh(&ifp2->lock);
2074 		ifp2->stable_privacy_retry = retries;
2075 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2076 		spin_unlock_bh(&ifp2->lock);
2077 
2078 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2079 		in6_ifa_put(ifp2);
2080 lock_errdad:
2081 		spin_lock_bh(&ifp->lock);
2082 	}
2083 
2084 errdad:
2085 	/* transition from _POSTDAD to _ERRDAD */
2086 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2087 	spin_unlock_bh(&ifp->lock);
2088 
2089 	addrconf_mod_dad_work(ifp, 0);
2090 	in6_ifa_put(ifp);
2091 }
2092 
2093 /* Join to solicited addr multicast group.
2094  * caller must hold RTNL */
2095 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2096 {
2097 	struct in6_addr maddr;
2098 
2099 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2100 		return;
2101 
2102 	addrconf_addr_solict_mult(addr, &maddr);
2103 	ipv6_dev_mc_inc(dev, &maddr);
2104 }
2105 
2106 /* caller must hold RTNL */
2107 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2108 {
2109 	struct in6_addr maddr;
2110 
2111 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2112 		return;
2113 
2114 	addrconf_addr_solict_mult(addr, &maddr);
2115 	__ipv6_dev_mc_dec(idev, &maddr);
2116 }
2117 
2118 /* caller must hold RTNL */
2119 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2120 {
2121 	struct in6_addr addr;
2122 
2123 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2124 		return;
2125 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2126 	if (ipv6_addr_any(&addr))
2127 		return;
2128 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2129 }
2130 
2131 /* caller must hold RTNL */
2132 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2133 {
2134 	struct in6_addr addr;
2135 
2136 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2137 		return;
2138 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2139 	if (ipv6_addr_any(&addr))
2140 		return;
2141 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2142 }
2143 
2144 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2145 {
2146 	switch (dev->addr_len) {
2147 	case ETH_ALEN:
2148 		memcpy(eui, dev->dev_addr, 3);
2149 		eui[3] = 0xFF;
2150 		eui[4] = 0xFE;
2151 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2152 		break;
2153 	case EUI64_ADDR_LEN:
2154 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2155 		eui[0] ^= 2;
2156 		break;
2157 	default:
2158 		return -1;
2159 	}
2160 
2161 	return 0;
2162 }
2163 
2164 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2165 {
2166 	union fwnet_hwaddr *ha;
2167 
2168 	if (dev->addr_len != FWNET_ALEN)
2169 		return -1;
2170 
2171 	ha = (union fwnet_hwaddr *)dev->dev_addr;
2172 
2173 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2174 	eui[0] ^= 2;
2175 	return 0;
2176 }
2177 
2178 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2179 {
2180 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2181 	if (dev->addr_len != ARCNET_ALEN)
2182 		return -1;
2183 	memset(eui, 0, 7);
2184 	eui[7] = *(u8 *)dev->dev_addr;
2185 	return 0;
2186 }
2187 
2188 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2189 {
2190 	if (dev->addr_len != INFINIBAND_ALEN)
2191 		return -1;
2192 	memcpy(eui, dev->dev_addr + 12, 8);
2193 	eui[0] |= 2;
2194 	return 0;
2195 }
2196 
2197 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2198 {
2199 	if (addr == 0)
2200 		return -1;
2201 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2202 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2203 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2204 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2205 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2206 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2207 	eui[1] = 0;
2208 	eui[2] = 0x5E;
2209 	eui[3] = 0xFE;
2210 	memcpy(eui + 4, &addr, 4);
2211 	return 0;
2212 }
2213 
2214 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2215 {
2216 	if (dev->priv_flags & IFF_ISATAP)
2217 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2218 	return -1;
2219 }
2220 
2221 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2222 {
2223 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2224 }
2225 
2226 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2227 {
2228 	memcpy(eui, dev->perm_addr, 3);
2229 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2230 	eui[3] = 0xFF;
2231 	eui[4] = 0xFE;
2232 	eui[0] ^= 2;
2233 	return 0;
2234 }
2235 
2236 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2237 {
2238 	switch (dev->type) {
2239 	case ARPHRD_ETHER:
2240 	case ARPHRD_FDDI:
2241 		return addrconf_ifid_eui48(eui, dev);
2242 	case ARPHRD_ARCNET:
2243 		return addrconf_ifid_arcnet(eui, dev);
2244 	case ARPHRD_INFINIBAND:
2245 		return addrconf_ifid_infiniband(eui, dev);
2246 	case ARPHRD_SIT:
2247 		return addrconf_ifid_sit(eui, dev);
2248 	case ARPHRD_IPGRE:
2249 	case ARPHRD_TUNNEL:
2250 		return addrconf_ifid_gre(eui, dev);
2251 	case ARPHRD_6LOWPAN:
2252 		return addrconf_ifid_6lowpan(eui, dev);
2253 	case ARPHRD_IEEE1394:
2254 		return addrconf_ifid_ieee1394(eui, dev);
2255 	case ARPHRD_TUNNEL6:
2256 	case ARPHRD_IP6GRE:
2257 		return addrconf_ifid_ip6tnl(eui, dev);
2258 	}
2259 	return -1;
2260 }
2261 
2262 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2263 {
2264 	int err = -1;
2265 	struct inet6_ifaddr *ifp;
2266 
2267 	read_lock_bh(&idev->lock);
2268 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2269 		if (ifp->scope > IFA_LINK)
2270 			break;
2271 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2272 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2273 			err = 0;
2274 			break;
2275 		}
2276 	}
2277 	read_unlock_bh(&idev->lock);
2278 	return err;
2279 }
2280 
2281 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2282 static void ipv6_regen_rndid(struct inet6_dev *idev)
2283 {
2284 regen:
2285 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
2286 	idev->rndid[0] &= ~0x02;
2287 
2288 	/*
2289 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2290 	 * check if generated address is not inappropriate
2291 	 *
2292 	 *  - Reserved subnet anycast (RFC 2526)
2293 	 *	11111101 11....11 1xxxxxxx
2294 	 *  - ISATAP (RFC4214) 6.1
2295 	 *	00-00-5E-FE-xx-xx-xx-xx
2296 	 *  - value 0
2297 	 *  - XXX: already assigned to an address on the device
2298 	 */
2299 	if (idev->rndid[0] == 0xfd &&
2300 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2301 	    (idev->rndid[7]&0x80))
2302 		goto regen;
2303 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2304 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2305 			goto regen;
2306 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2307 			goto regen;
2308 	}
2309 }
2310 
2311 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2312 {
2313 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2314 		ipv6_regen_rndid(idev);
2315 }
2316 
2317 /*
2318  *	Add prefix route.
2319  */
2320 
2321 static void
2322 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2323 		      unsigned long expires, u32 flags)
2324 {
2325 	struct fib6_config cfg = {
2326 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2327 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2328 		.fc_ifindex = dev->ifindex,
2329 		.fc_expires = expires,
2330 		.fc_dst_len = plen,
2331 		.fc_flags = RTF_UP | flags,
2332 		.fc_nlinfo.nl_net = dev_net(dev),
2333 		.fc_protocol = RTPROT_KERNEL,
2334 	};
2335 
2336 	cfg.fc_dst = *pfx;
2337 
2338 	/* Prevent useless cloning on PtP SIT.
2339 	   This thing is done here expecting that the whole
2340 	   class of non-broadcast devices need not cloning.
2341 	 */
2342 #if IS_ENABLED(CONFIG_IPV6_SIT)
2343 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2344 		cfg.fc_flags |= RTF_NONEXTHOP;
2345 #endif
2346 
2347 	ip6_route_add(&cfg, NULL);
2348 }
2349 
2350 
2351 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2352 						  int plen,
2353 						  const struct net_device *dev,
2354 						  u32 flags, u32 noflags)
2355 {
2356 	struct fib6_node *fn;
2357 	struct rt6_info *rt = NULL;
2358 	struct fib6_table *table;
2359 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2360 
2361 	table = fib6_get_table(dev_net(dev), tb_id);
2362 	if (!table)
2363 		return NULL;
2364 
2365 	rcu_read_lock();
2366 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2367 	if (!fn)
2368 		goto out;
2369 
2370 	for_each_fib6_node_rt_rcu(fn) {
2371 		if (rt->dst.dev->ifindex != dev->ifindex)
2372 			continue;
2373 		if ((rt->rt6i_flags & flags) != flags)
2374 			continue;
2375 		if ((rt->rt6i_flags & noflags) != 0)
2376 			continue;
2377 		if (!dst_hold_safe(&rt->dst))
2378 			rt = NULL;
2379 		break;
2380 	}
2381 out:
2382 	rcu_read_unlock();
2383 	return rt;
2384 }
2385 
2386 
2387 /* Create "default" multicast route to the interface */
2388 
2389 static void addrconf_add_mroute(struct net_device *dev)
2390 {
2391 	struct fib6_config cfg = {
2392 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2393 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2394 		.fc_ifindex = dev->ifindex,
2395 		.fc_dst_len = 8,
2396 		.fc_flags = RTF_UP,
2397 		.fc_nlinfo.nl_net = dev_net(dev),
2398 	};
2399 
2400 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2401 
2402 	ip6_route_add(&cfg, NULL);
2403 }
2404 
2405 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2406 {
2407 	struct inet6_dev *idev;
2408 
2409 	ASSERT_RTNL();
2410 
2411 	idev = ipv6_find_idev(dev);
2412 	if (!idev)
2413 		return ERR_PTR(-ENOBUFS);
2414 
2415 	if (idev->cnf.disable_ipv6)
2416 		return ERR_PTR(-EACCES);
2417 
2418 	/* Add default multicast route */
2419 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2420 		addrconf_add_mroute(dev);
2421 
2422 	return idev;
2423 }
2424 
2425 static void manage_tempaddrs(struct inet6_dev *idev,
2426 			     struct inet6_ifaddr *ifp,
2427 			     __u32 valid_lft, __u32 prefered_lft,
2428 			     bool create, unsigned long now)
2429 {
2430 	u32 flags;
2431 	struct inet6_ifaddr *ift;
2432 
2433 	read_lock_bh(&idev->lock);
2434 	/* update all temporary addresses in the list */
2435 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2436 		int age, max_valid, max_prefered;
2437 
2438 		if (ifp != ift->ifpub)
2439 			continue;
2440 
2441 		/* RFC 4941 section 3.3:
2442 		 * If a received option will extend the lifetime of a public
2443 		 * address, the lifetimes of temporary addresses should
2444 		 * be extended, subject to the overall constraint that no
2445 		 * temporary addresses should ever remain "valid" or "preferred"
2446 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2447 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2448 		 */
2449 		age = (now - ift->cstamp) / HZ;
2450 		max_valid = idev->cnf.temp_valid_lft - age;
2451 		if (max_valid < 0)
2452 			max_valid = 0;
2453 
2454 		max_prefered = idev->cnf.temp_prefered_lft -
2455 			       idev->desync_factor - age;
2456 		if (max_prefered < 0)
2457 			max_prefered = 0;
2458 
2459 		if (valid_lft > max_valid)
2460 			valid_lft = max_valid;
2461 
2462 		if (prefered_lft > max_prefered)
2463 			prefered_lft = max_prefered;
2464 
2465 		spin_lock(&ift->lock);
2466 		flags = ift->flags;
2467 		ift->valid_lft = valid_lft;
2468 		ift->prefered_lft = prefered_lft;
2469 		ift->tstamp = now;
2470 		if (prefered_lft > 0)
2471 			ift->flags &= ~IFA_F_DEPRECATED;
2472 
2473 		spin_unlock(&ift->lock);
2474 		if (!(flags&IFA_F_TENTATIVE))
2475 			ipv6_ifa_notify(0, ift);
2476 	}
2477 
2478 	if ((create || list_empty(&idev->tempaddr_list)) &&
2479 	    idev->cnf.use_tempaddr > 0) {
2480 		/* When a new public address is created as described
2481 		 * in [ADDRCONF], also create a new temporary address.
2482 		 * Also create a temporary address if it's enabled but
2483 		 * no temporary address currently exists.
2484 		 */
2485 		read_unlock_bh(&idev->lock);
2486 		ipv6_create_tempaddr(ifp, NULL, false);
2487 	} else {
2488 		read_unlock_bh(&idev->lock);
2489 	}
2490 }
2491 
2492 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2493 {
2494 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2495 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2496 }
2497 
2498 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2499 				 const struct prefix_info *pinfo,
2500 				 struct inet6_dev *in6_dev,
2501 				 const struct in6_addr *addr, int addr_type,
2502 				 u32 addr_flags, bool sllao, bool tokenized,
2503 				 __u32 valid_lft, u32 prefered_lft)
2504 {
2505 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2506 	int create = 0, update_lft = 0;
2507 
2508 	if (!ifp && valid_lft) {
2509 		int max_addresses = in6_dev->cnf.max_addresses;
2510 
2511 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2512 		if ((net->ipv6.devconf_all->optimistic_dad ||
2513 		     in6_dev->cnf.optimistic_dad) &&
2514 		    !net->ipv6.devconf_all->forwarding && sllao)
2515 			addr_flags |= IFA_F_OPTIMISTIC;
2516 #endif
2517 
2518 		/* Do not allow to create too much of autoconfigured
2519 		 * addresses; this would be too easy way to crash kernel.
2520 		 */
2521 		if (!max_addresses ||
2522 		    ipv6_count_addresses(in6_dev) < max_addresses)
2523 			ifp = ipv6_add_addr(in6_dev, addr, NULL,
2524 					    pinfo->prefix_len,
2525 					    addr_type&IPV6_ADDR_SCOPE_MASK,
2526 					    addr_flags, valid_lft,
2527 					    prefered_lft, false, NULL);
2528 
2529 		if (IS_ERR_OR_NULL(ifp))
2530 			return -1;
2531 
2532 		update_lft = 0;
2533 		create = 1;
2534 		spin_lock_bh(&ifp->lock);
2535 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2536 		ifp->cstamp = jiffies;
2537 		ifp->tokenized = tokenized;
2538 		spin_unlock_bh(&ifp->lock);
2539 		addrconf_dad_start(ifp);
2540 	}
2541 
2542 	if (ifp) {
2543 		u32 flags;
2544 		unsigned long now;
2545 		u32 stored_lft;
2546 
2547 		/* update lifetime (RFC2462 5.5.3 e) */
2548 		spin_lock_bh(&ifp->lock);
2549 		now = jiffies;
2550 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2551 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2552 		else
2553 			stored_lft = 0;
2554 		if (!update_lft && !create && stored_lft) {
2555 			const u32 minimum_lft = min_t(u32,
2556 				stored_lft, MIN_VALID_LIFETIME);
2557 			valid_lft = max(valid_lft, minimum_lft);
2558 
2559 			/* RFC4862 Section 5.5.3e:
2560 			 * "Note that the preferred lifetime of the
2561 			 *  corresponding address is always reset to
2562 			 *  the Preferred Lifetime in the received
2563 			 *  Prefix Information option, regardless of
2564 			 *  whether the valid lifetime is also reset or
2565 			 *  ignored."
2566 			 *
2567 			 * So we should always update prefered_lft here.
2568 			 */
2569 			update_lft = 1;
2570 		}
2571 
2572 		if (update_lft) {
2573 			ifp->valid_lft = valid_lft;
2574 			ifp->prefered_lft = prefered_lft;
2575 			ifp->tstamp = now;
2576 			flags = ifp->flags;
2577 			ifp->flags &= ~IFA_F_DEPRECATED;
2578 			spin_unlock_bh(&ifp->lock);
2579 
2580 			if (!(flags&IFA_F_TENTATIVE))
2581 				ipv6_ifa_notify(0, ifp);
2582 		} else
2583 			spin_unlock_bh(&ifp->lock);
2584 
2585 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2586 				 create, now);
2587 
2588 		in6_ifa_put(ifp);
2589 		addrconf_verify();
2590 	}
2591 
2592 	return 0;
2593 }
2594 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2595 
2596 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2597 {
2598 	struct prefix_info *pinfo;
2599 	__u32 valid_lft;
2600 	__u32 prefered_lft;
2601 	int addr_type, err;
2602 	u32 addr_flags = 0;
2603 	struct inet6_dev *in6_dev;
2604 	struct net *net = dev_net(dev);
2605 
2606 	pinfo = (struct prefix_info *) opt;
2607 
2608 	if (len < sizeof(struct prefix_info)) {
2609 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2610 		return;
2611 	}
2612 
2613 	/*
2614 	 *	Validation checks ([ADDRCONF], page 19)
2615 	 */
2616 
2617 	addr_type = ipv6_addr_type(&pinfo->prefix);
2618 
2619 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2620 		return;
2621 
2622 	valid_lft = ntohl(pinfo->valid);
2623 	prefered_lft = ntohl(pinfo->prefered);
2624 
2625 	if (prefered_lft > valid_lft) {
2626 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2627 		return;
2628 	}
2629 
2630 	in6_dev = in6_dev_get(dev);
2631 
2632 	if (!in6_dev) {
2633 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2634 				    dev->name);
2635 		return;
2636 	}
2637 
2638 	/*
2639 	 *	Two things going on here:
2640 	 *	1) Add routes for on-link prefixes
2641 	 *	2) Configure prefixes with the auto flag set
2642 	 */
2643 
2644 	if (pinfo->onlink) {
2645 		struct rt6_info *rt;
2646 		unsigned long rt_expires;
2647 
2648 		/* Avoid arithmetic overflow. Really, we could
2649 		 * save rt_expires in seconds, likely valid_lft,
2650 		 * but it would require division in fib gc, that it
2651 		 * not good.
2652 		 */
2653 		if (HZ > USER_HZ)
2654 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2655 		else
2656 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2657 
2658 		if (addrconf_finite_timeout(rt_expires))
2659 			rt_expires *= HZ;
2660 
2661 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2662 					       pinfo->prefix_len,
2663 					       dev,
2664 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2665 					       RTF_GATEWAY | RTF_DEFAULT);
2666 
2667 		if (rt) {
2668 			/* Autoconf prefix route */
2669 			if (valid_lft == 0) {
2670 				ip6_del_rt(rt);
2671 				rt = NULL;
2672 			} else if (addrconf_finite_timeout(rt_expires)) {
2673 				/* not infinity */
2674 				rt6_set_expires(rt, jiffies + rt_expires);
2675 			} else {
2676 				rt6_clean_expires(rt);
2677 			}
2678 		} else if (valid_lft) {
2679 			clock_t expires = 0;
2680 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2681 			if (addrconf_finite_timeout(rt_expires)) {
2682 				/* not infinity */
2683 				flags |= RTF_EXPIRES;
2684 				expires = jiffies_to_clock_t(rt_expires);
2685 			}
2686 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2687 					      dev, expires, flags);
2688 		}
2689 		ip6_rt_put(rt);
2690 	}
2691 
2692 	/* Try to figure out our local address for this prefix */
2693 
2694 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2695 		struct in6_addr addr;
2696 		bool tokenized = false, dev_addr_generated = false;
2697 
2698 		if (pinfo->prefix_len == 64) {
2699 			memcpy(&addr, &pinfo->prefix, 8);
2700 
2701 			if (!ipv6_addr_any(&in6_dev->token)) {
2702 				read_lock_bh(&in6_dev->lock);
2703 				memcpy(addr.s6_addr + 8,
2704 				       in6_dev->token.s6_addr + 8, 8);
2705 				read_unlock_bh(&in6_dev->lock);
2706 				tokenized = true;
2707 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2708 				   !ipv6_generate_stable_address(&addr, 0,
2709 								 in6_dev)) {
2710 				addr_flags |= IFA_F_STABLE_PRIVACY;
2711 				goto ok;
2712 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2713 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2714 				goto put;
2715 			} else {
2716 				dev_addr_generated = true;
2717 			}
2718 			goto ok;
2719 		}
2720 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2721 				    pinfo->prefix_len);
2722 		goto put;
2723 
2724 ok:
2725 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2726 						   &addr, addr_type,
2727 						   addr_flags, sllao,
2728 						   tokenized, valid_lft,
2729 						   prefered_lft);
2730 		if (err)
2731 			goto put;
2732 
2733 		/* Ignore error case here because previous prefix add addr was
2734 		 * successful which will be notified.
2735 		 */
2736 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2737 					      addr_type, addr_flags, sllao,
2738 					      tokenized, valid_lft,
2739 					      prefered_lft,
2740 					      dev_addr_generated);
2741 	}
2742 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2743 put:
2744 	in6_dev_put(in6_dev);
2745 }
2746 
2747 /*
2748  *	Set destination address.
2749  *	Special case for SIT interfaces where we create a new "virtual"
2750  *	device.
2751  */
2752 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2753 {
2754 	struct in6_ifreq ireq;
2755 	struct net_device *dev;
2756 	int err = -EINVAL;
2757 
2758 	rtnl_lock();
2759 
2760 	err = -EFAULT;
2761 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2762 		goto err_exit;
2763 
2764 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2765 
2766 	err = -ENODEV;
2767 	if (!dev)
2768 		goto err_exit;
2769 
2770 #if IS_ENABLED(CONFIG_IPV6_SIT)
2771 	if (dev->type == ARPHRD_SIT) {
2772 		const struct net_device_ops *ops = dev->netdev_ops;
2773 		struct ifreq ifr;
2774 		struct ip_tunnel_parm p;
2775 
2776 		err = -EADDRNOTAVAIL;
2777 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2778 			goto err_exit;
2779 
2780 		memset(&p, 0, sizeof(p));
2781 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2782 		p.iph.saddr = 0;
2783 		p.iph.version = 4;
2784 		p.iph.ihl = 5;
2785 		p.iph.protocol = IPPROTO_IPV6;
2786 		p.iph.ttl = 64;
2787 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2788 
2789 		if (ops->ndo_do_ioctl) {
2790 			mm_segment_t oldfs = get_fs();
2791 
2792 			set_fs(KERNEL_DS);
2793 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2794 			set_fs(oldfs);
2795 		} else
2796 			err = -EOPNOTSUPP;
2797 
2798 		if (err == 0) {
2799 			err = -ENOBUFS;
2800 			dev = __dev_get_by_name(net, p.name);
2801 			if (!dev)
2802 				goto err_exit;
2803 			err = dev_open(dev);
2804 		}
2805 	}
2806 #endif
2807 
2808 err_exit:
2809 	rtnl_unlock();
2810 	return err;
2811 }
2812 
2813 static int ipv6_mc_config(struct sock *sk, bool join,
2814 			  const struct in6_addr *addr, int ifindex)
2815 {
2816 	int ret;
2817 
2818 	ASSERT_RTNL();
2819 
2820 	lock_sock(sk);
2821 	if (join)
2822 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2823 	else
2824 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2825 	release_sock(sk);
2826 
2827 	return ret;
2828 }
2829 
2830 /*
2831  *	Manual configuration of address on an interface
2832  */
2833 static int inet6_addr_add(struct net *net, int ifindex,
2834 			  const struct in6_addr *pfx,
2835 			  const struct in6_addr *peer_pfx,
2836 			  unsigned int plen, __u32 ifa_flags,
2837 			  __u32 prefered_lft, __u32 valid_lft,
2838 			  struct netlink_ext_ack *extack)
2839 {
2840 	struct inet6_ifaddr *ifp;
2841 	struct inet6_dev *idev;
2842 	struct net_device *dev;
2843 	unsigned long timeout;
2844 	clock_t expires;
2845 	int scope;
2846 	u32 flags;
2847 
2848 	ASSERT_RTNL();
2849 
2850 	if (plen > 128)
2851 		return -EINVAL;
2852 
2853 	/* check the lifetime */
2854 	if (!valid_lft || prefered_lft > valid_lft)
2855 		return -EINVAL;
2856 
2857 	if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2858 		return -EINVAL;
2859 
2860 	dev = __dev_get_by_index(net, ifindex);
2861 	if (!dev)
2862 		return -ENODEV;
2863 
2864 	idev = addrconf_add_dev(dev);
2865 	if (IS_ERR(idev))
2866 		return PTR_ERR(idev);
2867 
2868 	if (ifa_flags & IFA_F_MCAUTOJOIN) {
2869 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2870 					 true, pfx, ifindex);
2871 
2872 		if (ret < 0)
2873 			return ret;
2874 	}
2875 
2876 	scope = ipv6_addr_scope(pfx);
2877 
2878 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2879 	if (addrconf_finite_timeout(timeout)) {
2880 		expires = jiffies_to_clock_t(timeout * HZ);
2881 		valid_lft = timeout;
2882 		flags = RTF_EXPIRES;
2883 	} else {
2884 		expires = 0;
2885 		flags = 0;
2886 		ifa_flags |= IFA_F_PERMANENT;
2887 	}
2888 
2889 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2890 	if (addrconf_finite_timeout(timeout)) {
2891 		if (timeout == 0)
2892 			ifa_flags |= IFA_F_DEPRECATED;
2893 		prefered_lft = timeout;
2894 	}
2895 
2896 	ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2897 			    valid_lft, prefered_lft, true, extack);
2898 
2899 	if (!IS_ERR(ifp)) {
2900 		if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2901 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2902 					      expires, flags);
2903 		}
2904 
2905 		/*
2906 		 * Note that section 3.1 of RFC 4429 indicates
2907 		 * that the Optimistic flag should not be set for
2908 		 * manually configured addresses
2909 		 */
2910 		addrconf_dad_start(ifp);
2911 		if (ifa_flags & IFA_F_MANAGETEMPADDR)
2912 			manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2913 					 true, jiffies);
2914 		in6_ifa_put(ifp);
2915 		addrconf_verify_rtnl();
2916 		return 0;
2917 	} else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2918 		ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2919 			       false, pfx, ifindex);
2920 	}
2921 
2922 	return PTR_ERR(ifp);
2923 }
2924 
2925 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2926 			  const struct in6_addr *pfx, unsigned int plen)
2927 {
2928 	struct inet6_ifaddr *ifp;
2929 	struct inet6_dev *idev;
2930 	struct net_device *dev;
2931 
2932 	if (plen > 128)
2933 		return -EINVAL;
2934 
2935 	dev = __dev_get_by_index(net, ifindex);
2936 	if (!dev)
2937 		return -ENODEV;
2938 
2939 	idev = __in6_dev_get(dev);
2940 	if (!idev)
2941 		return -ENXIO;
2942 
2943 	read_lock_bh(&idev->lock);
2944 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2945 		if (ifp->prefix_len == plen &&
2946 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2947 			in6_ifa_hold(ifp);
2948 			read_unlock_bh(&idev->lock);
2949 
2950 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
2951 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
2952 				manage_tempaddrs(idev, ifp, 0, 0, false,
2953 						 jiffies);
2954 			ipv6_del_addr(ifp);
2955 			addrconf_verify_rtnl();
2956 			if (ipv6_addr_is_multicast(pfx)) {
2957 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2958 					       false, pfx, dev->ifindex);
2959 			}
2960 			return 0;
2961 		}
2962 	}
2963 	read_unlock_bh(&idev->lock);
2964 	return -EADDRNOTAVAIL;
2965 }
2966 
2967 
2968 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2969 {
2970 	struct in6_ifreq ireq;
2971 	int err;
2972 
2973 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2974 		return -EPERM;
2975 
2976 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2977 		return -EFAULT;
2978 
2979 	rtnl_lock();
2980 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2981 			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2982 			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2983 	rtnl_unlock();
2984 	return err;
2985 }
2986 
2987 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2988 {
2989 	struct in6_ifreq ireq;
2990 	int err;
2991 
2992 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2993 		return -EPERM;
2994 
2995 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2996 		return -EFAULT;
2997 
2998 	rtnl_lock();
2999 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3000 			     ireq.ifr6_prefixlen);
3001 	rtnl_unlock();
3002 	return err;
3003 }
3004 
3005 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3006 		     int plen, int scope)
3007 {
3008 	struct inet6_ifaddr *ifp;
3009 
3010 	ifp = ipv6_add_addr(idev, addr, NULL, plen,
3011 			    scope, IFA_F_PERMANENT,
3012 			    INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
3013 			    true, NULL);
3014 	if (!IS_ERR(ifp)) {
3015 		spin_lock_bh(&ifp->lock);
3016 		ifp->flags &= ~IFA_F_TENTATIVE;
3017 		spin_unlock_bh(&ifp->lock);
3018 		rt_genid_bump_ipv6(dev_net(idev->dev));
3019 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3020 		in6_ifa_put(ifp);
3021 	}
3022 }
3023 
3024 #if IS_ENABLED(CONFIG_IPV6_SIT)
3025 static void sit_add_v4_addrs(struct inet6_dev *idev)
3026 {
3027 	struct in6_addr addr;
3028 	struct net_device *dev;
3029 	struct net *net = dev_net(idev->dev);
3030 	int scope, plen;
3031 	u32 pflags = 0;
3032 
3033 	ASSERT_RTNL();
3034 
3035 	memset(&addr, 0, sizeof(struct in6_addr));
3036 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3037 
3038 	if (idev->dev->flags&IFF_POINTOPOINT) {
3039 		addr.s6_addr32[0] = htonl(0xfe800000);
3040 		scope = IFA_LINK;
3041 		plen = 64;
3042 	} else {
3043 		scope = IPV6_ADDR_COMPATv4;
3044 		plen = 96;
3045 		pflags |= RTF_NONEXTHOP;
3046 	}
3047 
3048 	if (addr.s6_addr32[3]) {
3049 		add_addr(idev, &addr, plen, scope);
3050 		addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3051 		return;
3052 	}
3053 
3054 	for_each_netdev(net, dev) {
3055 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3056 		if (in_dev && (dev->flags & IFF_UP)) {
3057 			struct in_ifaddr *ifa;
3058 
3059 			int flag = scope;
3060 
3061 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3062 
3063 				addr.s6_addr32[3] = ifa->ifa_local;
3064 
3065 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3066 					continue;
3067 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3068 					if (idev->dev->flags&IFF_POINTOPOINT)
3069 						continue;
3070 					flag |= IFA_HOST;
3071 				}
3072 
3073 				add_addr(idev, &addr, plen, flag);
3074 				addrconf_prefix_route(&addr, plen, idev->dev, 0,
3075 						      pflags);
3076 			}
3077 		}
3078 	}
3079 }
3080 #endif
3081 
3082 static void init_loopback(struct net_device *dev)
3083 {
3084 	struct inet6_dev  *idev;
3085 
3086 	/* ::1 */
3087 
3088 	ASSERT_RTNL();
3089 
3090 	idev = ipv6_find_idev(dev);
3091 	if (!idev) {
3092 		pr_debug("%s: add_dev failed\n", __func__);
3093 		return;
3094 	}
3095 
3096 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3097 }
3098 
3099 void addrconf_add_linklocal(struct inet6_dev *idev,
3100 			    const struct in6_addr *addr, u32 flags)
3101 {
3102 	struct inet6_ifaddr *ifp;
3103 	u32 addr_flags = flags | IFA_F_PERMANENT;
3104 
3105 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3106 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3107 	     idev->cnf.optimistic_dad) &&
3108 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3109 		addr_flags |= IFA_F_OPTIMISTIC;
3110 #endif
3111 
3112 	ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3113 			    INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3114 	if (!IS_ERR(ifp)) {
3115 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3116 		addrconf_dad_start(ifp);
3117 		in6_ifa_put(ifp);
3118 	}
3119 }
3120 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3121 
3122 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3123 {
3124 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3125 		return true;
3126 
3127 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3128 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3129 		return true;
3130 
3131 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3132 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3133 		return true;
3134 
3135 	return false;
3136 }
3137 
3138 static int ipv6_generate_stable_address(struct in6_addr *address,
3139 					u8 dad_count,
3140 					const struct inet6_dev *idev)
3141 {
3142 	static DEFINE_SPINLOCK(lock);
3143 	static __u32 digest[SHA_DIGEST_WORDS];
3144 	static __u32 workspace[SHA_WORKSPACE_WORDS];
3145 
3146 	static union {
3147 		char __data[SHA_MESSAGE_BYTES];
3148 		struct {
3149 			struct in6_addr secret;
3150 			__be32 prefix[2];
3151 			unsigned char hwaddr[MAX_ADDR_LEN];
3152 			u8 dad_count;
3153 		} __packed;
3154 	} data;
3155 
3156 	struct in6_addr secret;
3157 	struct in6_addr temp;
3158 	struct net *net = dev_net(idev->dev);
3159 
3160 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3161 
3162 	if (idev->cnf.stable_secret.initialized)
3163 		secret = idev->cnf.stable_secret.secret;
3164 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3165 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3166 	else
3167 		return -1;
3168 
3169 retry:
3170 	spin_lock_bh(&lock);
3171 
3172 	sha_init(digest);
3173 	memset(&data, 0, sizeof(data));
3174 	memset(workspace, 0, sizeof(workspace));
3175 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3176 	data.prefix[0] = address->s6_addr32[0];
3177 	data.prefix[1] = address->s6_addr32[1];
3178 	data.secret = secret;
3179 	data.dad_count = dad_count;
3180 
3181 	sha_transform(digest, data.__data, workspace);
3182 
3183 	temp = *address;
3184 	temp.s6_addr32[2] = (__force __be32)digest[0];
3185 	temp.s6_addr32[3] = (__force __be32)digest[1];
3186 
3187 	spin_unlock_bh(&lock);
3188 
3189 	if (ipv6_reserved_interfaceid(temp)) {
3190 		dad_count++;
3191 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3192 			return -1;
3193 		goto retry;
3194 	}
3195 
3196 	*address = temp;
3197 	return 0;
3198 }
3199 
3200 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3201 {
3202 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3203 
3204 	if (s->initialized)
3205 		return;
3206 	s = &idev->cnf.stable_secret;
3207 	get_random_bytes(&s->secret, sizeof(s->secret));
3208 	s->initialized = true;
3209 }
3210 
3211 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3212 {
3213 	struct in6_addr addr;
3214 
3215 	/* no link local addresses on L3 master devices */
3216 	if (netif_is_l3_master(idev->dev))
3217 		return;
3218 
3219 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3220 
3221 	switch (idev->cnf.addr_gen_mode) {
3222 	case IN6_ADDR_GEN_MODE_RANDOM:
3223 		ipv6_gen_mode_random_init(idev);
3224 		/* fallthrough */
3225 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3226 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3227 			addrconf_add_linklocal(idev, &addr,
3228 					       IFA_F_STABLE_PRIVACY);
3229 		else if (prefix_route)
3230 			addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3231 		break;
3232 	case IN6_ADDR_GEN_MODE_EUI64:
3233 		/* addrconf_add_linklocal also adds a prefix_route and we
3234 		 * only need to care about prefix routes if ipv6_generate_eui64
3235 		 * couldn't generate one.
3236 		 */
3237 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3238 			addrconf_add_linklocal(idev, &addr, 0);
3239 		else if (prefix_route)
3240 			addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3241 		break;
3242 	case IN6_ADDR_GEN_MODE_NONE:
3243 	default:
3244 		/* will not add any link local address */
3245 		break;
3246 	}
3247 }
3248 
3249 static void addrconf_dev_config(struct net_device *dev)
3250 {
3251 	struct inet6_dev *idev;
3252 
3253 	ASSERT_RTNL();
3254 
3255 	if ((dev->type != ARPHRD_ETHER) &&
3256 	    (dev->type != ARPHRD_FDDI) &&
3257 	    (dev->type != ARPHRD_ARCNET) &&
3258 	    (dev->type != ARPHRD_INFINIBAND) &&
3259 	    (dev->type != ARPHRD_IEEE1394) &&
3260 	    (dev->type != ARPHRD_TUNNEL6) &&
3261 	    (dev->type != ARPHRD_6LOWPAN) &&
3262 	    (dev->type != ARPHRD_IP6GRE) &&
3263 	    (dev->type != ARPHRD_IPGRE) &&
3264 	    (dev->type != ARPHRD_TUNNEL) &&
3265 	    (dev->type != ARPHRD_NONE)) {
3266 		/* Alas, we support only Ethernet autoconfiguration. */
3267 		return;
3268 	}
3269 
3270 	idev = addrconf_add_dev(dev);
3271 	if (IS_ERR(idev))
3272 		return;
3273 
3274 	/* this device type has no EUI support */
3275 	if (dev->type == ARPHRD_NONE &&
3276 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3277 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3278 
3279 	addrconf_addr_gen(idev, false);
3280 }
3281 
3282 #if IS_ENABLED(CONFIG_IPV6_SIT)
3283 static void addrconf_sit_config(struct net_device *dev)
3284 {
3285 	struct inet6_dev *idev;
3286 
3287 	ASSERT_RTNL();
3288 
3289 	/*
3290 	 * Configure the tunnel with one of our IPv4
3291 	 * addresses... we should configure all of
3292 	 * our v4 addrs in the tunnel
3293 	 */
3294 
3295 	idev = ipv6_find_idev(dev);
3296 	if (!idev) {
3297 		pr_debug("%s: add_dev failed\n", __func__);
3298 		return;
3299 	}
3300 
3301 	if (dev->priv_flags & IFF_ISATAP) {
3302 		addrconf_addr_gen(idev, false);
3303 		return;
3304 	}
3305 
3306 	sit_add_v4_addrs(idev);
3307 
3308 	if (dev->flags&IFF_POINTOPOINT)
3309 		addrconf_add_mroute(dev);
3310 }
3311 #endif
3312 
3313 #if IS_ENABLED(CONFIG_NET_IPGRE)
3314 static void addrconf_gre_config(struct net_device *dev)
3315 {
3316 	struct inet6_dev *idev;
3317 
3318 	ASSERT_RTNL();
3319 
3320 	idev = ipv6_find_idev(dev);
3321 	if (!idev) {
3322 		pr_debug("%s: add_dev failed\n", __func__);
3323 		return;
3324 	}
3325 
3326 	addrconf_addr_gen(idev, true);
3327 	if (dev->flags & IFF_POINTOPOINT)
3328 		addrconf_add_mroute(dev);
3329 }
3330 #endif
3331 
3332 static int fixup_permanent_addr(struct inet6_dev *idev,
3333 				struct inet6_ifaddr *ifp)
3334 {
3335 	/* !rt6i_node means the host route was removed from the
3336 	 * FIB, for example, if 'lo' device is taken down. In that
3337 	 * case regenerate the host route.
3338 	 */
3339 	if (!ifp->rt || !ifp->rt->rt6i_node) {
3340 		struct rt6_info *rt, *prev;
3341 
3342 		rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3343 		if (IS_ERR(rt))
3344 			return PTR_ERR(rt);
3345 
3346 		/* ifp->rt can be accessed outside of rtnl */
3347 		spin_lock(&ifp->lock);
3348 		prev = ifp->rt;
3349 		ifp->rt = rt;
3350 		spin_unlock(&ifp->lock);
3351 
3352 		ip6_rt_put(prev);
3353 	}
3354 
3355 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3356 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3357 				      idev->dev, 0, 0);
3358 	}
3359 
3360 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3361 		addrconf_dad_start(ifp);
3362 
3363 	return 0;
3364 }
3365 
3366 static void addrconf_permanent_addr(struct net_device *dev)
3367 {
3368 	struct inet6_ifaddr *ifp, *tmp;
3369 	struct inet6_dev *idev;
3370 
3371 	idev = __in6_dev_get(dev);
3372 	if (!idev)
3373 		return;
3374 
3375 	write_lock_bh(&idev->lock);
3376 
3377 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3378 		if ((ifp->flags & IFA_F_PERMANENT) &&
3379 		    fixup_permanent_addr(idev, ifp) < 0) {
3380 			write_unlock_bh(&idev->lock);
3381 			in6_ifa_hold(ifp);
3382 			ipv6_del_addr(ifp);
3383 			write_lock_bh(&idev->lock);
3384 
3385 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3386 					     idev->dev->name, &ifp->addr);
3387 		}
3388 	}
3389 
3390 	write_unlock_bh(&idev->lock);
3391 }
3392 
3393 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3394 			   void *ptr)
3395 {
3396 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3397 	struct netdev_notifier_changeupper_info *info;
3398 	struct inet6_dev *idev = __in6_dev_get(dev);
3399 	struct net *net = dev_net(dev);
3400 	int run_pending = 0;
3401 	int err;
3402 
3403 	switch (event) {
3404 	case NETDEV_REGISTER:
3405 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3406 			idev = ipv6_add_dev(dev);
3407 			if (IS_ERR(idev))
3408 				return notifier_from_errno(PTR_ERR(idev));
3409 		}
3410 		break;
3411 
3412 	case NETDEV_CHANGEMTU:
3413 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3414 		if (dev->mtu < IPV6_MIN_MTU) {
3415 			addrconf_ifdown(dev, dev != net->loopback_dev);
3416 			break;
3417 		}
3418 
3419 		if (idev) {
3420 			rt6_mtu_change(dev, dev->mtu);
3421 			idev->cnf.mtu6 = dev->mtu;
3422 			break;
3423 		}
3424 
3425 		/* allocate new idev */
3426 		idev = ipv6_add_dev(dev);
3427 		if (IS_ERR(idev))
3428 			break;
3429 
3430 		/* device is still not ready */
3431 		if (!(idev->if_flags & IF_READY))
3432 			break;
3433 
3434 		run_pending = 1;
3435 
3436 		/* fall through */
3437 
3438 	case NETDEV_UP:
3439 	case NETDEV_CHANGE:
3440 		if (dev->flags & IFF_SLAVE)
3441 			break;
3442 
3443 		if (idev && idev->cnf.disable_ipv6)
3444 			break;
3445 
3446 		if (event == NETDEV_UP) {
3447 			/* restore routes for permanent addresses */
3448 			addrconf_permanent_addr(dev);
3449 
3450 			if (!addrconf_link_ready(dev)) {
3451 				/* device is not ready yet. */
3452 				pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3453 					dev->name);
3454 				break;
3455 			}
3456 
3457 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3458 				idev = ipv6_add_dev(dev);
3459 
3460 			if (!IS_ERR_OR_NULL(idev)) {
3461 				idev->if_flags |= IF_READY;
3462 				run_pending = 1;
3463 			}
3464 		} else if (event == NETDEV_CHANGE) {
3465 			if (!addrconf_link_ready(dev)) {
3466 				/* device is still not ready. */
3467 				rt6_sync_down_dev(dev, event);
3468 				break;
3469 			}
3470 
3471 			if (idev) {
3472 				if (idev->if_flags & IF_READY) {
3473 					/* device is already configured -
3474 					 * but resend MLD reports, we might
3475 					 * have roamed and need to update
3476 					 * multicast snooping switches
3477 					 */
3478 					ipv6_mc_up(idev);
3479 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3480 					break;
3481 				}
3482 				idev->if_flags |= IF_READY;
3483 			}
3484 
3485 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3486 				dev->name);
3487 
3488 			run_pending = 1;
3489 		}
3490 
3491 		switch (dev->type) {
3492 #if IS_ENABLED(CONFIG_IPV6_SIT)
3493 		case ARPHRD_SIT:
3494 			addrconf_sit_config(dev);
3495 			break;
3496 #endif
3497 #if IS_ENABLED(CONFIG_NET_IPGRE)
3498 		case ARPHRD_IPGRE:
3499 			addrconf_gre_config(dev);
3500 			break;
3501 #endif
3502 		case ARPHRD_LOOPBACK:
3503 			init_loopback(dev);
3504 			break;
3505 
3506 		default:
3507 			addrconf_dev_config(dev);
3508 			break;
3509 		}
3510 
3511 		if (!IS_ERR_OR_NULL(idev)) {
3512 			if (run_pending)
3513 				addrconf_dad_run(idev);
3514 
3515 			/* Device has an address by now */
3516 			rt6_sync_up(dev, RTNH_F_DEAD);
3517 
3518 			/*
3519 			 * If the MTU changed during the interface down,
3520 			 * when the interface up, the changed MTU must be
3521 			 * reflected in the idev as well as routers.
3522 			 */
3523 			if (idev->cnf.mtu6 != dev->mtu &&
3524 			    dev->mtu >= IPV6_MIN_MTU) {
3525 				rt6_mtu_change(dev, dev->mtu);
3526 				idev->cnf.mtu6 = dev->mtu;
3527 			}
3528 			idev->tstamp = jiffies;
3529 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3530 
3531 			/*
3532 			 * If the changed mtu during down is lower than
3533 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3534 			 */
3535 			if (dev->mtu < IPV6_MIN_MTU)
3536 				addrconf_ifdown(dev, dev != net->loopback_dev);
3537 		}
3538 		break;
3539 
3540 	case NETDEV_DOWN:
3541 	case NETDEV_UNREGISTER:
3542 		/*
3543 		 *	Remove all addresses from this interface.
3544 		 */
3545 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3546 		break;
3547 
3548 	case NETDEV_CHANGENAME:
3549 		if (idev) {
3550 			snmp6_unregister_dev(idev);
3551 			addrconf_sysctl_unregister(idev);
3552 			err = addrconf_sysctl_register(idev);
3553 			if (err)
3554 				return notifier_from_errno(err);
3555 			err = snmp6_register_dev(idev);
3556 			if (err) {
3557 				addrconf_sysctl_unregister(idev);
3558 				return notifier_from_errno(err);
3559 			}
3560 		}
3561 		break;
3562 
3563 	case NETDEV_PRE_TYPE_CHANGE:
3564 	case NETDEV_POST_TYPE_CHANGE:
3565 		if (idev)
3566 			addrconf_type_change(dev, event);
3567 		break;
3568 
3569 	case NETDEV_CHANGEUPPER:
3570 		info = ptr;
3571 
3572 		/* flush all routes if dev is linked to or unlinked from
3573 		 * an L3 master device (e.g., VRF)
3574 		 */
3575 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3576 			addrconf_ifdown(dev, 0);
3577 	}
3578 
3579 	return NOTIFY_OK;
3580 }
3581 
3582 /*
3583  *	addrconf module should be notified of a device going up
3584  */
3585 static struct notifier_block ipv6_dev_notf = {
3586 	.notifier_call = addrconf_notify,
3587 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3588 };
3589 
3590 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3591 {
3592 	struct inet6_dev *idev;
3593 	ASSERT_RTNL();
3594 
3595 	idev = __in6_dev_get(dev);
3596 
3597 	if (event == NETDEV_POST_TYPE_CHANGE)
3598 		ipv6_mc_remap(idev);
3599 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3600 		ipv6_mc_unmap(idev);
3601 }
3602 
3603 static bool addr_is_local(const struct in6_addr *addr)
3604 {
3605 	return ipv6_addr_type(addr) &
3606 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3607 }
3608 
3609 static int addrconf_ifdown(struct net_device *dev, int how)
3610 {
3611 	unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3612 	struct net *net = dev_net(dev);
3613 	struct inet6_dev *idev;
3614 	struct inet6_ifaddr *ifa, *tmp;
3615 	int _keep_addr;
3616 	bool keep_addr;
3617 	int state, i;
3618 
3619 	ASSERT_RTNL();
3620 
3621 	rt6_disable_ip(dev, event);
3622 
3623 	idev = __in6_dev_get(dev);
3624 	if (!idev)
3625 		return -ENODEV;
3626 
3627 	/*
3628 	 * Step 1: remove reference to ipv6 device from parent device.
3629 	 *	   Do not dev_put!
3630 	 */
3631 	if (how) {
3632 		idev->dead = 1;
3633 
3634 		/* protected by rtnl_lock */
3635 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3636 
3637 		/* Step 1.5: remove snmp6 entry */
3638 		snmp6_unregister_dev(idev);
3639 
3640 	}
3641 
3642 	/* aggregate the system setting and interface setting */
3643 	_keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3644 	if (!_keep_addr)
3645 		_keep_addr = idev->cnf.keep_addr_on_down;
3646 
3647 	/* combine the user config with event to determine if permanent
3648 	 * addresses are to be removed from address hash table
3649 	 */
3650 	keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3651 
3652 	/* Step 2: clear hash table */
3653 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3654 		struct hlist_head *h = &inet6_addr_lst[i];
3655 
3656 		spin_lock_bh(&addrconf_hash_lock);
3657 restart:
3658 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3659 			if (ifa->idev == idev) {
3660 				addrconf_del_dad_work(ifa);
3661 				/* combined flag + permanent flag decide if
3662 				 * address is retained on a down event
3663 				 */
3664 				if (!keep_addr ||
3665 				    !(ifa->flags & IFA_F_PERMANENT) ||
3666 				    addr_is_local(&ifa->addr)) {
3667 					hlist_del_init_rcu(&ifa->addr_lst);
3668 					goto restart;
3669 				}
3670 			}
3671 		}
3672 		spin_unlock_bh(&addrconf_hash_lock);
3673 	}
3674 
3675 	write_lock_bh(&idev->lock);
3676 
3677 	addrconf_del_rs_timer(idev);
3678 
3679 	/* Step 2: clear flags for stateless addrconf */
3680 	if (!how)
3681 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3682 
3683 	/* Step 3: clear tempaddr list */
3684 	while (!list_empty(&idev->tempaddr_list)) {
3685 		ifa = list_first_entry(&idev->tempaddr_list,
3686 				       struct inet6_ifaddr, tmp_list);
3687 		list_del(&ifa->tmp_list);
3688 		write_unlock_bh(&idev->lock);
3689 		spin_lock_bh(&ifa->lock);
3690 
3691 		if (ifa->ifpub) {
3692 			in6_ifa_put(ifa->ifpub);
3693 			ifa->ifpub = NULL;
3694 		}
3695 		spin_unlock_bh(&ifa->lock);
3696 		in6_ifa_put(ifa);
3697 		write_lock_bh(&idev->lock);
3698 	}
3699 
3700 	/* re-combine the user config with event to determine if permanent
3701 	 * addresses are to be removed from the interface list
3702 	 */
3703 	keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3704 
3705 	list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3706 		struct rt6_info *rt = NULL;
3707 		bool keep;
3708 
3709 		addrconf_del_dad_work(ifa);
3710 
3711 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3712 			!addr_is_local(&ifa->addr);
3713 
3714 		write_unlock_bh(&idev->lock);
3715 		spin_lock_bh(&ifa->lock);
3716 
3717 		if (keep) {
3718 			/* set state to skip the notifier below */
3719 			state = INET6_IFADDR_STATE_DEAD;
3720 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3721 			if (!(ifa->flags & IFA_F_NODAD))
3722 				ifa->flags |= IFA_F_TENTATIVE;
3723 
3724 			rt = ifa->rt;
3725 			ifa->rt = NULL;
3726 		} else {
3727 			state = ifa->state;
3728 			ifa->state = INET6_IFADDR_STATE_DEAD;
3729 		}
3730 
3731 		spin_unlock_bh(&ifa->lock);
3732 
3733 		if (rt)
3734 			ip6_del_rt(rt);
3735 
3736 		if (state != INET6_IFADDR_STATE_DEAD) {
3737 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3738 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3739 		} else {
3740 			if (idev->cnf.forwarding)
3741 				addrconf_leave_anycast(ifa);
3742 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3743 		}
3744 
3745 		write_lock_bh(&idev->lock);
3746 		if (!keep) {
3747 			list_del_rcu(&ifa->if_list);
3748 			in6_ifa_put(ifa);
3749 		}
3750 	}
3751 
3752 	write_unlock_bh(&idev->lock);
3753 
3754 	/* Step 5: Discard anycast and multicast list */
3755 	if (how) {
3756 		ipv6_ac_destroy_dev(idev);
3757 		ipv6_mc_destroy_dev(idev);
3758 	} else {
3759 		ipv6_mc_down(idev);
3760 	}
3761 
3762 	idev->tstamp = jiffies;
3763 
3764 	/* Last: Shot the device (if unregistered) */
3765 	if (how) {
3766 		addrconf_sysctl_unregister(idev);
3767 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3768 		neigh_ifdown(&nd_tbl, dev);
3769 		in6_dev_put(idev);
3770 	}
3771 	return 0;
3772 }
3773 
3774 static void addrconf_rs_timer(struct timer_list *t)
3775 {
3776 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3777 	struct net_device *dev = idev->dev;
3778 	struct in6_addr lladdr;
3779 
3780 	write_lock(&idev->lock);
3781 	if (idev->dead || !(idev->if_flags & IF_READY))
3782 		goto out;
3783 
3784 	if (!ipv6_accept_ra(idev))
3785 		goto out;
3786 
3787 	/* Announcement received after solicitation was sent */
3788 	if (idev->if_flags & IF_RA_RCVD)
3789 		goto out;
3790 
3791 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3792 		write_unlock(&idev->lock);
3793 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3794 			ndisc_send_rs(dev, &lladdr,
3795 				      &in6addr_linklocal_allrouters);
3796 		else
3797 			goto put;
3798 
3799 		write_lock(&idev->lock);
3800 		idev->rs_interval = rfc3315_s14_backoff_update(
3801 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3802 		/* The wait after the last probe can be shorter */
3803 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3804 					     idev->cnf.rtr_solicits) ?
3805 				      idev->cnf.rtr_solicit_delay :
3806 				      idev->rs_interval);
3807 	} else {
3808 		/*
3809 		 * Note: we do not support deprecated "all on-link"
3810 		 * assumption any longer.
3811 		 */
3812 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3813 	}
3814 
3815 out:
3816 	write_unlock(&idev->lock);
3817 put:
3818 	in6_dev_put(idev);
3819 }
3820 
3821 /*
3822  *	Duplicate Address Detection
3823  */
3824 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3825 {
3826 	unsigned long rand_num;
3827 	struct inet6_dev *idev = ifp->idev;
3828 	u64 nonce;
3829 
3830 	if (ifp->flags & IFA_F_OPTIMISTIC)
3831 		rand_num = 0;
3832 	else
3833 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3834 
3835 	nonce = 0;
3836 	if (idev->cnf.enhanced_dad ||
3837 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3838 		do
3839 			get_random_bytes(&nonce, 6);
3840 		while (nonce == 0);
3841 	}
3842 	ifp->dad_nonce = nonce;
3843 	ifp->dad_probes = idev->cnf.dad_transmits;
3844 	addrconf_mod_dad_work(ifp, rand_num);
3845 }
3846 
3847 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3848 {
3849 	struct inet6_dev *idev = ifp->idev;
3850 	struct net_device *dev = idev->dev;
3851 	bool bump_id, notify = false;
3852 
3853 	addrconf_join_solict(dev, &ifp->addr);
3854 
3855 	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3856 
3857 	read_lock_bh(&idev->lock);
3858 	spin_lock(&ifp->lock);
3859 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3860 		goto out;
3861 
3862 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3863 	    (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3864 	     idev->cnf.accept_dad < 1) ||
3865 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3866 	    ifp->flags & IFA_F_NODAD) {
3867 		bool send_na = false;
3868 
3869 		if (ifp->flags & IFA_F_TENTATIVE &&
3870 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3871 			send_na = true;
3872 		bump_id = ifp->flags & IFA_F_TENTATIVE;
3873 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3874 		spin_unlock(&ifp->lock);
3875 		read_unlock_bh(&idev->lock);
3876 
3877 		addrconf_dad_completed(ifp, bump_id, send_na);
3878 		return;
3879 	}
3880 
3881 	if (!(idev->if_flags & IF_READY)) {
3882 		spin_unlock(&ifp->lock);
3883 		read_unlock_bh(&idev->lock);
3884 		/*
3885 		 * If the device is not ready:
3886 		 * - keep it tentative if it is a permanent address.
3887 		 * - otherwise, kill it.
3888 		 */
3889 		in6_ifa_hold(ifp);
3890 		addrconf_dad_stop(ifp, 0);
3891 		return;
3892 	}
3893 
3894 	/*
3895 	 * Optimistic nodes can start receiving
3896 	 * Frames right away
3897 	 */
3898 	if (ifp->flags & IFA_F_OPTIMISTIC) {
3899 		ip6_ins_rt(ifp->rt);
3900 		if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3901 			/* Because optimistic nodes can use this address,
3902 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3903 			 */
3904 			notify = true;
3905 		}
3906 	}
3907 
3908 	addrconf_dad_kick(ifp);
3909 out:
3910 	spin_unlock(&ifp->lock);
3911 	read_unlock_bh(&idev->lock);
3912 	if (notify)
3913 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3914 }
3915 
3916 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3917 {
3918 	bool begin_dad = false;
3919 
3920 	spin_lock_bh(&ifp->lock);
3921 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3922 		ifp->state = INET6_IFADDR_STATE_PREDAD;
3923 		begin_dad = true;
3924 	}
3925 	spin_unlock_bh(&ifp->lock);
3926 
3927 	if (begin_dad)
3928 		addrconf_mod_dad_work(ifp, 0);
3929 }
3930 
3931 static void addrconf_dad_work(struct work_struct *w)
3932 {
3933 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3934 						struct inet6_ifaddr,
3935 						dad_work);
3936 	struct inet6_dev *idev = ifp->idev;
3937 	bool bump_id, disable_ipv6 = false;
3938 	struct in6_addr mcaddr;
3939 
3940 	enum {
3941 		DAD_PROCESS,
3942 		DAD_BEGIN,
3943 		DAD_ABORT,
3944 	} action = DAD_PROCESS;
3945 
3946 	rtnl_lock();
3947 
3948 	spin_lock_bh(&ifp->lock);
3949 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3950 		action = DAD_BEGIN;
3951 		ifp->state = INET6_IFADDR_STATE_DAD;
3952 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3953 		action = DAD_ABORT;
3954 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
3955 
3956 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3957 		     idev->cnf.accept_dad > 1) &&
3958 		    !idev->cnf.disable_ipv6 &&
3959 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3960 			struct in6_addr addr;
3961 
3962 			addr.s6_addr32[0] = htonl(0xfe800000);
3963 			addr.s6_addr32[1] = 0;
3964 
3965 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3966 			    ipv6_addr_equal(&ifp->addr, &addr)) {
3967 				/* DAD failed for link-local based on MAC */
3968 				idev->cnf.disable_ipv6 = 1;
3969 
3970 				pr_info("%s: IPv6 being disabled!\n",
3971 					ifp->idev->dev->name);
3972 				disable_ipv6 = true;
3973 			}
3974 		}
3975 	}
3976 	spin_unlock_bh(&ifp->lock);
3977 
3978 	if (action == DAD_BEGIN) {
3979 		addrconf_dad_begin(ifp);
3980 		goto out;
3981 	} else if (action == DAD_ABORT) {
3982 		in6_ifa_hold(ifp);
3983 		addrconf_dad_stop(ifp, 1);
3984 		if (disable_ipv6)
3985 			addrconf_ifdown(idev->dev, 0);
3986 		goto out;
3987 	}
3988 
3989 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
3990 		goto out;
3991 
3992 	write_lock_bh(&idev->lock);
3993 	if (idev->dead || !(idev->if_flags & IF_READY)) {
3994 		write_unlock_bh(&idev->lock);
3995 		goto out;
3996 	}
3997 
3998 	spin_lock(&ifp->lock);
3999 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4000 		spin_unlock(&ifp->lock);
4001 		write_unlock_bh(&idev->lock);
4002 		goto out;
4003 	}
4004 
4005 	if (ifp->dad_probes == 0) {
4006 		bool send_na = false;
4007 
4008 		/*
4009 		 * DAD was successful
4010 		 */
4011 
4012 		if (ifp->flags & IFA_F_TENTATIVE &&
4013 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4014 			send_na = true;
4015 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4016 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4017 		spin_unlock(&ifp->lock);
4018 		write_unlock_bh(&idev->lock);
4019 
4020 		addrconf_dad_completed(ifp, bump_id, send_na);
4021 
4022 		goto out;
4023 	}
4024 
4025 	ifp->dad_probes--;
4026 	addrconf_mod_dad_work(ifp,
4027 			      NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4028 	spin_unlock(&ifp->lock);
4029 	write_unlock_bh(&idev->lock);
4030 
4031 	/* send a neighbour solicitation for our addr */
4032 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4033 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4034 		      ifp->dad_nonce);
4035 out:
4036 	in6_ifa_put(ifp);
4037 	rtnl_unlock();
4038 }
4039 
4040 /* ifp->idev must be at least read locked */
4041 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4042 {
4043 	struct inet6_ifaddr *ifpiter;
4044 	struct inet6_dev *idev = ifp->idev;
4045 
4046 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4047 		if (ifpiter->scope > IFA_LINK)
4048 			break;
4049 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4050 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4051 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4052 		    IFA_F_PERMANENT)
4053 			return false;
4054 	}
4055 	return true;
4056 }
4057 
4058 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4059 				   bool send_na)
4060 {
4061 	struct net_device *dev = ifp->idev->dev;
4062 	struct in6_addr lladdr;
4063 	bool send_rs, send_mld;
4064 
4065 	addrconf_del_dad_work(ifp);
4066 
4067 	/*
4068 	 *	Configure the address for reception. Now it is valid.
4069 	 */
4070 
4071 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4072 
4073 	/* If added prefix is link local and we are prepared to process
4074 	   router advertisements, start sending router solicitations.
4075 	 */
4076 
4077 	read_lock_bh(&ifp->idev->lock);
4078 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4079 	send_rs = send_mld &&
4080 		  ipv6_accept_ra(ifp->idev) &&
4081 		  ifp->idev->cnf.rtr_solicits != 0 &&
4082 		  (dev->flags&IFF_LOOPBACK) == 0;
4083 	read_unlock_bh(&ifp->idev->lock);
4084 
4085 	/* While dad is in progress mld report's source address is in6_addrany.
4086 	 * Resend with proper ll now.
4087 	 */
4088 	if (send_mld)
4089 		ipv6_mc_dad_complete(ifp->idev);
4090 
4091 	/* send unsolicited NA if enabled */
4092 	if (send_na &&
4093 	    (ifp->idev->cnf.ndisc_notify ||
4094 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4095 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4096 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4097 			      /*solicited=*/ false, /*override=*/ true,
4098 			      /*inc_opt=*/ true);
4099 	}
4100 
4101 	if (send_rs) {
4102 		/*
4103 		 *	If a host as already performed a random delay
4104 		 *	[...] as part of DAD [...] there is no need
4105 		 *	to delay again before sending the first RS
4106 		 */
4107 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4108 			return;
4109 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4110 
4111 		write_lock_bh(&ifp->idev->lock);
4112 		spin_lock(&ifp->lock);
4113 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4114 			ifp->idev->cnf.rtr_solicit_interval);
4115 		ifp->idev->rs_probes = 1;
4116 		ifp->idev->if_flags |= IF_RS_SENT;
4117 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4118 		spin_unlock(&ifp->lock);
4119 		write_unlock_bh(&ifp->idev->lock);
4120 	}
4121 
4122 	if (bump_id)
4123 		rt_genid_bump_ipv6(dev_net(dev));
4124 
4125 	/* Make sure that a new temporary address will be created
4126 	 * before this temporary address becomes deprecated.
4127 	 */
4128 	if (ifp->flags & IFA_F_TEMPORARY)
4129 		addrconf_verify_rtnl();
4130 }
4131 
4132 static void addrconf_dad_run(struct inet6_dev *idev)
4133 {
4134 	struct inet6_ifaddr *ifp;
4135 
4136 	read_lock_bh(&idev->lock);
4137 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4138 		spin_lock(&ifp->lock);
4139 		if (ifp->flags & IFA_F_TENTATIVE &&
4140 		    ifp->state == INET6_IFADDR_STATE_DAD)
4141 			addrconf_dad_kick(ifp);
4142 		spin_unlock(&ifp->lock);
4143 	}
4144 	read_unlock_bh(&idev->lock);
4145 }
4146 
4147 #ifdef CONFIG_PROC_FS
4148 struct if6_iter_state {
4149 	struct seq_net_private p;
4150 	int bucket;
4151 	int offset;
4152 };
4153 
4154 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4155 {
4156 	struct if6_iter_state *state = seq->private;
4157 	struct net *net = seq_file_net(seq);
4158 	struct inet6_ifaddr *ifa = NULL;
4159 	int p = 0;
4160 
4161 	/* initial bucket if pos is 0 */
4162 	if (pos == 0) {
4163 		state->bucket = 0;
4164 		state->offset = 0;
4165 	}
4166 
4167 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4168 		hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4169 					 addr_lst) {
4170 			if (!net_eq(dev_net(ifa->idev->dev), net))
4171 				continue;
4172 			/* sync with offset */
4173 			if (p < state->offset) {
4174 				p++;
4175 				continue;
4176 			}
4177 			state->offset++;
4178 			return ifa;
4179 		}
4180 
4181 		/* prepare for next bucket */
4182 		state->offset = 0;
4183 		p = 0;
4184 	}
4185 	return NULL;
4186 }
4187 
4188 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4189 					 struct inet6_ifaddr *ifa)
4190 {
4191 	struct if6_iter_state *state = seq->private;
4192 	struct net *net = seq_file_net(seq);
4193 
4194 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4195 		if (!net_eq(dev_net(ifa->idev->dev), net))
4196 			continue;
4197 		state->offset++;
4198 		return ifa;
4199 	}
4200 
4201 	while (++state->bucket < IN6_ADDR_HSIZE) {
4202 		state->offset = 0;
4203 		hlist_for_each_entry_rcu(ifa,
4204 				     &inet6_addr_lst[state->bucket], addr_lst) {
4205 			if (!net_eq(dev_net(ifa->idev->dev), net))
4206 				continue;
4207 			state->offset++;
4208 			return ifa;
4209 		}
4210 	}
4211 
4212 	return NULL;
4213 }
4214 
4215 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4216 	__acquires(rcu)
4217 {
4218 	rcu_read_lock();
4219 	return if6_get_first(seq, *pos);
4220 }
4221 
4222 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4223 {
4224 	struct inet6_ifaddr *ifa;
4225 
4226 	ifa = if6_get_next(seq, v);
4227 	++*pos;
4228 	return ifa;
4229 }
4230 
4231 static void if6_seq_stop(struct seq_file *seq, void *v)
4232 	__releases(rcu)
4233 {
4234 	rcu_read_unlock();
4235 }
4236 
4237 static int if6_seq_show(struct seq_file *seq, void *v)
4238 {
4239 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4240 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4241 		   &ifp->addr,
4242 		   ifp->idev->dev->ifindex,
4243 		   ifp->prefix_len,
4244 		   ifp->scope,
4245 		   (u8) ifp->flags,
4246 		   ifp->idev->dev->name);
4247 	return 0;
4248 }
4249 
4250 static const struct seq_operations if6_seq_ops = {
4251 	.start	= if6_seq_start,
4252 	.next	= if6_seq_next,
4253 	.show	= if6_seq_show,
4254 	.stop	= if6_seq_stop,
4255 };
4256 
4257 static int __net_init if6_proc_net_init(struct net *net)
4258 {
4259 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4260 			sizeof(struct if6_iter_state)))
4261 		return -ENOMEM;
4262 	return 0;
4263 }
4264 
4265 static void __net_exit if6_proc_net_exit(struct net *net)
4266 {
4267 	remove_proc_entry("if_inet6", net->proc_net);
4268 }
4269 
4270 static struct pernet_operations if6_proc_net_ops = {
4271 	.init = if6_proc_net_init,
4272 	.exit = if6_proc_net_exit,
4273 };
4274 
4275 int __init if6_proc_init(void)
4276 {
4277 	return register_pernet_subsys(&if6_proc_net_ops);
4278 }
4279 
4280 void if6_proc_exit(void)
4281 {
4282 	unregister_pernet_subsys(&if6_proc_net_ops);
4283 }
4284 #endif	/* CONFIG_PROC_FS */
4285 
4286 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4287 /* Check if address is a home address configured on any interface. */
4288 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4289 {
4290 	unsigned int hash = inet6_addr_hash(net, addr);
4291 	struct inet6_ifaddr *ifp = NULL;
4292 	int ret = 0;
4293 
4294 	rcu_read_lock();
4295 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4296 		if (!net_eq(dev_net(ifp->idev->dev), net))
4297 			continue;
4298 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4299 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4300 			ret = 1;
4301 			break;
4302 		}
4303 	}
4304 	rcu_read_unlock();
4305 	return ret;
4306 }
4307 #endif
4308 
4309 /*
4310  *	Periodic address status verification
4311  */
4312 
4313 static void addrconf_verify_rtnl(void)
4314 {
4315 	unsigned long now, next, next_sec, next_sched;
4316 	struct inet6_ifaddr *ifp;
4317 	int i;
4318 
4319 	ASSERT_RTNL();
4320 
4321 	rcu_read_lock_bh();
4322 	now = jiffies;
4323 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4324 
4325 	cancel_delayed_work(&addr_chk_work);
4326 
4327 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4328 restart:
4329 		hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4330 			unsigned long age;
4331 
4332 			/* When setting preferred_lft to a value not zero or
4333 			 * infinity, while valid_lft is infinity
4334 			 * IFA_F_PERMANENT has a non-infinity life time.
4335 			 */
4336 			if ((ifp->flags & IFA_F_PERMANENT) &&
4337 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4338 				continue;
4339 
4340 			spin_lock(&ifp->lock);
4341 			/* We try to batch several events at once. */
4342 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4343 
4344 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4345 			    age >= ifp->valid_lft) {
4346 				spin_unlock(&ifp->lock);
4347 				in6_ifa_hold(ifp);
4348 				ipv6_del_addr(ifp);
4349 				goto restart;
4350 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4351 				spin_unlock(&ifp->lock);
4352 				continue;
4353 			} else if (age >= ifp->prefered_lft) {
4354 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4355 				int deprecate = 0;
4356 
4357 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4358 					deprecate = 1;
4359 					ifp->flags |= IFA_F_DEPRECATED;
4360 				}
4361 
4362 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4363 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4364 					next = ifp->tstamp + ifp->valid_lft * HZ;
4365 
4366 				spin_unlock(&ifp->lock);
4367 
4368 				if (deprecate) {
4369 					in6_ifa_hold(ifp);
4370 
4371 					ipv6_ifa_notify(0, ifp);
4372 					in6_ifa_put(ifp);
4373 					goto restart;
4374 				}
4375 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4376 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4377 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4378 					ifp->idev->cnf.dad_transmits *
4379 					NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4380 
4381 				if (age >= ifp->prefered_lft - regen_advance) {
4382 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4383 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4384 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4385 					if (!ifp->regen_count && ifpub) {
4386 						ifp->regen_count++;
4387 						in6_ifa_hold(ifp);
4388 						in6_ifa_hold(ifpub);
4389 						spin_unlock(&ifp->lock);
4390 
4391 						spin_lock(&ifpub->lock);
4392 						ifpub->regen_count = 0;
4393 						spin_unlock(&ifpub->lock);
4394 						rcu_read_unlock_bh();
4395 						ipv6_create_tempaddr(ifpub, ifp, true);
4396 						in6_ifa_put(ifpub);
4397 						in6_ifa_put(ifp);
4398 						rcu_read_lock_bh();
4399 						goto restart;
4400 					}
4401 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4402 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4403 				spin_unlock(&ifp->lock);
4404 			} else {
4405 				/* ifp->prefered_lft <= ifp->valid_lft */
4406 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4407 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4408 				spin_unlock(&ifp->lock);
4409 			}
4410 		}
4411 	}
4412 
4413 	next_sec = round_jiffies_up(next);
4414 	next_sched = next;
4415 
4416 	/* If rounded timeout is accurate enough, accept it. */
4417 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4418 		next_sched = next_sec;
4419 
4420 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4421 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4422 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4423 
4424 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4425 		 now, next, next_sec, next_sched);
4426 	mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4427 	rcu_read_unlock_bh();
4428 }
4429 
4430 static void addrconf_verify_work(struct work_struct *w)
4431 {
4432 	rtnl_lock();
4433 	addrconf_verify_rtnl();
4434 	rtnl_unlock();
4435 }
4436 
4437 static void addrconf_verify(void)
4438 {
4439 	mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4440 }
4441 
4442 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4443 				     struct in6_addr **peer_pfx)
4444 {
4445 	struct in6_addr *pfx = NULL;
4446 
4447 	*peer_pfx = NULL;
4448 
4449 	if (addr)
4450 		pfx = nla_data(addr);
4451 
4452 	if (local) {
4453 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4454 			*peer_pfx = pfx;
4455 		pfx = nla_data(local);
4456 	}
4457 
4458 	return pfx;
4459 }
4460 
4461 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4462 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4463 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4464 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4465 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4466 };
4467 
4468 static int
4469 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4470 		  struct netlink_ext_ack *extack)
4471 {
4472 	struct net *net = sock_net(skb->sk);
4473 	struct ifaddrmsg *ifm;
4474 	struct nlattr *tb[IFA_MAX+1];
4475 	struct in6_addr *pfx, *peer_pfx;
4476 	u32 ifa_flags;
4477 	int err;
4478 
4479 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4480 			  extack);
4481 	if (err < 0)
4482 		return err;
4483 
4484 	ifm = nlmsg_data(nlh);
4485 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4486 	if (!pfx)
4487 		return -EINVAL;
4488 
4489 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4490 
4491 	/* We ignore other flags so far. */
4492 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4493 
4494 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4495 			      ifm->ifa_prefixlen);
4496 }
4497 
4498 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4499 			     u32 prefered_lft, u32 valid_lft)
4500 {
4501 	u32 flags;
4502 	clock_t expires;
4503 	unsigned long timeout;
4504 	bool was_managetempaddr;
4505 	bool had_prefixroute;
4506 
4507 	ASSERT_RTNL();
4508 
4509 	if (!valid_lft || (prefered_lft > valid_lft))
4510 		return -EINVAL;
4511 
4512 	if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4513 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4514 		return -EINVAL;
4515 
4516 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4517 		ifa_flags &= ~IFA_F_OPTIMISTIC;
4518 
4519 	timeout = addrconf_timeout_fixup(valid_lft, HZ);
4520 	if (addrconf_finite_timeout(timeout)) {
4521 		expires = jiffies_to_clock_t(timeout * HZ);
4522 		valid_lft = timeout;
4523 		flags = RTF_EXPIRES;
4524 	} else {
4525 		expires = 0;
4526 		flags = 0;
4527 		ifa_flags |= IFA_F_PERMANENT;
4528 	}
4529 
4530 	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4531 	if (addrconf_finite_timeout(timeout)) {
4532 		if (timeout == 0)
4533 			ifa_flags |= IFA_F_DEPRECATED;
4534 		prefered_lft = timeout;
4535 	}
4536 
4537 	spin_lock_bh(&ifp->lock);
4538 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4539 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4540 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4541 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4542 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4543 			IFA_F_NOPREFIXROUTE);
4544 	ifp->flags |= ifa_flags;
4545 	ifp->tstamp = jiffies;
4546 	ifp->valid_lft = valid_lft;
4547 	ifp->prefered_lft = prefered_lft;
4548 
4549 	spin_unlock_bh(&ifp->lock);
4550 	if (!(ifp->flags&IFA_F_TENTATIVE))
4551 		ipv6_ifa_notify(0, ifp);
4552 
4553 	if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4554 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4555 				      expires, flags);
4556 	} else if (had_prefixroute) {
4557 		enum cleanup_prefix_rt_t action;
4558 		unsigned long rt_expires;
4559 
4560 		write_lock_bh(&ifp->idev->lock);
4561 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4562 		write_unlock_bh(&ifp->idev->lock);
4563 
4564 		if (action != CLEANUP_PREFIX_RT_NOP) {
4565 			cleanup_prefix_route(ifp, rt_expires,
4566 				action == CLEANUP_PREFIX_RT_DEL);
4567 		}
4568 	}
4569 
4570 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4571 		if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4572 			valid_lft = prefered_lft = 0;
4573 		manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4574 				 !was_managetempaddr, jiffies);
4575 	}
4576 
4577 	addrconf_verify_rtnl();
4578 
4579 	return 0;
4580 }
4581 
4582 static int
4583 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4584 		  struct netlink_ext_ack *extack)
4585 {
4586 	struct net *net = sock_net(skb->sk);
4587 	struct ifaddrmsg *ifm;
4588 	struct nlattr *tb[IFA_MAX+1];
4589 	struct in6_addr *pfx, *peer_pfx;
4590 	struct inet6_ifaddr *ifa;
4591 	struct net_device *dev;
4592 	struct inet6_dev *idev;
4593 	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4594 	u32 ifa_flags;
4595 	int err;
4596 
4597 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4598 			  extack);
4599 	if (err < 0)
4600 		return err;
4601 
4602 	ifm = nlmsg_data(nlh);
4603 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4604 	if (!pfx)
4605 		return -EINVAL;
4606 
4607 	if (tb[IFA_CACHEINFO]) {
4608 		struct ifa_cacheinfo *ci;
4609 
4610 		ci = nla_data(tb[IFA_CACHEINFO]);
4611 		valid_lft = ci->ifa_valid;
4612 		preferred_lft = ci->ifa_prefered;
4613 	} else {
4614 		preferred_lft = INFINITY_LIFE_TIME;
4615 		valid_lft = INFINITY_LIFE_TIME;
4616 	}
4617 
4618 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4619 	if (!dev)
4620 		return -ENODEV;
4621 
4622 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4623 
4624 	/* We ignore other flags so far. */
4625 	ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4626 		     IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4627 
4628 	idev = ipv6_find_idev(dev);
4629 	if (IS_ERR(idev))
4630 		return PTR_ERR(idev);
4631 
4632 	if (!ipv6_allow_optimistic_dad(net, idev))
4633 		ifa_flags &= ~IFA_F_OPTIMISTIC;
4634 
4635 	if (ifa_flags & IFA_F_NODAD && ifa_flags & IFA_F_OPTIMISTIC) {
4636 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4637 		return -EINVAL;
4638 	}
4639 
4640 	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4641 	if (!ifa) {
4642 		/*
4643 		 * It would be best to check for !NLM_F_CREATE here but
4644 		 * userspace already relies on not having to provide this.
4645 		 */
4646 		return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4647 				      ifm->ifa_prefixlen, ifa_flags,
4648 				      preferred_lft, valid_lft, extack);
4649 	}
4650 
4651 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4652 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4653 		err = -EEXIST;
4654 	else
4655 		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4656 
4657 	in6_ifa_put(ifa);
4658 
4659 	return err;
4660 }
4661 
4662 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4663 			  u8 scope, int ifindex)
4664 {
4665 	struct ifaddrmsg *ifm;
4666 
4667 	ifm = nlmsg_data(nlh);
4668 	ifm->ifa_family = AF_INET6;
4669 	ifm->ifa_prefixlen = prefixlen;
4670 	ifm->ifa_flags = flags;
4671 	ifm->ifa_scope = scope;
4672 	ifm->ifa_index = ifindex;
4673 }
4674 
4675 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4676 			 unsigned long tstamp, u32 preferred, u32 valid)
4677 {
4678 	struct ifa_cacheinfo ci;
4679 
4680 	ci.cstamp = cstamp_delta(cstamp);
4681 	ci.tstamp = cstamp_delta(tstamp);
4682 	ci.ifa_prefered = preferred;
4683 	ci.ifa_valid = valid;
4684 
4685 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4686 }
4687 
4688 static inline int rt_scope(int ifa_scope)
4689 {
4690 	if (ifa_scope & IFA_HOST)
4691 		return RT_SCOPE_HOST;
4692 	else if (ifa_scope & IFA_LINK)
4693 		return RT_SCOPE_LINK;
4694 	else if (ifa_scope & IFA_SITE)
4695 		return RT_SCOPE_SITE;
4696 	else
4697 		return RT_SCOPE_UNIVERSE;
4698 }
4699 
4700 static inline int inet6_ifaddr_msgsize(void)
4701 {
4702 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4703 	       + nla_total_size(16) /* IFA_LOCAL */
4704 	       + nla_total_size(16) /* IFA_ADDRESS */
4705 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4706 	       + nla_total_size(4)  /* IFA_FLAGS */;
4707 }
4708 
4709 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4710 			     u32 portid, u32 seq, int event, unsigned int flags)
4711 {
4712 	struct nlmsghdr  *nlh;
4713 	u32 preferred, valid;
4714 
4715 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4716 	if (!nlh)
4717 		return -EMSGSIZE;
4718 
4719 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4720 		      ifa->idev->dev->ifindex);
4721 
4722 	if (!((ifa->flags&IFA_F_PERMANENT) &&
4723 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4724 		preferred = ifa->prefered_lft;
4725 		valid = ifa->valid_lft;
4726 		if (preferred != INFINITY_LIFE_TIME) {
4727 			long tval = (jiffies - ifa->tstamp)/HZ;
4728 			if (preferred > tval)
4729 				preferred -= tval;
4730 			else
4731 				preferred = 0;
4732 			if (valid != INFINITY_LIFE_TIME) {
4733 				if (valid > tval)
4734 					valid -= tval;
4735 				else
4736 					valid = 0;
4737 			}
4738 		}
4739 	} else {
4740 		preferred = INFINITY_LIFE_TIME;
4741 		valid = INFINITY_LIFE_TIME;
4742 	}
4743 
4744 	if (!ipv6_addr_any(&ifa->peer_addr)) {
4745 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4746 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4747 			goto error;
4748 	} else
4749 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4750 			goto error;
4751 
4752 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4753 		goto error;
4754 
4755 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4756 		goto error;
4757 
4758 	nlmsg_end(skb, nlh);
4759 	return 0;
4760 
4761 error:
4762 	nlmsg_cancel(skb, nlh);
4763 	return -EMSGSIZE;
4764 }
4765 
4766 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4767 				u32 portid, u32 seq, int event, u16 flags)
4768 {
4769 	struct nlmsghdr  *nlh;
4770 	u8 scope = RT_SCOPE_UNIVERSE;
4771 	int ifindex = ifmca->idev->dev->ifindex;
4772 
4773 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4774 		scope = RT_SCOPE_SITE;
4775 
4776 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4777 	if (!nlh)
4778 		return -EMSGSIZE;
4779 
4780 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4781 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4782 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4783 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4784 		nlmsg_cancel(skb, nlh);
4785 		return -EMSGSIZE;
4786 	}
4787 
4788 	nlmsg_end(skb, nlh);
4789 	return 0;
4790 }
4791 
4792 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4793 				u32 portid, u32 seq, int event, unsigned int flags)
4794 {
4795 	struct nlmsghdr  *nlh;
4796 	u8 scope = RT_SCOPE_UNIVERSE;
4797 	int ifindex = ifaca->aca_idev->dev->ifindex;
4798 
4799 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4800 		scope = RT_SCOPE_SITE;
4801 
4802 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4803 	if (!nlh)
4804 		return -EMSGSIZE;
4805 
4806 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4807 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4808 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4809 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4810 		nlmsg_cancel(skb, nlh);
4811 		return -EMSGSIZE;
4812 	}
4813 
4814 	nlmsg_end(skb, nlh);
4815 	return 0;
4816 }
4817 
4818 enum addr_type_t {
4819 	UNICAST_ADDR,
4820 	MULTICAST_ADDR,
4821 	ANYCAST_ADDR,
4822 };
4823 
4824 /* called with rcu_read_lock() */
4825 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4826 			  struct netlink_callback *cb, enum addr_type_t type,
4827 			  int s_ip_idx, int *p_ip_idx)
4828 {
4829 	struct ifmcaddr6 *ifmca;
4830 	struct ifacaddr6 *ifaca;
4831 	int err = 1;
4832 	int ip_idx = *p_ip_idx;
4833 
4834 	read_lock_bh(&idev->lock);
4835 	switch (type) {
4836 	case UNICAST_ADDR: {
4837 		struct inet6_ifaddr *ifa;
4838 
4839 		/* unicast address incl. temp addr */
4840 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
4841 			if (++ip_idx < s_ip_idx)
4842 				continue;
4843 			err = inet6_fill_ifaddr(skb, ifa,
4844 						NETLINK_CB(cb->skb).portid,
4845 						cb->nlh->nlmsg_seq,
4846 						RTM_NEWADDR,
4847 						NLM_F_MULTI);
4848 			if (err < 0)
4849 				break;
4850 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4851 		}
4852 		break;
4853 	}
4854 	case MULTICAST_ADDR:
4855 		/* multicast address */
4856 		for (ifmca = idev->mc_list; ifmca;
4857 		     ifmca = ifmca->next, ip_idx++) {
4858 			if (ip_idx < s_ip_idx)
4859 				continue;
4860 			err = inet6_fill_ifmcaddr(skb, ifmca,
4861 						  NETLINK_CB(cb->skb).portid,
4862 						  cb->nlh->nlmsg_seq,
4863 						  RTM_GETMULTICAST,
4864 						  NLM_F_MULTI);
4865 			if (err < 0)
4866 				break;
4867 		}
4868 		break;
4869 	case ANYCAST_ADDR:
4870 		/* anycast address */
4871 		for (ifaca = idev->ac_list; ifaca;
4872 		     ifaca = ifaca->aca_next, ip_idx++) {
4873 			if (ip_idx < s_ip_idx)
4874 				continue;
4875 			err = inet6_fill_ifacaddr(skb, ifaca,
4876 						  NETLINK_CB(cb->skb).portid,
4877 						  cb->nlh->nlmsg_seq,
4878 						  RTM_GETANYCAST,
4879 						  NLM_F_MULTI);
4880 			if (err < 0)
4881 				break;
4882 		}
4883 		break;
4884 	default:
4885 		break;
4886 	}
4887 	read_unlock_bh(&idev->lock);
4888 	*p_ip_idx = ip_idx;
4889 	return err;
4890 }
4891 
4892 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4893 			   enum addr_type_t type)
4894 {
4895 	struct net *net = sock_net(skb->sk);
4896 	int h, s_h;
4897 	int idx, ip_idx;
4898 	int s_idx, s_ip_idx;
4899 	struct net_device *dev;
4900 	struct inet6_dev *idev;
4901 	struct hlist_head *head;
4902 
4903 	s_h = cb->args[0];
4904 	s_idx = idx = cb->args[1];
4905 	s_ip_idx = ip_idx = cb->args[2];
4906 
4907 	rcu_read_lock();
4908 	cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4909 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4910 		idx = 0;
4911 		head = &net->dev_index_head[h];
4912 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
4913 			if (idx < s_idx)
4914 				goto cont;
4915 			if (h > s_h || idx > s_idx)
4916 				s_ip_idx = 0;
4917 			ip_idx = 0;
4918 			idev = __in6_dev_get(dev);
4919 			if (!idev)
4920 				goto cont;
4921 
4922 			if (in6_dump_addrs(idev, skb, cb, type,
4923 					   s_ip_idx, &ip_idx) < 0)
4924 				goto done;
4925 cont:
4926 			idx++;
4927 		}
4928 	}
4929 done:
4930 	rcu_read_unlock();
4931 	cb->args[0] = h;
4932 	cb->args[1] = idx;
4933 	cb->args[2] = ip_idx;
4934 
4935 	return skb->len;
4936 }
4937 
4938 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4939 {
4940 	enum addr_type_t type = UNICAST_ADDR;
4941 
4942 	return inet6_dump_addr(skb, cb, type);
4943 }
4944 
4945 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4946 {
4947 	enum addr_type_t type = MULTICAST_ADDR;
4948 
4949 	return inet6_dump_addr(skb, cb, type);
4950 }
4951 
4952 
4953 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4954 {
4955 	enum addr_type_t type = ANYCAST_ADDR;
4956 
4957 	return inet6_dump_addr(skb, cb, type);
4958 }
4959 
4960 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4961 			     struct netlink_ext_ack *extack)
4962 {
4963 	struct net *net = sock_net(in_skb->sk);
4964 	struct ifaddrmsg *ifm;
4965 	struct nlattr *tb[IFA_MAX+1];
4966 	struct in6_addr *addr = NULL, *peer;
4967 	struct net_device *dev = NULL;
4968 	struct inet6_ifaddr *ifa;
4969 	struct sk_buff *skb;
4970 	int err;
4971 
4972 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4973 			  extack);
4974 	if (err < 0)
4975 		return err;
4976 
4977 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4978 	if (!addr)
4979 		return -EINVAL;
4980 
4981 	ifm = nlmsg_data(nlh);
4982 	if (ifm->ifa_index)
4983 		dev = dev_get_by_index(net, ifm->ifa_index);
4984 
4985 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4986 	if (!ifa) {
4987 		err = -EADDRNOTAVAIL;
4988 		goto errout;
4989 	}
4990 
4991 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4992 	if (!skb) {
4993 		err = -ENOBUFS;
4994 		goto errout_ifa;
4995 	}
4996 
4997 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4998 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
4999 	if (err < 0) {
5000 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5001 		WARN_ON(err == -EMSGSIZE);
5002 		kfree_skb(skb);
5003 		goto errout_ifa;
5004 	}
5005 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5006 errout_ifa:
5007 	in6_ifa_put(ifa);
5008 errout:
5009 	if (dev)
5010 		dev_put(dev);
5011 	return err;
5012 }
5013 
5014 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5015 {
5016 	struct sk_buff *skb;
5017 	struct net *net = dev_net(ifa->idev->dev);
5018 	int err = -ENOBUFS;
5019 
5020 	/* Don't send DELADDR notification for TENTATIVE address,
5021 	 * since NEWADDR notification is sent only after removing
5022 	 * TENTATIVE flag, if DAD has not failed.
5023 	 */
5024 	if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
5025 	    event == RTM_DELADDR)
5026 		return;
5027 
5028 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5029 	if (!skb)
5030 		goto errout;
5031 
5032 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5033 	if (err < 0) {
5034 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5035 		WARN_ON(err == -EMSGSIZE);
5036 		kfree_skb(skb);
5037 		goto errout;
5038 	}
5039 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5040 	return;
5041 errout:
5042 	if (err < 0)
5043 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5044 }
5045 
5046 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5047 				__s32 *array, int bytes)
5048 {
5049 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5050 
5051 	memset(array, 0, bytes);
5052 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5053 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5054 	array[DEVCONF_MTU6] = cnf->mtu6;
5055 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5056 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5057 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5058 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5059 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5060 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5061 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5062 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5063 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5064 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5065 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5066 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5067 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5068 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5069 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5070 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5071 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5072 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5073 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5074 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5075 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5076 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5077 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5078 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5079 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5080 #ifdef CONFIG_IPV6_ROUTER_PREF
5081 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5082 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5083 		jiffies_to_msecs(cnf->rtr_probe_interval);
5084 #ifdef CONFIG_IPV6_ROUTE_INFO
5085 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5086 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5087 #endif
5088 #endif
5089 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5090 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5091 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5092 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5093 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5094 #endif
5095 #ifdef CONFIG_IPV6_MROUTE
5096 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5097 #endif
5098 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5099 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5100 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5101 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5102 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5103 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5104 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5105 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5106 	/* we omit DEVCONF_STABLE_SECRET for now */
5107 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5108 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5109 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5110 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5111 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5112 #ifdef CONFIG_IPV6_SEG6_HMAC
5113 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5114 #endif
5115 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5116 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5117 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5118 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5119 }
5120 
5121 static inline size_t inet6_ifla6_size(void)
5122 {
5123 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5124 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5125 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5126 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5127 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5128 	     + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5129 }
5130 
5131 static inline size_t inet6_if_nlmsg_size(void)
5132 {
5133 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5134 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5135 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5136 	       + nla_total_size(4) /* IFLA_MTU */
5137 	       + nla_total_size(4) /* IFLA_LINK */
5138 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5139 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5140 }
5141 
5142 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5143 					int bytes)
5144 {
5145 	int i;
5146 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5147 	BUG_ON(pad < 0);
5148 
5149 	/* Use put_unaligned() because stats may not be aligned for u64. */
5150 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5151 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5152 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5153 
5154 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5155 }
5156 
5157 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5158 					int bytes, size_t syncpoff)
5159 {
5160 	int i, c;
5161 	u64 buff[IPSTATS_MIB_MAX];
5162 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5163 
5164 	BUG_ON(pad < 0);
5165 
5166 	memset(buff, 0, sizeof(buff));
5167 	buff[0] = IPSTATS_MIB_MAX;
5168 
5169 	for_each_possible_cpu(c) {
5170 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5171 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5172 	}
5173 
5174 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5175 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5176 }
5177 
5178 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5179 			     int bytes)
5180 {
5181 	switch (attrtype) {
5182 	case IFLA_INET6_STATS:
5183 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5184 				     offsetof(struct ipstats_mib, syncp));
5185 		break;
5186 	case IFLA_INET6_ICMP6STATS:
5187 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5188 		break;
5189 	}
5190 }
5191 
5192 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5193 				  u32 ext_filter_mask)
5194 {
5195 	struct nlattr *nla;
5196 	struct ifla_cacheinfo ci;
5197 
5198 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5199 		goto nla_put_failure;
5200 	ci.max_reasm_len = IPV6_MAXPLEN;
5201 	ci.tstamp = cstamp_delta(idev->tstamp);
5202 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5203 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5204 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5205 		goto nla_put_failure;
5206 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5207 	if (!nla)
5208 		goto nla_put_failure;
5209 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5210 
5211 	/* XXX - MC not implemented */
5212 
5213 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5214 		return 0;
5215 
5216 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5217 	if (!nla)
5218 		goto nla_put_failure;
5219 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5220 
5221 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5222 	if (!nla)
5223 		goto nla_put_failure;
5224 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5225 
5226 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5227 	if (!nla)
5228 		goto nla_put_failure;
5229 
5230 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5231 		goto nla_put_failure;
5232 
5233 	read_lock_bh(&idev->lock);
5234 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5235 	read_unlock_bh(&idev->lock);
5236 
5237 	return 0;
5238 
5239 nla_put_failure:
5240 	return -EMSGSIZE;
5241 }
5242 
5243 static size_t inet6_get_link_af_size(const struct net_device *dev,
5244 				     u32 ext_filter_mask)
5245 {
5246 	if (!__in6_dev_get(dev))
5247 		return 0;
5248 
5249 	return inet6_ifla6_size();
5250 }
5251 
5252 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5253 			      u32 ext_filter_mask)
5254 {
5255 	struct inet6_dev *idev = __in6_dev_get(dev);
5256 
5257 	if (!idev)
5258 		return -ENODATA;
5259 
5260 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5261 		return -EMSGSIZE;
5262 
5263 	return 0;
5264 }
5265 
5266 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5267 {
5268 	struct inet6_ifaddr *ifp;
5269 	struct net_device *dev = idev->dev;
5270 	bool clear_token, update_rs = false;
5271 	struct in6_addr ll_addr;
5272 
5273 	ASSERT_RTNL();
5274 
5275 	if (!token)
5276 		return -EINVAL;
5277 	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5278 		return -EINVAL;
5279 	if (!ipv6_accept_ra(idev))
5280 		return -EINVAL;
5281 	if (idev->cnf.rtr_solicits == 0)
5282 		return -EINVAL;
5283 
5284 	write_lock_bh(&idev->lock);
5285 
5286 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5287 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5288 
5289 	write_unlock_bh(&idev->lock);
5290 
5291 	clear_token = ipv6_addr_any(token);
5292 	if (clear_token)
5293 		goto update_lft;
5294 
5295 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5296 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5297 			     IFA_F_OPTIMISTIC)) {
5298 		/* If we're not ready, then normal ifup will take care
5299 		 * of this. Otherwise, we need to request our rs here.
5300 		 */
5301 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5302 		update_rs = true;
5303 	}
5304 
5305 update_lft:
5306 	write_lock_bh(&idev->lock);
5307 
5308 	if (update_rs) {
5309 		idev->if_flags |= IF_RS_SENT;
5310 		idev->rs_interval = rfc3315_s14_backoff_init(
5311 			idev->cnf.rtr_solicit_interval);
5312 		idev->rs_probes = 1;
5313 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5314 	}
5315 
5316 	/* Well, that's kinda nasty ... */
5317 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5318 		spin_lock(&ifp->lock);
5319 		if (ifp->tokenized) {
5320 			ifp->valid_lft = 0;
5321 			ifp->prefered_lft = 0;
5322 		}
5323 		spin_unlock(&ifp->lock);
5324 	}
5325 
5326 	write_unlock_bh(&idev->lock);
5327 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5328 	addrconf_verify_rtnl();
5329 	return 0;
5330 }
5331 
5332 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5333 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5334 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5335 };
5336 
5337 static int inet6_validate_link_af(const struct net_device *dev,
5338 				  const struct nlattr *nla)
5339 {
5340 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5341 
5342 	if (dev && !__in6_dev_get(dev))
5343 		return -EAFNOSUPPORT;
5344 
5345 	return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5346 				NULL);
5347 }
5348 
5349 static int check_addr_gen_mode(int mode)
5350 {
5351 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5352 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5353 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5354 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5355 		return -EINVAL;
5356 	return 1;
5357 }
5358 
5359 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5360 				int mode)
5361 {
5362 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5363 	    !idev->cnf.stable_secret.initialized &&
5364 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5365 		return -EINVAL;
5366 	return 1;
5367 }
5368 
5369 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5370 {
5371 	int err = -EINVAL;
5372 	struct inet6_dev *idev = __in6_dev_get(dev);
5373 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5374 
5375 	if (!idev)
5376 		return -EAFNOSUPPORT;
5377 
5378 	if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5379 		BUG();
5380 
5381 	if (tb[IFLA_INET6_TOKEN]) {
5382 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5383 		if (err)
5384 			return err;
5385 	}
5386 
5387 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5388 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5389 
5390 		if (check_addr_gen_mode(mode) < 0 ||
5391 		    check_stable_privacy(idev, dev_net(dev), mode) < 0)
5392 			return -EINVAL;
5393 
5394 		idev->cnf.addr_gen_mode = mode;
5395 		err = 0;
5396 	}
5397 
5398 	return err;
5399 }
5400 
5401 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5402 			     u32 portid, u32 seq, int event, unsigned int flags)
5403 {
5404 	struct net_device *dev = idev->dev;
5405 	struct ifinfomsg *hdr;
5406 	struct nlmsghdr *nlh;
5407 	void *protoinfo;
5408 
5409 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5410 	if (!nlh)
5411 		return -EMSGSIZE;
5412 
5413 	hdr = nlmsg_data(nlh);
5414 	hdr->ifi_family = AF_INET6;
5415 	hdr->__ifi_pad = 0;
5416 	hdr->ifi_type = dev->type;
5417 	hdr->ifi_index = dev->ifindex;
5418 	hdr->ifi_flags = dev_get_flags(dev);
5419 	hdr->ifi_change = 0;
5420 
5421 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5422 	    (dev->addr_len &&
5423 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5424 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5425 	    (dev->ifindex != dev_get_iflink(dev) &&
5426 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5427 	    nla_put_u8(skb, IFLA_OPERSTATE,
5428 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5429 		goto nla_put_failure;
5430 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5431 	if (!protoinfo)
5432 		goto nla_put_failure;
5433 
5434 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5435 		goto nla_put_failure;
5436 
5437 	nla_nest_end(skb, protoinfo);
5438 	nlmsg_end(skb, nlh);
5439 	return 0;
5440 
5441 nla_put_failure:
5442 	nlmsg_cancel(skb, nlh);
5443 	return -EMSGSIZE;
5444 }
5445 
5446 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5447 {
5448 	struct net *net = sock_net(skb->sk);
5449 	int h, s_h;
5450 	int idx = 0, s_idx;
5451 	struct net_device *dev;
5452 	struct inet6_dev *idev;
5453 	struct hlist_head *head;
5454 
5455 	s_h = cb->args[0];
5456 	s_idx = cb->args[1];
5457 
5458 	rcu_read_lock();
5459 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5460 		idx = 0;
5461 		head = &net->dev_index_head[h];
5462 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5463 			if (idx < s_idx)
5464 				goto cont;
5465 			idev = __in6_dev_get(dev);
5466 			if (!idev)
5467 				goto cont;
5468 			if (inet6_fill_ifinfo(skb, idev,
5469 					      NETLINK_CB(cb->skb).portid,
5470 					      cb->nlh->nlmsg_seq,
5471 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
5472 				goto out;
5473 cont:
5474 			idx++;
5475 		}
5476 	}
5477 out:
5478 	rcu_read_unlock();
5479 	cb->args[1] = idx;
5480 	cb->args[0] = h;
5481 
5482 	return skb->len;
5483 }
5484 
5485 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5486 {
5487 	struct sk_buff *skb;
5488 	struct net *net = dev_net(idev->dev);
5489 	int err = -ENOBUFS;
5490 
5491 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5492 	if (!skb)
5493 		goto errout;
5494 
5495 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5496 	if (err < 0) {
5497 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5498 		WARN_ON(err == -EMSGSIZE);
5499 		kfree_skb(skb);
5500 		goto errout;
5501 	}
5502 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5503 	return;
5504 errout:
5505 	if (err < 0)
5506 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5507 }
5508 
5509 static inline size_t inet6_prefix_nlmsg_size(void)
5510 {
5511 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
5512 	       + nla_total_size(sizeof(struct in6_addr))
5513 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
5514 }
5515 
5516 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5517 			     struct prefix_info *pinfo, u32 portid, u32 seq,
5518 			     int event, unsigned int flags)
5519 {
5520 	struct prefixmsg *pmsg;
5521 	struct nlmsghdr *nlh;
5522 	struct prefix_cacheinfo	ci;
5523 
5524 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5525 	if (!nlh)
5526 		return -EMSGSIZE;
5527 
5528 	pmsg = nlmsg_data(nlh);
5529 	pmsg->prefix_family = AF_INET6;
5530 	pmsg->prefix_pad1 = 0;
5531 	pmsg->prefix_pad2 = 0;
5532 	pmsg->prefix_ifindex = idev->dev->ifindex;
5533 	pmsg->prefix_len = pinfo->prefix_len;
5534 	pmsg->prefix_type = pinfo->type;
5535 	pmsg->prefix_pad3 = 0;
5536 	pmsg->prefix_flags = 0;
5537 	if (pinfo->onlink)
5538 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5539 	if (pinfo->autoconf)
5540 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5541 
5542 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5543 		goto nla_put_failure;
5544 	ci.preferred_time = ntohl(pinfo->prefered);
5545 	ci.valid_time = ntohl(pinfo->valid);
5546 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5547 		goto nla_put_failure;
5548 	nlmsg_end(skb, nlh);
5549 	return 0;
5550 
5551 nla_put_failure:
5552 	nlmsg_cancel(skb, nlh);
5553 	return -EMSGSIZE;
5554 }
5555 
5556 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5557 			 struct prefix_info *pinfo)
5558 {
5559 	struct sk_buff *skb;
5560 	struct net *net = dev_net(idev->dev);
5561 	int err = -ENOBUFS;
5562 
5563 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5564 	if (!skb)
5565 		goto errout;
5566 
5567 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5568 	if (err < 0) {
5569 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5570 		WARN_ON(err == -EMSGSIZE);
5571 		kfree_skb(skb);
5572 		goto errout;
5573 	}
5574 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5575 	return;
5576 errout:
5577 	if (err < 0)
5578 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5579 }
5580 
5581 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5582 {
5583 	struct net *net = dev_net(ifp->idev->dev);
5584 
5585 	if (event)
5586 		ASSERT_RTNL();
5587 
5588 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5589 
5590 	switch (event) {
5591 	case RTM_NEWADDR:
5592 		/*
5593 		 * If the address was optimistic
5594 		 * we inserted the route at the start of
5595 		 * our DAD process, so we don't need
5596 		 * to do it again
5597 		 */
5598 		if (!rcu_access_pointer(ifp->rt->rt6i_node))
5599 			ip6_ins_rt(ifp->rt);
5600 		if (ifp->idev->cnf.forwarding)
5601 			addrconf_join_anycast(ifp);
5602 		if (!ipv6_addr_any(&ifp->peer_addr))
5603 			addrconf_prefix_route(&ifp->peer_addr, 128,
5604 					      ifp->idev->dev, 0, 0);
5605 		break;
5606 	case RTM_DELADDR:
5607 		if (ifp->idev->cnf.forwarding)
5608 			addrconf_leave_anycast(ifp);
5609 		addrconf_leave_solict(ifp->idev, &ifp->addr);
5610 		if (!ipv6_addr_any(&ifp->peer_addr)) {
5611 			struct rt6_info *rt;
5612 
5613 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5614 						       ifp->idev->dev, 0, 0);
5615 			if (rt)
5616 				ip6_del_rt(rt);
5617 		}
5618 		if (ifp->rt) {
5619 			if (dst_hold_safe(&ifp->rt->dst))
5620 				ip6_del_rt(ifp->rt);
5621 		}
5622 		rt_genid_bump_ipv6(net);
5623 		break;
5624 	}
5625 	atomic_inc(&net->ipv6.dev_addr_genid);
5626 }
5627 
5628 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5629 {
5630 	rcu_read_lock_bh();
5631 	if (likely(ifp->idev->dead == 0))
5632 		__ipv6_ifa_notify(event, ifp);
5633 	rcu_read_unlock_bh();
5634 }
5635 
5636 #ifdef CONFIG_SYSCTL
5637 
5638 static
5639 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5640 			   void __user *buffer, size_t *lenp, loff_t *ppos)
5641 {
5642 	int *valp = ctl->data;
5643 	int val = *valp;
5644 	loff_t pos = *ppos;
5645 	struct ctl_table lctl;
5646 	int ret;
5647 
5648 	/*
5649 	 * ctl->data points to idev->cnf.forwarding, we should
5650 	 * not modify it until we get the rtnl lock.
5651 	 */
5652 	lctl = *ctl;
5653 	lctl.data = &val;
5654 
5655 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5656 
5657 	if (write)
5658 		ret = addrconf_fixup_forwarding(ctl, valp, val);
5659 	if (ret)
5660 		*ppos = pos;
5661 	return ret;
5662 }
5663 
5664 static
5665 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5666 			void __user *buffer, size_t *lenp, loff_t *ppos)
5667 {
5668 	struct inet6_dev *idev = ctl->extra1;
5669 	int min_mtu = IPV6_MIN_MTU;
5670 	struct ctl_table lctl;
5671 
5672 	lctl = *ctl;
5673 	lctl.extra1 = &min_mtu;
5674 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5675 
5676 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5677 }
5678 
5679 static void dev_disable_change(struct inet6_dev *idev)
5680 {
5681 	struct netdev_notifier_info info;
5682 
5683 	if (!idev || !idev->dev)
5684 		return;
5685 
5686 	netdev_notifier_info_init(&info, idev->dev);
5687 	if (idev->cnf.disable_ipv6)
5688 		addrconf_notify(NULL, NETDEV_DOWN, &info);
5689 	else
5690 		addrconf_notify(NULL, NETDEV_UP, &info);
5691 }
5692 
5693 static void addrconf_disable_change(struct net *net, __s32 newf)
5694 {
5695 	struct net_device *dev;
5696 	struct inet6_dev *idev;
5697 
5698 	for_each_netdev(net, dev) {
5699 		idev = __in6_dev_get(dev);
5700 		if (idev) {
5701 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5702 			idev->cnf.disable_ipv6 = newf;
5703 			if (changed)
5704 				dev_disable_change(idev);
5705 		}
5706 	}
5707 }
5708 
5709 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5710 {
5711 	struct net *net;
5712 	int old;
5713 
5714 	if (!rtnl_trylock())
5715 		return restart_syscall();
5716 
5717 	net = (struct net *)table->extra2;
5718 	old = *p;
5719 	*p = newf;
5720 
5721 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5722 		rtnl_unlock();
5723 		return 0;
5724 	}
5725 
5726 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
5727 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
5728 		addrconf_disable_change(net, newf);
5729 	} else if ((!newf) ^ (!old))
5730 		dev_disable_change((struct inet6_dev *)table->extra1);
5731 
5732 	rtnl_unlock();
5733 	return 0;
5734 }
5735 
5736 static
5737 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5738 			    void __user *buffer, size_t *lenp, loff_t *ppos)
5739 {
5740 	int *valp = ctl->data;
5741 	int val = *valp;
5742 	loff_t pos = *ppos;
5743 	struct ctl_table lctl;
5744 	int ret;
5745 
5746 	/*
5747 	 * ctl->data points to idev->cnf.disable_ipv6, we should
5748 	 * not modify it until we get the rtnl lock.
5749 	 */
5750 	lctl = *ctl;
5751 	lctl.data = &val;
5752 
5753 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5754 
5755 	if (write)
5756 		ret = addrconf_disable_ipv6(ctl, valp, val);
5757 	if (ret)
5758 		*ppos = pos;
5759 	return ret;
5760 }
5761 
5762 static
5763 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5764 			      void __user *buffer, size_t *lenp, loff_t *ppos)
5765 {
5766 	int *valp = ctl->data;
5767 	int ret;
5768 	int old, new;
5769 
5770 	old = *valp;
5771 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5772 	new = *valp;
5773 
5774 	if (write && old != new) {
5775 		struct net *net = ctl->extra2;
5776 
5777 		if (!rtnl_trylock())
5778 			return restart_syscall();
5779 
5780 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5781 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5782 						     NETCONFA_PROXY_NEIGH,
5783 						     NETCONFA_IFINDEX_DEFAULT,
5784 						     net->ipv6.devconf_dflt);
5785 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5786 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5787 						     NETCONFA_PROXY_NEIGH,
5788 						     NETCONFA_IFINDEX_ALL,
5789 						     net->ipv6.devconf_all);
5790 		else {
5791 			struct inet6_dev *idev = ctl->extra1;
5792 
5793 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5794 						     NETCONFA_PROXY_NEIGH,
5795 						     idev->dev->ifindex,
5796 						     &idev->cnf);
5797 		}
5798 		rtnl_unlock();
5799 	}
5800 
5801 	return ret;
5802 }
5803 
5804 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5805 					 void __user *buffer, size_t *lenp,
5806 					 loff_t *ppos)
5807 {
5808 	int ret = 0;
5809 	int new_val;
5810 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5811 	struct net *net = (struct net *)ctl->extra2;
5812 
5813 	if (!rtnl_trylock())
5814 		return restart_syscall();
5815 
5816 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5817 
5818 	if (write) {
5819 		new_val = *((int *)ctl->data);
5820 
5821 		if (check_addr_gen_mode(new_val) < 0) {
5822 			ret = -EINVAL;
5823 			goto out;
5824 		}
5825 
5826 		/* request for default */
5827 		if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5828 			ipv6_devconf_dflt.addr_gen_mode = new_val;
5829 
5830 		/* request for individual net device */
5831 		} else {
5832 			if (!idev)
5833 				goto out;
5834 
5835 			if (check_stable_privacy(idev, net, new_val) < 0) {
5836 				ret = -EINVAL;
5837 				goto out;
5838 			}
5839 
5840 			if (idev->cnf.addr_gen_mode != new_val) {
5841 				idev->cnf.addr_gen_mode = new_val;
5842 				addrconf_dev_config(idev->dev);
5843 			}
5844 		}
5845 	}
5846 
5847 out:
5848 	rtnl_unlock();
5849 
5850 	return ret;
5851 }
5852 
5853 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5854 					 void __user *buffer, size_t *lenp,
5855 					 loff_t *ppos)
5856 {
5857 	int err;
5858 	struct in6_addr addr;
5859 	char str[IPV6_MAX_STRLEN];
5860 	struct ctl_table lctl = *ctl;
5861 	struct net *net = ctl->extra2;
5862 	struct ipv6_stable_secret *secret = ctl->data;
5863 
5864 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5865 		return -EIO;
5866 
5867 	lctl.maxlen = IPV6_MAX_STRLEN;
5868 	lctl.data = str;
5869 
5870 	if (!rtnl_trylock())
5871 		return restart_syscall();
5872 
5873 	if (!write && !secret->initialized) {
5874 		err = -EIO;
5875 		goto out;
5876 	}
5877 
5878 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5879 	if (err >= sizeof(str)) {
5880 		err = -EIO;
5881 		goto out;
5882 	}
5883 
5884 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5885 	if (err || !write)
5886 		goto out;
5887 
5888 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5889 		err = -EIO;
5890 		goto out;
5891 	}
5892 
5893 	secret->initialized = true;
5894 	secret->secret = addr;
5895 
5896 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5897 		struct net_device *dev;
5898 
5899 		for_each_netdev(net, dev) {
5900 			struct inet6_dev *idev = __in6_dev_get(dev);
5901 
5902 			if (idev) {
5903 				idev->cnf.addr_gen_mode =
5904 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5905 			}
5906 		}
5907 	} else {
5908 		struct inet6_dev *idev = ctl->extra1;
5909 
5910 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5911 	}
5912 
5913 out:
5914 	rtnl_unlock();
5915 
5916 	return err;
5917 }
5918 
5919 static
5920 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5921 						int write,
5922 						void __user *buffer,
5923 						size_t *lenp,
5924 						loff_t *ppos)
5925 {
5926 	int *valp = ctl->data;
5927 	int val = *valp;
5928 	loff_t pos = *ppos;
5929 	struct ctl_table lctl;
5930 	int ret;
5931 
5932 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5933 	 * we should not modify it until we get the rtnl lock.
5934 	 */
5935 	lctl = *ctl;
5936 	lctl.data = &val;
5937 
5938 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5939 
5940 	if (write)
5941 		ret = addrconf_fixup_linkdown(ctl, valp, val);
5942 	if (ret)
5943 		*ppos = pos;
5944 	return ret;
5945 }
5946 
5947 static
5948 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5949 {
5950 	if (rt) {
5951 		if (action)
5952 			rt->dst.flags |= DST_NOPOLICY;
5953 		else
5954 			rt->dst.flags &= ~DST_NOPOLICY;
5955 	}
5956 }
5957 
5958 static
5959 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5960 {
5961 	struct inet6_ifaddr *ifa;
5962 
5963 	read_lock_bh(&idev->lock);
5964 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
5965 		spin_lock(&ifa->lock);
5966 		if (ifa->rt) {
5967 			struct rt6_info *rt = ifa->rt;
5968 			int cpu;
5969 
5970 			rcu_read_lock();
5971 			addrconf_set_nopolicy(ifa->rt, val);
5972 			if (rt->rt6i_pcpu) {
5973 				for_each_possible_cpu(cpu) {
5974 					struct rt6_info **rtp;
5975 
5976 					rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5977 					addrconf_set_nopolicy(*rtp, val);
5978 				}
5979 			}
5980 			rcu_read_unlock();
5981 		}
5982 		spin_unlock(&ifa->lock);
5983 	}
5984 	read_unlock_bh(&idev->lock);
5985 }
5986 
5987 static
5988 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5989 {
5990 	struct inet6_dev *idev;
5991 	struct net *net;
5992 
5993 	if (!rtnl_trylock())
5994 		return restart_syscall();
5995 
5996 	*valp = val;
5997 
5998 	net = (struct net *)ctl->extra2;
5999 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6000 		rtnl_unlock();
6001 		return 0;
6002 	}
6003 
6004 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6005 		struct net_device *dev;
6006 
6007 		for_each_netdev(net, dev) {
6008 			idev = __in6_dev_get(dev);
6009 			if (idev)
6010 				addrconf_disable_policy_idev(idev, val);
6011 		}
6012 	} else {
6013 		idev = (struct inet6_dev *)ctl->extra1;
6014 		addrconf_disable_policy_idev(idev, val);
6015 	}
6016 
6017 	rtnl_unlock();
6018 	return 0;
6019 }
6020 
6021 static
6022 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6023 				   void __user *buffer, size_t *lenp,
6024 				   loff_t *ppos)
6025 {
6026 	int *valp = ctl->data;
6027 	int val = *valp;
6028 	loff_t pos = *ppos;
6029 	struct ctl_table lctl;
6030 	int ret;
6031 
6032 	lctl = *ctl;
6033 	lctl.data = &val;
6034 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6035 
6036 	if (write && (*valp != val))
6037 		ret = addrconf_disable_policy(ctl, valp, val);
6038 
6039 	if (ret)
6040 		*ppos = pos;
6041 
6042 	return ret;
6043 }
6044 
6045 static int minus_one = -1;
6046 static const int zero = 0;
6047 static const int one = 1;
6048 static const int two_five_five = 255;
6049 
6050 static const struct ctl_table addrconf_sysctl[] = {
6051 	{
6052 		.procname	= "forwarding",
6053 		.data		= &ipv6_devconf.forwarding,
6054 		.maxlen		= sizeof(int),
6055 		.mode		= 0644,
6056 		.proc_handler	= addrconf_sysctl_forward,
6057 	},
6058 	{
6059 		.procname	= "hop_limit",
6060 		.data		= &ipv6_devconf.hop_limit,
6061 		.maxlen		= sizeof(int),
6062 		.mode		= 0644,
6063 		.proc_handler	= proc_dointvec_minmax,
6064 		.extra1		= (void *)&one,
6065 		.extra2		= (void *)&two_five_five,
6066 	},
6067 	{
6068 		.procname	= "mtu",
6069 		.data		= &ipv6_devconf.mtu6,
6070 		.maxlen		= sizeof(int),
6071 		.mode		= 0644,
6072 		.proc_handler	= addrconf_sysctl_mtu,
6073 	},
6074 	{
6075 		.procname	= "accept_ra",
6076 		.data		= &ipv6_devconf.accept_ra,
6077 		.maxlen		= sizeof(int),
6078 		.mode		= 0644,
6079 		.proc_handler	= proc_dointvec,
6080 	},
6081 	{
6082 		.procname	= "accept_redirects",
6083 		.data		= &ipv6_devconf.accept_redirects,
6084 		.maxlen		= sizeof(int),
6085 		.mode		= 0644,
6086 		.proc_handler	= proc_dointvec,
6087 	},
6088 	{
6089 		.procname	= "autoconf",
6090 		.data		= &ipv6_devconf.autoconf,
6091 		.maxlen		= sizeof(int),
6092 		.mode		= 0644,
6093 		.proc_handler	= proc_dointvec,
6094 	},
6095 	{
6096 		.procname	= "dad_transmits",
6097 		.data		= &ipv6_devconf.dad_transmits,
6098 		.maxlen		= sizeof(int),
6099 		.mode		= 0644,
6100 		.proc_handler	= proc_dointvec,
6101 	},
6102 	{
6103 		.procname	= "router_solicitations",
6104 		.data		= &ipv6_devconf.rtr_solicits,
6105 		.maxlen		= sizeof(int),
6106 		.mode		= 0644,
6107 		.proc_handler	= proc_dointvec_minmax,
6108 		.extra1		= &minus_one,
6109 	},
6110 	{
6111 		.procname	= "router_solicitation_interval",
6112 		.data		= &ipv6_devconf.rtr_solicit_interval,
6113 		.maxlen		= sizeof(int),
6114 		.mode		= 0644,
6115 		.proc_handler	= proc_dointvec_jiffies,
6116 	},
6117 	{
6118 		.procname	= "router_solicitation_max_interval",
6119 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6120 		.maxlen		= sizeof(int),
6121 		.mode		= 0644,
6122 		.proc_handler	= proc_dointvec_jiffies,
6123 	},
6124 	{
6125 		.procname	= "router_solicitation_delay",
6126 		.data		= &ipv6_devconf.rtr_solicit_delay,
6127 		.maxlen		= sizeof(int),
6128 		.mode		= 0644,
6129 		.proc_handler	= proc_dointvec_jiffies,
6130 	},
6131 	{
6132 		.procname	= "force_mld_version",
6133 		.data		= &ipv6_devconf.force_mld_version,
6134 		.maxlen		= sizeof(int),
6135 		.mode		= 0644,
6136 		.proc_handler	= proc_dointvec,
6137 	},
6138 	{
6139 		.procname	= "mldv1_unsolicited_report_interval",
6140 		.data		=
6141 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6142 		.maxlen		= sizeof(int),
6143 		.mode		= 0644,
6144 		.proc_handler	= proc_dointvec_ms_jiffies,
6145 	},
6146 	{
6147 		.procname	= "mldv2_unsolicited_report_interval",
6148 		.data		=
6149 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6150 		.maxlen		= sizeof(int),
6151 		.mode		= 0644,
6152 		.proc_handler	= proc_dointvec_ms_jiffies,
6153 	},
6154 	{
6155 		.procname	= "use_tempaddr",
6156 		.data		= &ipv6_devconf.use_tempaddr,
6157 		.maxlen		= sizeof(int),
6158 		.mode		= 0644,
6159 		.proc_handler	= proc_dointvec,
6160 	},
6161 	{
6162 		.procname	= "temp_valid_lft",
6163 		.data		= &ipv6_devconf.temp_valid_lft,
6164 		.maxlen		= sizeof(int),
6165 		.mode		= 0644,
6166 		.proc_handler	= proc_dointvec,
6167 	},
6168 	{
6169 		.procname	= "temp_prefered_lft",
6170 		.data		= &ipv6_devconf.temp_prefered_lft,
6171 		.maxlen		= sizeof(int),
6172 		.mode		= 0644,
6173 		.proc_handler	= proc_dointvec,
6174 	},
6175 	{
6176 		.procname	= "regen_max_retry",
6177 		.data		= &ipv6_devconf.regen_max_retry,
6178 		.maxlen		= sizeof(int),
6179 		.mode		= 0644,
6180 		.proc_handler	= proc_dointvec,
6181 	},
6182 	{
6183 		.procname	= "max_desync_factor",
6184 		.data		= &ipv6_devconf.max_desync_factor,
6185 		.maxlen		= sizeof(int),
6186 		.mode		= 0644,
6187 		.proc_handler	= proc_dointvec,
6188 	},
6189 	{
6190 		.procname	= "max_addresses",
6191 		.data		= &ipv6_devconf.max_addresses,
6192 		.maxlen		= sizeof(int),
6193 		.mode		= 0644,
6194 		.proc_handler	= proc_dointvec,
6195 	},
6196 	{
6197 		.procname	= "accept_ra_defrtr",
6198 		.data		= &ipv6_devconf.accept_ra_defrtr,
6199 		.maxlen		= sizeof(int),
6200 		.mode		= 0644,
6201 		.proc_handler	= proc_dointvec,
6202 	},
6203 	{
6204 		.procname	= "accept_ra_min_hop_limit",
6205 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6206 		.maxlen		= sizeof(int),
6207 		.mode		= 0644,
6208 		.proc_handler	= proc_dointvec,
6209 	},
6210 	{
6211 		.procname	= "accept_ra_pinfo",
6212 		.data		= &ipv6_devconf.accept_ra_pinfo,
6213 		.maxlen		= sizeof(int),
6214 		.mode		= 0644,
6215 		.proc_handler	= proc_dointvec,
6216 	},
6217 #ifdef CONFIG_IPV6_ROUTER_PREF
6218 	{
6219 		.procname	= "accept_ra_rtr_pref",
6220 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6221 		.maxlen		= sizeof(int),
6222 		.mode		= 0644,
6223 		.proc_handler	= proc_dointvec,
6224 	},
6225 	{
6226 		.procname	= "router_probe_interval",
6227 		.data		= &ipv6_devconf.rtr_probe_interval,
6228 		.maxlen		= sizeof(int),
6229 		.mode		= 0644,
6230 		.proc_handler	= proc_dointvec_jiffies,
6231 	},
6232 #ifdef CONFIG_IPV6_ROUTE_INFO
6233 	{
6234 		.procname	= "accept_ra_rt_info_min_plen",
6235 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6236 		.maxlen		= sizeof(int),
6237 		.mode		= 0644,
6238 		.proc_handler	= proc_dointvec,
6239 	},
6240 	{
6241 		.procname	= "accept_ra_rt_info_max_plen",
6242 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6243 		.maxlen		= sizeof(int),
6244 		.mode		= 0644,
6245 		.proc_handler	= proc_dointvec,
6246 	},
6247 #endif
6248 #endif
6249 	{
6250 		.procname	= "proxy_ndp",
6251 		.data		= &ipv6_devconf.proxy_ndp,
6252 		.maxlen		= sizeof(int),
6253 		.mode		= 0644,
6254 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6255 	},
6256 	{
6257 		.procname	= "accept_source_route",
6258 		.data		= &ipv6_devconf.accept_source_route,
6259 		.maxlen		= sizeof(int),
6260 		.mode		= 0644,
6261 		.proc_handler	= proc_dointvec,
6262 	},
6263 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6264 	{
6265 		.procname	= "optimistic_dad",
6266 		.data		= &ipv6_devconf.optimistic_dad,
6267 		.maxlen		= sizeof(int),
6268 		.mode		= 0644,
6269 		.proc_handler   = proc_dointvec,
6270 	},
6271 	{
6272 		.procname	= "use_optimistic",
6273 		.data		= &ipv6_devconf.use_optimistic,
6274 		.maxlen		= sizeof(int),
6275 		.mode		= 0644,
6276 		.proc_handler	= proc_dointvec,
6277 	},
6278 #endif
6279 #ifdef CONFIG_IPV6_MROUTE
6280 	{
6281 		.procname	= "mc_forwarding",
6282 		.data		= &ipv6_devconf.mc_forwarding,
6283 		.maxlen		= sizeof(int),
6284 		.mode		= 0444,
6285 		.proc_handler	= proc_dointvec,
6286 	},
6287 #endif
6288 	{
6289 		.procname	= "disable_ipv6",
6290 		.data		= &ipv6_devconf.disable_ipv6,
6291 		.maxlen		= sizeof(int),
6292 		.mode		= 0644,
6293 		.proc_handler	= addrconf_sysctl_disable,
6294 	},
6295 	{
6296 		.procname	= "accept_dad",
6297 		.data		= &ipv6_devconf.accept_dad,
6298 		.maxlen		= sizeof(int),
6299 		.mode		= 0644,
6300 		.proc_handler	= proc_dointvec,
6301 	},
6302 	{
6303 		.procname	= "force_tllao",
6304 		.data		= &ipv6_devconf.force_tllao,
6305 		.maxlen		= sizeof(int),
6306 		.mode		= 0644,
6307 		.proc_handler	= proc_dointvec
6308 	},
6309 	{
6310 		.procname	= "ndisc_notify",
6311 		.data		= &ipv6_devconf.ndisc_notify,
6312 		.maxlen		= sizeof(int),
6313 		.mode		= 0644,
6314 		.proc_handler	= proc_dointvec
6315 	},
6316 	{
6317 		.procname	= "suppress_frag_ndisc",
6318 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6319 		.maxlen		= sizeof(int),
6320 		.mode		= 0644,
6321 		.proc_handler	= proc_dointvec
6322 	},
6323 	{
6324 		.procname	= "accept_ra_from_local",
6325 		.data		= &ipv6_devconf.accept_ra_from_local,
6326 		.maxlen		= sizeof(int),
6327 		.mode		= 0644,
6328 		.proc_handler	= proc_dointvec,
6329 	},
6330 	{
6331 		.procname	= "accept_ra_mtu",
6332 		.data		= &ipv6_devconf.accept_ra_mtu,
6333 		.maxlen		= sizeof(int),
6334 		.mode		= 0644,
6335 		.proc_handler	= proc_dointvec,
6336 	},
6337 	{
6338 		.procname	= "stable_secret",
6339 		.data		= &ipv6_devconf.stable_secret,
6340 		.maxlen		= IPV6_MAX_STRLEN,
6341 		.mode		= 0600,
6342 		.proc_handler	= addrconf_sysctl_stable_secret,
6343 	},
6344 	{
6345 		.procname	= "use_oif_addrs_only",
6346 		.data		= &ipv6_devconf.use_oif_addrs_only,
6347 		.maxlen		= sizeof(int),
6348 		.mode		= 0644,
6349 		.proc_handler	= proc_dointvec,
6350 	},
6351 	{
6352 		.procname	= "ignore_routes_with_linkdown",
6353 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6354 		.maxlen		= sizeof(int),
6355 		.mode		= 0644,
6356 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6357 	},
6358 	{
6359 		.procname	= "drop_unicast_in_l2_multicast",
6360 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6361 		.maxlen		= sizeof(int),
6362 		.mode		= 0644,
6363 		.proc_handler	= proc_dointvec,
6364 	},
6365 	{
6366 		.procname	= "drop_unsolicited_na",
6367 		.data		= &ipv6_devconf.drop_unsolicited_na,
6368 		.maxlen		= sizeof(int),
6369 		.mode		= 0644,
6370 		.proc_handler	= proc_dointvec,
6371 	},
6372 	{
6373 		.procname	= "keep_addr_on_down",
6374 		.data		= &ipv6_devconf.keep_addr_on_down,
6375 		.maxlen		= sizeof(int),
6376 		.mode		= 0644,
6377 		.proc_handler	= proc_dointvec,
6378 
6379 	},
6380 	{
6381 		.procname	= "seg6_enabled",
6382 		.data		= &ipv6_devconf.seg6_enabled,
6383 		.maxlen		= sizeof(int),
6384 		.mode		= 0644,
6385 		.proc_handler	= proc_dointvec,
6386 	},
6387 #ifdef CONFIG_IPV6_SEG6_HMAC
6388 	{
6389 		.procname	= "seg6_require_hmac",
6390 		.data		= &ipv6_devconf.seg6_require_hmac,
6391 		.maxlen		= sizeof(int),
6392 		.mode		= 0644,
6393 		.proc_handler	= proc_dointvec,
6394 	},
6395 #endif
6396 	{
6397 		.procname       = "enhanced_dad",
6398 		.data           = &ipv6_devconf.enhanced_dad,
6399 		.maxlen         = sizeof(int),
6400 		.mode           = 0644,
6401 		.proc_handler   = proc_dointvec,
6402 	},
6403 	{
6404 		.procname		= "addr_gen_mode",
6405 		.data			= &ipv6_devconf.addr_gen_mode,
6406 		.maxlen			= sizeof(int),
6407 		.mode			= 0644,
6408 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6409 	},
6410 	{
6411 		.procname       = "disable_policy",
6412 		.data           = &ipv6_devconf.disable_policy,
6413 		.maxlen         = sizeof(int),
6414 		.mode           = 0644,
6415 		.proc_handler   = addrconf_sysctl_disable_policy,
6416 	},
6417 	{
6418 		.procname	= "ndisc_tclass",
6419 		.data		= &ipv6_devconf.ndisc_tclass,
6420 		.maxlen		= sizeof(int),
6421 		.mode		= 0644,
6422 		.proc_handler	= proc_dointvec_minmax,
6423 		.extra1		= (void *)&zero,
6424 		.extra2		= (void *)&two_five_five,
6425 	},
6426 	{
6427 		/* sentinel */
6428 	}
6429 };
6430 
6431 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6432 		struct inet6_dev *idev, struct ipv6_devconf *p)
6433 {
6434 	int i, ifindex;
6435 	struct ctl_table *table;
6436 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6437 
6438 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6439 	if (!table)
6440 		goto out;
6441 
6442 	for (i = 0; table[i].data; i++) {
6443 		table[i].data += (char *)p - (char *)&ipv6_devconf;
6444 		/* If one of these is already set, then it is not safe to
6445 		 * overwrite either of them: this makes proc_dointvec_minmax
6446 		 * usable.
6447 		 */
6448 		if (!table[i].extra1 && !table[i].extra2) {
6449 			table[i].extra1 = idev; /* embedded; no ref */
6450 			table[i].extra2 = net;
6451 		}
6452 	}
6453 
6454 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6455 
6456 	p->sysctl_header = register_net_sysctl(net, path, table);
6457 	if (!p->sysctl_header)
6458 		goto free;
6459 
6460 	if (!strcmp(dev_name, "all"))
6461 		ifindex = NETCONFA_IFINDEX_ALL;
6462 	else if (!strcmp(dev_name, "default"))
6463 		ifindex = NETCONFA_IFINDEX_DEFAULT;
6464 	else
6465 		ifindex = idev->dev->ifindex;
6466 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6467 				     ifindex, p);
6468 	return 0;
6469 
6470 free:
6471 	kfree(table);
6472 out:
6473 	return -ENOBUFS;
6474 }
6475 
6476 static void __addrconf_sysctl_unregister(struct net *net,
6477 					 struct ipv6_devconf *p, int ifindex)
6478 {
6479 	struct ctl_table *table;
6480 
6481 	if (!p->sysctl_header)
6482 		return;
6483 
6484 	table = p->sysctl_header->ctl_table_arg;
6485 	unregister_net_sysctl_table(p->sysctl_header);
6486 	p->sysctl_header = NULL;
6487 	kfree(table);
6488 
6489 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6490 }
6491 
6492 static int addrconf_sysctl_register(struct inet6_dev *idev)
6493 {
6494 	int err;
6495 
6496 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
6497 		return -EINVAL;
6498 
6499 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6500 				    &ndisc_ifinfo_sysctl_change);
6501 	if (err)
6502 		return err;
6503 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6504 					 idev, &idev->cnf);
6505 	if (err)
6506 		neigh_sysctl_unregister(idev->nd_parms);
6507 
6508 	return err;
6509 }
6510 
6511 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6512 {
6513 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6514 				     idev->dev->ifindex);
6515 	neigh_sysctl_unregister(idev->nd_parms);
6516 }
6517 
6518 
6519 #endif
6520 
6521 static int __net_init addrconf_init_net(struct net *net)
6522 {
6523 	int err = -ENOMEM;
6524 	struct ipv6_devconf *all, *dflt;
6525 
6526 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6527 	if (!all)
6528 		goto err_alloc_all;
6529 
6530 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6531 	if (!dflt)
6532 		goto err_alloc_dflt;
6533 
6534 	/* these will be inherited by all namespaces */
6535 	dflt->autoconf = ipv6_defaults.autoconf;
6536 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6537 
6538 	dflt->stable_secret.initialized = false;
6539 	all->stable_secret.initialized = false;
6540 
6541 	net->ipv6.devconf_all = all;
6542 	net->ipv6.devconf_dflt = dflt;
6543 
6544 #ifdef CONFIG_SYSCTL
6545 	err = __addrconf_sysctl_register(net, "all", NULL, all);
6546 	if (err < 0)
6547 		goto err_reg_all;
6548 
6549 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6550 	if (err < 0)
6551 		goto err_reg_dflt;
6552 #endif
6553 	return 0;
6554 
6555 #ifdef CONFIG_SYSCTL
6556 err_reg_dflt:
6557 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6558 err_reg_all:
6559 	kfree(dflt);
6560 #endif
6561 err_alloc_dflt:
6562 	kfree(all);
6563 err_alloc_all:
6564 	return err;
6565 }
6566 
6567 static void __net_exit addrconf_exit_net(struct net *net)
6568 {
6569 #ifdef CONFIG_SYSCTL
6570 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6571 				     NETCONFA_IFINDEX_DEFAULT);
6572 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6573 				     NETCONFA_IFINDEX_ALL);
6574 #endif
6575 	kfree(net->ipv6.devconf_dflt);
6576 	kfree(net->ipv6.devconf_all);
6577 }
6578 
6579 static struct pernet_operations addrconf_ops = {
6580 	.init = addrconf_init_net,
6581 	.exit = addrconf_exit_net,
6582 };
6583 
6584 static struct rtnl_af_ops inet6_ops __read_mostly = {
6585 	.family		  = AF_INET6,
6586 	.fill_link_af	  = inet6_fill_link_af,
6587 	.get_link_af_size = inet6_get_link_af_size,
6588 	.validate_link_af = inet6_validate_link_af,
6589 	.set_link_af	  = inet6_set_link_af,
6590 };
6591 
6592 /*
6593  *	Init / cleanup code
6594  */
6595 
6596 int __init addrconf_init(void)
6597 {
6598 	struct inet6_dev *idev;
6599 	int i, err;
6600 
6601 	err = ipv6_addr_label_init();
6602 	if (err < 0) {
6603 		pr_crit("%s: cannot initialize default policy table: %d\n",
6604 			__func__, err);
6605 		goto out;
6606 	}
6607 
6608 	err = register_pernet_subsys(&addrconf_ops);
6609 	if (err < 0)
6610 		goto out_addrlabel;
6611 
6612 	addrconf_wq = create_workqueue("ipv6_addrconf");
6613 	if (!addrconf_wq) {
6614 		err = -ENOMEM;
6615 		goto out_nowq;
6616 	}
6617 
6618 	/* The addrconf netdev notifier requires that loopback_dev
6619 	 * has it's ipv6 private information allocated and setup
6620 	 * before it can bring up and give link-local addresses
6621 	 * to other devices which are up.
6622 	 *
6623 	 * Unfortunately, loopback_dev is not necessarily the first
6624 	 * entry in the global dev_base list of net devices.  In fact,
6625 	 * it is likely to be the very last entry on that list.
6626 	 * So this causes the notifier registry below to try and
6627 	 * give link-local addresses to all devices besides loopback_dev
6628 	 * first, then loopback_dev, which cases all the non-loopback_dev
6629 	 * devices to fail to get a link-local address.
6630 	 *
6631 	 * So, as a temporary fix, allocate the ipv6 structure for
6632 	 * loopback_dev first by hand.
6633 	 * Longer term, all of the dependencies ipv6 has upon the loopback
6634 	 * device and it being up should be removed.
6635 	 */
6636 	rtnl_lock();
6637 	idev = ipv6_add_dev(init_net.loopback_dev);
6638 	rtnl_unlock();
6639 	if (IS_ERR(idev)) {
6640 		err = PTR_ERR(idev);
6641 		goto errlo;
6642 	}
6643 
6644 	ip6_route_init_special_entries();
6645 
6646 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
6647 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6648 
6649 	register_netdevice_notifier(&ipv6_dev_notf);
6650 
6651 	addrconf_verify();
6652 
6653 	rtnl_af_register(&inet6_ops);
6654 
6655 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6656 				   NULL, inet6_dump_ifinfo, 0);
6657 	if (err < 0)
6658 		goto errout;
6659 
6660 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6661 				   inet6_rtm_newaddr, NULL, 0);
6662 	if (err < 0)
6663 		goto errout;
6664 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6665 				   inet6_rtm_deladdr, NULL, 0);
6666 	if (err < 0)
6667 		goto errout;
6668 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6669 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
6670 				   RTNL_FLAG_DOIT_UNLOCKED);
6671 	if (err < 0)
6672 		goto errout;
6673 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6674 				   NULL, inet6_dump_ifmcaddr, 0);
6675 	if (err < 0)
6676 		goto errout;
6677 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6678 				   NULL, inet6_dump_ifacaddr, 0);
6679 	if (err < 0)
6680 		goto errout;
6681 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6682 				   inet6_netconf_get_devconf,
6683 				   inet6_netconf_dump_devconf,
6684 				   RTNL_FLAG_DOIT_UNLOCKED);
6685 	if (err < 0)
6686 		goto errout;
6687 	err = ipv6_addr_label_rtnl_register();
6688 	if (err < 0)
6689 		goto errout;
6690 
6691 	return 0;
6692 errout:
6693 	rtnl_unregister_all(PF_INET6);
6694 	rtnl_af_unregister(&inet6_ops);
6695 	unregister_netdevice_notifier(&ipv6_dev_notf);
6696 errlo:
6697 	destroy_workqueue(addrconf_wq);
6698 out_nowq:
6699 	unregister_pernet_subsys(&addrconf_ops);
6700 out_addrlabel:
6701 	ipv6_addr_label_cleanup();
6702 out:
6703 	return err;
6704 }
6705 
6706 void addrconf_cleanup(void)
6707 {
6708 	struct net_device *dev;
6709 	int i;
6710 
6711 	unregister_netdevice_notifier(&ipv6_dev_notf);
6712 	unregister_pernet_subsys(&addrconf_ops);
6713 	ipv6_addr_label_cleanup();
6714 
6715 	rtnl_af_unregister(&inet6_ops);
6716 
6717 	rtnl_lock();
6718 
6719 	/* clean dev list */
6720 	for_each_netdev(&init_net, dev) {
6721 		if (__in6_dev_get(dev) == NULL)
6722 			continue;
6723 		addrconf_ifdown(dev, 1);
6724 	}
6725 	addrconf_ifdown(init_net.loopback_dev, 2);
6726 
6727 	/*
6728 	 *	Check hash table.
6729 	 */
6730 	spin_lock_bh(&addrconf_hash_lock);
6731 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
6732 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6733 	spin_unlock_bh(&addrconf_hash_lock);
6734 	cancel_delayed_work(&addr_chk_work);
6735 	rtnl_unlock();
6736 
6737 	destroy_workqueue(addrconf_wq);
6738 }
6739