xref: /openbmc/linux/net/ipv6/addrconf.c (revision 6d99a79c)
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 fib6_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 
389 	ndev->cnf.mtu6 = dev->mtu;
390 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
391 	if (!ndev->nd_parms) {
392 		kfree(ndev);
393 		return ERR_PTR(err);
394 	}
395 	if (ndev->cnf.forwarding)
396 		dev_disable_lro(dev);
397 	/* We refer to the device */
398 	dev_hold(dev);
399 
400 	if (snmp6_alloc_dev(ndev) < 0) {
401 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
402 			   __func__);
403 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
404 		dev_put(dev);
405 		kfree(ndev);
406 		return ERR_PTR(err);
407 	}
408 
409 	if (snmp6_register_dev(ndev) < 0) {
410 		netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
411 			   __func__, dev->name);
412 		goto err_release;
413 	}
414 
415 	/* One reference from device. */
416 	refcount_set(&ndev->refcnt, 1);
417 
418 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
419 		ndev->cnf.accept_dad = -1;
420 
421 #if IS_ENABLED(CONFIG_IPV6_SIT)
422 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
423 		pr_info("%s: Disabled Multicast RS\n", dev->name);
424 		ndev->cnf.rtr_solicits = 0;
425 	}
426 #endif
427 
428 	INIT_LIST_HEAD(&ndev->tempaddr_list);
429 	ndev->desync_factor = U32_MAX;
430 	if ((dev->flags&IFF_LOOPBACK) ||
431 	    dev->type == ARPHRD_TUNNEL ||
432 	    dev->type == ARPHRD_TUNNEL6 ||
433 	    dev->type == ARPHRD_SIT ||
434 	    dev->type == ARPHRD_NONE) {
435 		ndev->cnf.use_tempaddr = -1;
436 	} else
437 		ipv6_regen_rndid(ndev);
438 
439 	ndev->token = in6addr_any;
440 
441 	if (netif_running(dev) && addrconf_link_ready(dev))
442 		ndev->if_flags |= IF_READY;
443 
444 	ipv6_mc_init_dev(ndev);
445 	ndev->tstamp = jiffies;
446 	err = addrconf_sysctl_register(ndev);
447 	if (err) {
448 		ipv6_mc_destroy_dev(ndev);
449 		snmp6_unregister_dev(ndev);
450 		goto err_release;
451 	}
452 	/* protected by rtnl_lock */
453 	rcu_assign_pointer(dev->ip6_ptr, ndev);
454 
455 	/* Join interface-local all-node multicast group */
456 	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
457 
458 	/* Join all-node multicast group */
459 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
460 
461 	/* Join all-router multicast group if forwarding is set */
462 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
463 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
464 
465 	return ndev;
466 
467 err_release:
468 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
469 	ndev->dead = 1;
470 	in6_dev_finish_destroy(ndev);
471 	return ERR_PTR(err);
472 }
473 
474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
475 {
476 	struct inet6_dev *idev;
477 
478 	ASSERT_RTNL();
479 
480 	idev = __in6_dev_get(dev);
481 	if (!idev) {
482 		idev = ipv6_add_dev(dev);
483 		if (IS_ERR(idev))
484 			return NULL;
485 	}
486 
487 	if (dev->flags&IFF_UP)
488 		ipv6_mc_up(idev);
489 	return idev;
490 }
491 
492 static int inet6_netconf_msgsize_devconf(int type)
493 {
494 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
495 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
496 	bool all = false;
497 
498 	if (type == NETCONFA_ALL)
499 		all = true;
500 
501 	if (all || type == NETCONFA_FORWARDING)
502 		size += nla_total_size(4);
503 #ifdef CONFIG_IPV6_MROUTE
504 	if (all || type == NETCONFA_MC_FORWARDING)
505 		size += nla_total_size(4);
506 #endif
507 	if (all || type == NETCONFA_PROXY_NEIGH)
508 		size += nla_total_size(4);
509 
510 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
511 		size += nla_total_size(4);
512 
513 	return size;
514 }
515 
516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
517 				      struct ipv6_devconf *devconf, u32 portid,
518 				      u32 seq, int event, unsigned int flags,
519 				      int type)
520 {
521 	struct nlmsghdr  *nlh;
522 	struct netconfmsg *ncm;
523 	bool all = false;
524 
525 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
526 			flags);
527 	if (!nlh)
528 		return -EMSGSIZE;
529 
530 	if (type == NETCONFA_ALL)
531 		all = true;
532 
533 	ncm = nlmsg_data(nlh);
534 	ncm->ncm_family = AF_INET6;
535 
536 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
537 		goto nla_put_failure;
538 
539 	if (!devconf)
540 		goto out;
541 
542 	if ((all || type == NETCONFA_FORWARDING) &&
543 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
544 		goto nla_put_failure;
545 #ifdef CONFIG_IPV6_MROUTE
546 	if ((all || type == NETCONFA_MC_FORWARDING) &&
547 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
548 			devconf->mc_forwarding) < 0)
549 		goto nla_put_failure;
550 #endif
551 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
552 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
553 		goto nla_put_failure;
554 
555 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
556 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
557 			devconf->ignore_routes_with_linkdown) < 0)
558 		goto nla_put_failure;
559 
560 out:
561 	nlmsg_end(skb, nlh);
562 	return 0;
563 
564 nla_put_failure:
565 	nlmsg_cancel(skb, nlh);
566 	return -EMSGSIZE;
567 }
568 
569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
570 				  int ifindex, struct ipv6_devconf *devconf)
571 {
572 	struct sk_buff *skb;
573 	int err = -ENOBUFS;
574 
575 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
576 	if (!skb)
577 		goto errout;
578 
579 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
580 					 event, 0, type);
581 	if (err < 0) {
582 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
583 		WARN_ON(err == -EMSGSIZE);
584 		kfree_skb(skb);
585 		goto errout;
586 	}
587 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
588 	return;
589 errout:
590 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
591 }
592 
593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
594 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
595 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
596 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
597 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
598 };
599 
600 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
601 				     struct nlmsghdr *nlh,
602 				     struct netlink_ext_ack *extack)
603 {
604 	struct net *net = sock_net(in_skb->sk);
605 	struct nlattr *tb[NETCONFA_MAX+1];
606 	struct inet6_dev *in6_dev = NULL;
607 	struct net_device *dev = NULL;
608 	struct netconfmsg *ncm;
609 	struct sk_buff *skb;
610 	struct ipv6_devconf *devconf;
611 	int ifindex;
612 	int err;
613 
614 	err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
615 			  devconf_ipv6_policy, extack);
616 	if (err < 0)
617 		return err;
618 
619 	if (!tb[NETCONFA_IFINDEX])
620 		return -EINVAL;
621 
622 	err = -EINVAL;
623 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
624 	switch (ifindex) {
625 	case NETCONFA_IFINDEX_ALL:
626 		devconf = net->ipv6.devconf_all;
627 		break;
628 	case NETCONFA_IFINDEX_DEFAULT:
629 		devconf = net->ipv6.devconf_dflt;
630 		break;
631 	default:
632 		dev = dev_get_by_index(net, ifindex);
633 		if (!dev)
634 			return -EINVAL;
635 		in6_dev = in6_dev_get(dev);
636 		if (!in6_dev)
637 			goto errout;
638 		devconf = &in6_dev->cnf;
639 		break;
640 	}
641 
642 	err = -ENOBUFS;
643 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
644 	if (!skb)
645 		goto errout;
646 
647 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
648 					 NETLINK_CB(in_skb).portid,
649 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
650 					 NETCONFA_ALL);
651 	if (err < 0) {
652 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
653 		WARN_ON(err == -EMSGSIZE);
654 		kfree_skb(skb);
655 		goto errout;
656 	}
657 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
658 errout:
659 	if (in6_dev)
660 		in6_dev_put(in6_dev);
661 	if (dev)
662 		dev_put(dev);
663 	return err;
664 }
665 
666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
667 				      struct netlink_callback *cb)
668 {
669 	const struct nlmsghdr *nlh = cb->nlh;
670 	struct net *net = sock_net(skb->sk);
671 	int h, s_h;
672 	int idx, s_idx;
673 	struct net_device *dev;
674 	struct inet6_dev *idev;
675 	struct hlist_head *head;
676 
677 	if (cb->strict_check) {
678 		struct netlink_ext_ack *extack = cb->extack;
679 		struct netconfmsg *ncm;
680 
681 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
682 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
683 			return -EINVAL;
684 		}
685 
686 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
687 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
688 			return -EINVAL;
689 		}
690 	}
691 
692 	s_h = cb->args[0];
693 	s_idx = idx = cb->args[1];
694 
695 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
696 		idx = 0;
697 		head = &net->dev_index_head[h];
698 		rcu_read_lock();
699 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
700 			  net->dev_base_seq;
701 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
702 			if (idx < s_idx)
703 				goto cont;
704 			idev = __in6_dev_get(dev);
705 			if (!idev)
706 				goto cont;
707 
708 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
709 						       &idev->cnf,
710 						       NETLINK_CB(cb->skb).portid,
711 						       nlh->nlmsg_seq,
712 						       RTM_NEWNETCONF,
713 						       NLM_F_MULTI,
714 						       NETCONFA_ALL) < 0) {
715 				rcu_read_unlock();
716 				goto done;
717 			}
718 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
719 cont:
720 			idx++;
721 		}
722 		rcu_read_unlock();
723 	}
724 	if (h == NETDEV_HASHENTRIES) {
725 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
726 					       net->ipv6.devconf_all,
727 					       NETLINK_CB(cb->skb).portid,
728 					       nlh->nlmsg_seq,
729 					       RTM_NEWNETCONF, NLM_F_MULTI,
730 					       NETCONFA_ALL) < 0)
731 			goto done;
732 		else
733 			h++;
734 	}
735 	if (h == NETDEV_HASHENTRIES + 1) {
736 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
737 					       net->ipv6.devconf_dflt,
738 					       NETLINK_CB(cb->skb).portid,
739 					       nlh->nlmsg_seq,
740 					       RTM_NEWNETCONF, NLM_F_MULTI,
741 					       NETCONFA_ALL) < 0)
742 			goto done;
743 		else
744 			h++;
745 	}
746 done:
747 	cb->args[0] = h;
748 	cb->args[1] = idx;
749 
750 	return skb->len;
751 }
752 
753 #ifdef CONFIG_SYSCTL
754 static void dev_forward_change(struct inet6_dev *idev)
755 {
756 	struct net_device *dev;
757 	struct inet6_ifaddr *ifa;
758 
759 	if (!idev)
760 		return;
761 	dev = idev->dev;
762 	if (idev->cnf.forwarding)
763 		dev_disable_lro(dev);
764 	if (dev->flags & IFF_MULTICAST) {
765 		if (idev->cnf.forwarding) {
766 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
767 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
768 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
769 		} else {
770 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
771 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
772 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
773 		}
774 	}
775 
776 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
777 		if (ifa->flags&IFA_F_TENTATIVE)
778 			continue;
779 		if (idev->cnf.forwarding)
780 			addrconf_join_anycast(ifa);
781 		else
782 			addrconf_leave_anycast(ifa);
783 	}
784 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
785 				     NETCONFA_FORWARDING,
786 				     dev->ifindex, &idev->cnf);
787 }
788 
789 
790 static void addrconf_forward_change(struct net *net, __s32 newf)
791 {
792 	struct net_device *dev;
793 	struct inet6_dev *idev;
794 
795 	for_each_netdev(net, dev) {
796 		idev = __in6_dev_get(dev);
797 		if (idev) {
798 			int changed = (!idev->cnf.forwarding) ^ (!newf);
799 			idev->cnf.forwarding = newf;
800 			if (changed)
801 				dev_forward_change(idev);
802 		}
803 	}
804 }
805 
806 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
807 {
808 	struct net *net;
809 	int old;
810 
811 	if (!rtnl_trylock())
812 		return restart_syscall();
813 
814 	net = (struct net *)table->extra2;
815 	old = *p;
816 	*p = newf;
817 
818 	if (p == &net->ipv6.devconf_dflt->forwarding) {
819 		if ((!newf) ^ (!old))
820 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
821 						     NETCONFA_FORWARDING,
822 						     NETCONFA_IFINDEX_DEFAULT,
823 						     net->ipv6.devconf_dflt);
824 		rtnl_unlock();
825 		return 0;
826 	}
827 
828 	if (p == &net->ipv6.devconf_all->forwarding) {
829 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
830 
831 		net->ipv6.devconf_dflt->forwarding = newf;
832 		if ((!newf) ^ (!old_dflt))
833 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
834 						     NETCONFA_FORWARDING,
835 						     NETCONFA_IFINDEX_DEFAULT,
836 						     net->ipv6.devconf_dflt);
837 
838 		addrconf_forward_change(net, newf);
839 		if ((!newf) ^ (!old))
840 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
841 						     NETCONFA_FORWARDING,
842 						     NETCONFA_IFINDEX_ALL,
843 						     net->ipv6.devconf_all);
844 	} else if ((!newf) ^ (!old))
845 		dev_forward_change((struct inet6_dev *)table->extra1);
846 	rtnl_unlock();
847 
848 	if (newf)
849 		rt6_purge_dflt_routers(net);
850 	return 1;
851 }
852 
853 static void addrconf_linkdown_change(struct net *net, __s32 newf)
854 {
855 	struct net_device *dev;
856 	struct inet6_dev *idev;
857 
858 	for_each_netdev(net, dev) {
859 		idev = __in6_dev_get(dev);
860 		if (idev) {
861 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
862 
863 			idev->cnf.ignore_routes_with_linkdown = newf;
864 			if (changed)
865 				inet6_netconf_notify_devconf(dev_net(dev),
866 							     RTM_NEWNETCONF,
867 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
868 							     dev->ifindex,
869 							     &idev->cnf);
870 		}
871 	}
872 }
873 
874 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
875 {
876 	struct net *net;
877 	int old;
878 
879 	if (!rtnl_trylock())
880 		return restart_syscall();
881 
882 	net = (struct net *)table->extra2;
883 	old = *p;
884 	*p = newf;
885 
886 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
887 		if ((!newf) ^ (!old))
888 			inet6_netconf_notify_devconf(net,
889 						     RTM_NEWNETCONF,
890 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
891 						     NETCONFA_IFINDEX_DEFAULT,
892 						     net->ipv6.devconf_dflt);
893 		rtnl_unlock();
894 		return 0;
895 	}
896 
897 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
898 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
899 		addrconf_linkdown_change(net, newf);
900 		if ((!newf) ^ (!old))
901 			inet6_netconf_notify_devconf(net,
902 						     RTM_NEWNETCONF,
903 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
904 						     NETCONFA_IFINDEX_ALL,
905 						     net->ipv6.devconf_all);
906 	}
907 	rtnl_unlock();
908 
909 	return 1;
910 }
911 
912 #endif
913 
914 /* Nobody refers to this ifaddr, destroy it */
915 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
916 {
917 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
918 
919 #ifdef NET_REFCNT_DEBUG
920 	pr_debug("%s\n", __func__);
921 #endif
922 
923 	in6_dev_put(ifp->idev);
924 
925 	if (cancel_delayed_work(&ifp->dad_work))
926 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
927 			  ifp);
928 
929 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
930 		pr_warn("Freeing alive inet6 address %p\n", ifp);
931 		return;
932 	}
933 
934 	kfree_rcu(ifp, rcu);
935 }
936 
937 static void
938 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
939 {
940 	struct list_head *p;
941 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
942 
943 	/*
944 	 * Each device address list is sorted in order of scope -
945 	 * global before linklocal.
946 	 */
947 	list_for_each(p, &idev->addr_list) {
948 		struct inet6_ifaddr *ifa
949 			= list_entry(p, struct inet6_ifaddr, if_list);
950 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
951 			break;
952 	}
953 
954 	list_add_tail_rcu(&ifp->if_list, p);
955 }
956 
957 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
958 {
959 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
960 
961 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
962 }
963 
964 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
965 			       struct net_device *dev, unsigned int hash)
966 {
967 	struct inet6_ifaddr *ifp;
968 
969 	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
970 		if (!net_eq(dev_net(ifp->idev->dev), net))
971 			continue;
972 		if (ipv6_addr_equal(&ifp->addr, addr)) {
973 			if (!dev || ifp->idev->dev == dev)
974 				return true;
975 		}
976 	}
977 	return false;
978 }
979 
980 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
981 {
982 	unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
983 	int err = 0;
984 
985 	spin_lock(&addrconf_hash_lock);
986 
987 	/* Ignore adding duplicate addresses on an interface */
988 	if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
989 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
990 		err = -EEXIST;
991 	} else {
992 		hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
993 	}
994 
995 	spin_unlock(&addrconf_hash_lock);
996 
997 	return err;
998 }
999 
1000 /* On success it returns ifp with increased reference count */
1001 
1002 static struct inet6_ifaddr *
1003 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1004 	      bool can_block, struct netlink_ext_ack *extack)
1005 {
1006 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1007 	int addr_type = ipv6_addr_type(cfg->pfx);
1008 	struct net *net = dev_net(idev->dev);
1009 	struct inet6_ifaddr *ifa = NULL;
1010 	struct fib6_info *f6i = NULL;
1011 	int err = 0;
1012 
1013 	if (addr_type == IPV6_ADDR_ANY ||
1014 	    addr_type & IPV6_ADDR_MULTICAST ||
1015 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1016 	     !netif_is_l3_master(idev->dev) &&
1017 	     addr_type & IPV6_ADDR_LOOPBACK))
1018 		return ERR_PTR(-EADDRNOTAVAIL);
1019 
1020 	if (idev->dead) {
1021 		err = -ENODEV;			/*XXX*/
1022 		goto out;
1023 	}
1024 
1025 	if (idev->cnf.disable_ipv6) {
1026 		err = -EACCES;
1027 		goto out;
1028 	}
1029 
1030 	/* validator notifier needs to be blocking;
1031 	 * do not call in atomic context
1032 	 */
1033 	if (can_block) {
1034 		struct in6_validator_info i6vi = {
1035 			.i6vi_addr = *cfg->pfx,
1036 			.i6vi_dev = idev,
1037 			.extack = extack,
1038 		};
1039 
1040 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1041 		err = notifier_to_errno(err);
1042 		if (err < 0)
1043 			goto out;
1044 	}
1045 
1046 	ifa = kzalloc(sizeof(*ifa), gfp_flags);
1047 	if (!ifa) {
1048 		err = -ENOBUFS;
1049 		goto out;
1050 	}
1051 
1052 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1053 	if (IS_ERR(f6i)) {
1054 		err = PTR_ERR(f6i);
1055 		f6i = NULL;
1056 		goto out;
1057 	}
1058 
1059 	if (net->ipv6.devconf_all->disable_policy ||
1060 	    idev->cnf.disable_policy)
1061 		f6i->dst_nopolicy = true;
1062 
1063 	neigh_parms_data_state_setall(idev->nd_parms);
1064 
1065 	ifa->addr = *cfg->pfx;
1066 	if (cfg->peer_pfx)
1067 		ifa->peer_addr = *cfg->peer_pfx;
1068 
1069 	spin_lock_init(&ifa->lock);
1070 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1071 	INIT_HLIST_NODE(&ifa->addr_lst);
1072 	ifa->scope = cfg->scope;
1073 	ifa->prefix_len = cfg->plen;
1074 	ifa->rt_priority = cfg->rt_priority;
1075 	ifa->flags = cfg->ifa_flags;
1076 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1077 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1078 		ifa->flags |= IFA_F_TENTATIVE;
1079 	ifa->valid_lft = cfg->valid_lft;
1080 	ifa->prefered_lft = cfg->preferred_lft;
1081 	ifa->cstamp = ifa->tstamp = jiffies;
1082 	ifa->tokenized = false;
1083 
1084 	ifa->rt = f6i;
1085 
1086 	ifa->idev = idev;
1087 	in6_dev_hold(idev);
1088 
1089 	/* For caller */
1090 	refcount_set(&ifa->refcnt, 1);
1091 
1092 	rcu_read_lock_bh();
1093 
1094 	err = ipv6_add_addr_hash(idev->dev, ifa);
1095 	if (err < 0) {
1096 		rcu_read_unlock_bh();
1097 		goto out;
1098 	}
1099 
1100 	write_lock(&idev->lock);
1101 
1102 	/* Add to inet6_dev unicast addr list. */
1103 	ipv6_link_dev_addr(idev, ifa);
1104 
1105 	if (ifa->flags&IFA_F_TEMPORARY) {
1106 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1107 		in6_ifa_hold(ifa);
1108 	}
1109 
1110 	in6_ifa_hold(ifa);
1111 	write_unlock(&idev->lock);
1112 
1113 	rcu_read_unlock_bh();
1114 
1115 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1116 out:
1117 	if (unlikely(err < 0)) {
1118 		fib6_info_release(f6i);
1119 
1120 		if (ifa) {
1121 			if (ifa->idev)
1122 				in6_dev_put(ifa->idev);
1123 			kfree(ifa);
1124 		}
1125 		ifa = ERR_PTR(err);
1126 	}
1127 
1128 	return ifa;
1129 }
1130 
1131 enum cleanup_prefix_rt_t {
1132 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1133 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1134 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1135 };
1136 
1137 /*
1138  * Check, whether the prefix for ifp would still need a prefix route
1139  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1140  * constants.
1141  *
1142  * 1) we don't purge prefix if address was not permanent.
1143  *    prefix is managed by its own lifetime.
1144  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1145  * 3) if there are no addresses, delete prefix.
1146  * 4) if there are still other permanent address(es),
1147  *    corresponding prefix is still permanent.
1148  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1149  *    don't purge the prefix, assume user space is managing it.
1150  * 6) otherwise, update prefix lifetime to the
1151  *    longest valid lifetime among the corresponding
1152  *    addresses on the device.
1153  *    Note: subsequent RA will update lifetime.
1154  **/
1155 static enum cleanup_prefix_rt_t
1156 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1157 {
1158 	struct inet6_ifaddr *ifa;
1159 	struct inet6_dev *idev = ifp->idev;
1160 	unsigned long lifetime;
1161 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1162 
1163 	*expires = jiffies;
1164 
1165 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1166 		if (ifa == ifp)
1167 			continue;
1168 		if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1169 				       ifp->prefix_len))
1170 			continue;
1171 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1172 			return CLEANUP_PREFIX_RT_NOP;
1173 
1174 		action = CLEANUP_PREFIX_RT_EXPIRE;
1175 
1176 		spin_lock(&ifa->lock);
1177 
1178 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1179 		/*
1180 		 * Note: Because this address is
1181 		 * not permanent, lifetime <
1182 		 * LONG_MAX / HZ here.
1183 		 */
1184 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1185 			*expires = ifa->tstamp + lifetime * HZ;
1186 		spin_unlock(&ifa->lock);
1187 	}
1188 
1189 	return action;
1190 }
1191 
1192 static void
1193 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1194 {
1195 	struct fib6_info *f6i;
1196 
1197 	f6i = addrconf_get_prefix_route(&ifp->addr,
1198 				       ifp->prefix_len,
1199 				       ifp->idev->dev,
1200 				       0, RTF_GATEWAY | RTF_DEFAULT);
1201 	if (f6i) {
1202 		if (del_rt)
1203 			ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1204 		else {
1205 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1206 				fib6_set_expires(f6i, expires);
1207 			fib6_info_release(f6i);
1208 		}
1209 	}
1210 }
1211 
1212 
1213 /* This function wants to get referenced ifp and releases it before return */
1214 
1215 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1216 {
1217 	int state;
1218 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1219 	unsigned long expires;
1220 
1221 	ASSERT_RTNL();
1222 
1223 	spin_lock_bh(&ifp->lock);
1224 	state = ifp->state;
1225 	ifp->state = INET6_IFADDR_STATE_DEAD;
1226 	spin_unlock_bh(&ifp->lock);
1227 
1228 	if (state == INET6_IFADDR_STATE_DEAD)
1229 		goto out;
1230 
1231 	spin_lock_bh(&addrconf_hash_lock);
1232 	hlist_del_init_rcu(&ifp->addr_lst);
1233 	spin_unlock_bh(&addrconf_hash_lock);
1234 
1235 	write_lock_bh(&ifp->idev->lock);
1236 
1237 	if (ifp->flags&IFA_F_TEMPORARY) {
1238 		list_del(&ifp->tmp_list);
1239 		if (ifp->ifpub) {
1240 			in6_ifa_put(ifp->ifpub);
1241 			ifp->ifpub = NULL;
1242 		}
1243 		__in6_ifa_put(ifp);
1244 	}
1245 
1246 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1247 		action = check_cleanup_prefix_route(ifp, &expires);
1248 
1249 	list_del_rcu(&ifp->if_list);
1250 	__in6_ifa_put(ifp);
1251 
1252 	write_unlock_bh(&ifp->idev->lock);
1253 
1254 	addrconf_del_dad_work(ifp);
1255 
1256 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1257 
1258 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1259 
1260 	if (action != CLEANUP_PREFIX_RT_NOP) {
1261 		cleanup_prefix_route(ifp, expires,
1262 			action == CLEANUP_PREFIX_RT_DEL);
1263 	}
1264 
1265 	/* clean up prefsrc entries */
1266 	rt6_remove_prefsrc(ifp);
1267 out:
1268 	in6_ifa_put(ifp);
1269 }
1270 
1271 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1272 				struct inet6_ifaddr *ift,
1273 				bool block)
1274 {
1275 	struct inet6_dev *idev = ifp->idev;
1276 	struct in6_addr addr, *tmpaddr;
1277 	unsigned long tmp_tstamp, age;
1278 	unsigned long regen_advance;
1279 	struct ifa6_config cfg;
1280 	int ret = 0;
1281 	unsigned long now = jiffies;
1282 	long max_desync_factor;
1283 	s32 cnf_temp_preferred_lft;
1284 
1285 	write_lock_bh(&idev->lock);
1286 	if (ift) {
1287 		spin_lock_bh(&ift->lock);
1288 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1289 		spin_unlock_bh(&ift->lock);
1290 		tmpaddr = &addr;
1291 	} else {
1292 		tmpaddr = NULL;
1293 	}
1294 retry:
1295 	in6_dev_hold(idev);
1296 	if (idev->cnf.use_tempaddr <= 0) {
1297 		write_unlock_bh(&idev->lock);
1298 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1299 		in6_dev_put(idev);
1300 		ret = -1;
1301 		goto out;
1302 	}
1303 	spin_lock_bh(&ifp->lock);
1304 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1305 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1306 		spin_unlock_bh(&ifp->lock);
1307 		write_unlock_bh(&idev->lock);
1308 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1309 			__func__);
1310 		in6_dev_put(idev);
1311 		ret = -1;
1312 		goto out;
1313 	}
1314 	in6_ifa_hold(ifp);
1315 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1316 	ipv6_try_regen_rndid(idev, tmpaddr);
1317 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
1318 	age = (now - ifp->tstamp) / HZ;
1319 
1320 	regen_advance = idev->cnf.regen_max_retry *
1321 			idev->cnf.dad_transmits *
1322 			NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1323 
1324 	/* recalculate max_desync_factor each time and update
1325 	 * idev->desync_factor if it's larger
1326 	 */
1327 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1328 	max_desync_factor = min_t(__u32,
1329 				  idev->cnf.max_desync_factor,
1330 				  cnf_temp_preferred_lft - regen_advance);
1331 
1332 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1333 		if (max_desync_factor > 0) {
1334 			get_random_bytes(&idev->desync_factor,
1335 					 sizeof(idev->desync_factor));
1336 			idev->desync_factor %= max_desync_factor;
1337 		} else {
1338 			idev->desync_factor = 0;
1339 		}
1340 	}
1341 
1342 	memset(&cfg, 0, sizeof(cfg));
1343 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1344 			      idev->cnf.temp_valid_lft + age);
1345 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1346 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1347 
1348 	cfg.plen = ifp->prefix_len;
1349 	tmp_tstamp = ifp->tstamp;
1350 	spin_unlock_bh(&ifp->lock);
1351 
1352 	write_unlock_bh(&idev->lock);
1353 
1354 	/* A temporary address is created only if this calculated Preferred
1355 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1356 	 * an implementation must not create a temporary address with a zero
1357 	 * Preferred Lifetime.
1358 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1359 	 * temporary addresses being generated.
1360 	 */
1361 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1362 	if (cfg.preferred_lft <= regen_advance + age) {
1363 		in6_ifa_put(ifp);
1364 		in6_dev_put(idev);
1365 		ret = -1;
1366 		goto out;
1367 	}
1368 
1369 	cfg.ifa_flags = IFA_F_TEMPORARY;
1370 	/* set in addrconf_prefix_rcv() */
1371 	if (ifp->flags & IFA_F_OPTIMISTIC)
1372 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1373 
1374 	cfg.pfx = &addr;
1375 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1376 
1377 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1378 	if (IS_ERR(ift)) {
1379 		in6_ifa_put(ifp);
1380 		in6_dev_put(idev);
1381 		pr_info("%s: retry temporary address regeneration\n", __func__);
1382 		tmpaddr = &addr;
1383 		write_lock_bh(&idev->lock);
1384 		goto retry;
1385 	}
1386 
1387 	spin_lock_bh(&ift->lock);
1388 	ift->ifpub = ifp;
1389 	ift->cstamp = now;
1390 	ift->tstamp = tmp_tstamp;
1391 	spin_unlock_bh(&ift->lock);
1392 
1393 	addrconf_dad_start(ift);
1394 	in6_ifa_put(ift);
1395 	in6_dev_put(idev);
1396 out:
1397 	return ret;
1398 }
1399 
1400 /*
1401  *	Choose an appropriate source address (RFC3484)
1402  */
1403 enum {
1404 	IPV6_SADDR_RULE_INIT = 0,
1405 	IPV6_SADDR_RULE_LOCAL,
1406 	IPV6_SADDR_RULE_SCOPE,
1407 	IPV6_SADDR_RULE_PREFERRED,
1408 #ifdef CONFIG_IPV6_MIP6
1409 	IPV6_SADDR_RULE_HOA,
1410 #endif
1411 	IPV6_SADDR_RULE_OIF,
1412 	IPV6_SADDR_RULE_LABEL,
1413 	IPV6_SADDR_RULE_PRIVACY,
1414 	IPV6_SADDR_RULE_ORCHID,
1415 	IPV6_SADDR_RULE_PREFIX,
1416 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1417 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1418 #endif
1419 	IPV6_SADDR_RULE_MAX
1420 };
1421 
1422 struct ipv6_saddr_score {
1423 	int			rule;
1424 	int			addr_type;
1425 	struct inet6_ifaddr	*ifa;
1426 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1427 	int			scopedist;
1428 	int			matchlen;
1429 };
1430 
1431 struct ipv6_saddr_dst {
1432 	const struct in6_addr *addr;
1433 	int ifindex;
1434 	int scope;
1435 	int label;
1436 	unsigned int prefs;
1437 };
1438 
1439 static inline int ipv6_saddr_preferred(int type)
1440 {
1441 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1442 		return 1;
1443 	return 0;
1444 }
1445 
1446 static bool ipv6_use_optimistic_addr(struct net *net,
1447 				     struct inet6_dev *idev)
1448 {
1449 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1450 	if (!idev)
1451 		return false;
1452 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1453 		return false;
1454 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1455 		return false;
1456 
1457 	return true;
1458 #else
1459 	return false;
1460 #endif
1461 }
1462 
1463 static bool ipv6_allow_optimistic_dad(struct net *net,
1464 				      struct inet6_dev *idev)
1465 {
1466 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1467 	if (!idev)
1468 		return false;
1469 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1470 		return false;
1471 
1472 	return true;
1473 #else
1474 	return false;
1475 #endif
1476 }
1477 
1478 static int ipv6_get_saddr_eval(struct net *net,
1479 			       struct ipv6_saddr_score *score,
1480 			       struct ipv6_saddr_dst *dst,
1481 			       int i)
1482 {
1483 	int ret;
1484 
1485 	if (i <= score->rule) {
1486 		switch (i) {
1487 		case IPV6_SADDR_RULE_SCOPE:
1488 			ret = score->scopedist;
1489 			break;
1490 		case IPV6_SADDR_RULE_PREFIX:
1491 			ret = score->matchlen;
1492 			break;
1493 		default:
1494 			ret = !!test_bit(i, score->scorebits);
1495 		}
1496 		goto out;
1497 	}
1498 
1499 	switch (i) {
1500 	case IPV6_SADDR_RULE_INIT:
1501 		/* Rule 0: remember if hiscore is not ready yet */
1502 		ret = !!score->ifa;
1503 		break;
1504 	case IPV6_SADDR_RULE_LOCAL:
1505 		/* Rule 1: Prefer same address */
1506 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1507 		break;
1508 	case IPV6_SADDR_RULE_SCOPE:
1509 		/* Rule 2: Prefer appropriate scope
1510 		 *
1511 		 *      ret
1512 		 *       ^
1513 		 *    -1 |  d 15
1514 		 *    ---+--+-+---> scope
1515 		 *       |
1516 		 *       |             d is scope of the destination.
1517 		 *  B-d  |  \
1518 		 *       |   \      <- smaller scope is better if
1519 		 *  B-15 |    \        if scope is enough for destination.
1520 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1521 		 * d-C-1 | /
1522 		 *       |/         <- greater is better
1523 		 *   -C  /             if scope is not enough for destination.
1524 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1525 		 *
1526 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1527 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1528 		 * Assume B = 0 and we get C > 29.
1529 		 */
1530 		ret = __ipv6_addr_src_scope(score->addr_type);
1531 		if (ret >= dst->scope)
1532 			ret = -ret;
1533 		else
1534 			ret -= 128;	/* 30 is enough */
1535 		score->scopedist = ret;
1536 		break;
1537 	case IPV6_SADDR_RULE_PREFERRED:
1538 	    {
1539 		/* Rule 3: Avoid deprecated and optimistic addresses */
1540 		u8 avoid = IFA_F_DEPRECATED;
1541 
1542 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1543 			avoid |= IFA_F_OPTIMISTIC;
1544 		ret = ipv6_saddr_preferred(score->addr_type) ||
1545 		      !(score->ifa->flags & avoid);
1546 		break;
1547 	    }
1548 #ifdef CONFIG_IPV6_MIP6
1549 	case IPV6_SADDR_RULE_HOA:
1550 	    {
1551 		/* Rule 4: Prefer home address */
1552 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1553 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1554 		break;
1555 	    }
1556 #endif
1557 	case IPV6_SADDR_RULE_OIF:
1558 		/* Rule 5: Prefer outgoing interface */
1559 		ret = (!dst->ifindex ||
1560 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1561 		break;
1562 	case IPV6_SADDR_RULE_LABEL:
1563 		/* Rule 6: Prefer matching label */
1564 		ret = ipv6_addr_label(net,
1565 				      &score->ifa->addr, score->addr_type,
1566 				      score->ifa->idev->dev->ifindex) == dst->label;
1567 		break;
1568 	case IPV6_SADDR_RULE_PRIVACY:
1569 	    {
1570 		/* Rule 7: Prefer public address
1571 		 * Note: prefer temporary address if use_tempaddr >= 2
1572 		 */
1573 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1574 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1575 				score->ifa->idev->cnf.use_tempaddr >= 2;
1576 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1577 		break;
1578 	    }
1579 	case IPV6_SADDR_RULE_ORCHID:
1580 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1581 		 *	    non-ORCHID vs non-ORCHID
1582 		 */
1583 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1584 			ipv6_addr_orchid(dst->addr));
1585 		break;
1586 	case IPV6_SADDR_RULE_PREFIX:
1587 		/* Rule 8: Use longest matching prefix */
1588 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1589 		if (ret > score->ifa->prefix_len)
1590 			ret = score->ifa->prefix_len;
1591 		score->matchlen = ret;
1592 		break;
1593 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1594 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1595 		/* Optimistic addresses still have lower precedence than other
1596 		 * preferred addresses.
1597 		 */
1598 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1599 		break;
1600 #endif
1601 	default:
1602 		ret = 0;
1603 	}
1604 
1605 	if (ret)
1606 		__set_bit(i, score->scorebits);
1607 	score->rule = i;
1608 out:
1609 	return ret;
1610 }
1611 
1612 static int __ipv6_dev_get_saddr(struct net *net,
1613 				struct ipv6_saddr_dst *dst,
1614 				struct inet6_dev *idev,
1615 				struct ipv6_saddr_score *scores,
1616 				int hiscore_idx)
1617 {
1618 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1619 
1620 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1621 		int i;
1622 
1623 		/*
1624 		 * - Tentative Address (RFC2462 section 5.4)
1625 		 *  - A tentative address is not considered
1626 		 *    "assigned to an interface" in the traditional
1627 		 *    sense, unless it is also flagged as optimistic.
1628 		 * - Candidate Source Address (section 4)
1629 		 *  - In any case, anycast addresses, multicast
1630 		 *    addresses, and the unspecified address MUST
1631 		 *    NOT be included in a candidate set.
1632 		 */
1633 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1634 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1635 			continue;
1636 
1637 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1638 
1639 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1640 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1641 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1642 					    idev->dev->name);
1643 			continue;
1644 		}
1645 
1646 		score->rule = -1;
1647 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1648 
1649 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1650 			int minihiscore, miniscore;
1651 
1652 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1653 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1654 
1655 			if (minihiscore > miniscore) {
1656 				if (i == IPV6_SADDR_RULE_SCOPE &&
1657 				    score->scopedist > 0) {
1658 					/*
1659 					 * special case:
1660 					 * each remaining entry
1661 					 * has too small (not enough)
1662 					 * scope, because ifa entries
1663 					 * are sorted by their scope
1664 					 * values.
1665 					 */
1666 					goto out;
1667 				}
1668 				break;
1669 			} else if (minihiscore < miniscore) {
1670 				swap(hiscore, score);
1671 				hiscore_idx = 1 - hiscore_idx;
1672 
1673 				/* restore our iterator */
1674 				score->ifa = hiscore->ifa;
1675 
1676 				break;
1677 			}
1678 		}
1679 	}
1680 out:
1681 	return hiscore_idx;
1682 }
1683 
1684 static int ipv6_get_saddr_master(struct net *net,
1685 				 const struct net_device *dst_dev,
1686 				 const struct net_device *master,
1687 				 struct ipv6_saddr_dst *dst,
1688 				 struct ipv6_saddr_score *scores,
1689 				 int hiscore_idx)
1690 {
1691 	struct inet6_dev *idev;
1692 
1693 	idev = __in6_dev_get(dst_dev);
1694 	if (idev)
1695 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1696 						   scores, hiscore_idx);
1697 
1698 	idev = __in6_dev_get(master);
1699 	if (idev)
1700 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1701 						   scores, hiscore_idx);
1702 
1703 	return hiscore_idx;
1704 }
1705 
1706 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1707 		       const struct in6_addr *daddr, unsigned int prefs,
1708 		       struct in6_addr *saddr)
1709 {
1710 	struct ipv6_saddr_score scores[2], *hiscore;
1711 	struct ipv6_saddr_dst dst;
1712 	struct inet6_dev *idev;
1713 	struct net_device *dev;
1714 	int dst_type;
1715 	bool use_oif_addr = false;
1716 	int hiscore_idx = 0;
1717 	int ret = 0;
1718 
1719 	dst_type = __ipv6_addr_type(daddr);
1720 	dst.addr = daddr;
1721 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1722 	dst.scope = __ipv6_addr_src_scope(dst_type);
1723 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1724 	dst.prefs = prefs;
1725 
1726 	scores[hiscore_idx].rule = -1;
1727 	scores[hiscore_idx].ifa = NULL;
1728 
1729 	rcu_read_lock();
1730 
1731 	/* Candidate Source Address (section 4)
1732 	 *  - multicast and link-local destination address,
1733 	 *    the set of candidate source address MUST only
1734 	 *    include addresses assigned to interfaces
1735 	 *    belonging to the same link as the outgoing
1736 	 *    interface.
1737 	 * (- For site-local destination addresses, the
1738 	 *    set of candidate source addresses MUST only
1739 	 *    include addresses assigned to interfaces
1740 	 *    belonging to the same site as the outgoing
1741 	 *    interface.)
1742 	 *  - "It is RECOMMENDED that the candidate source addresses
1743 	 *    be the set of unicast addresses assigned to the
1744 	 *    interface that will be used to send to the destination
1745 	 *    (the 'outgoing' interface)." (RFC 6724)
1746 	 */
1747 	if (dst_dev) {
1748 		idev = __in6_dev_get(dst_dev);
1749 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1750 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1751 		    (idev && idev->cnf.use_oif_addrs_only)) {
1752 			use_oif_addr = true;
1753 		}
1754 	}
1755 
1756 	if (use_oif_addr) {
1757 		if (idev)
1758 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1759 	} else {
1760 		const struct net_device *master;
1761 		int master_idx = 0;
1762 
1763 		/* if dst_dev exists and is enslaved to an L3 device, then
1764 		 * prefer addresses from dst_dev and then the master over
1765 		 * any other enslaved devices in the L3 domain.
1766 		 */
1767 		master = l3mdev_master_dev_rcu(dst_dev);
1768 		if (master) {
1769 			master_idx = master->ifindex;
1770 
1771 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1772 							    master, &dst,
1773 							    scores, hiscore_idx);
1774 
1775 			if (scores[hiscore_idx].ifa)
1776 				goto out;
1777 		}
1778 
1779 		for_each_netdev_rcu(net, dev) {
1780 			/* only consider addresses on devices in the
1781 			 * same L3 domain
1782 			 */
1783 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1784 				continue;
1785 			idev = __in6_dev_get(dev);
1786 			if (!idev)
1787 				continue;
1788 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1789 		}
1790 	}
1791 
1792 out:
1793 	hiscore = &scores[hiscore_idx];
1794 	if (!hiscore->ifa)
1795 		ret = -EADDRNOTAVAIL;
1796 	else
1797 		*saddr = hiscore->ifa->addr;
1798 
1799 	rcu_read_unlock();
1800 	return ret;
1801 }
1802 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1803 
1804 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1805 		      u32 banned_flags)
1806 {
1807 	struct inet6_ifaddr *ifp;
1808 	int err = -EADDRNOTAVAIL;
1809 
1810 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1811 		if (ifp->scope > IFA_LINK)
1812 			break;
1813 		if (ifp->scope == IFA_LINK &&
1814 		    !(ifp->flags & banned_flags)) {
1815 			*addr = ifp->addr;
1816 			err = 0;
1817 			break;
1818 		}
1819 	}
1820 	return err;
1821 }
1822 
1823 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1824 		    u32 banned_flags)
1825 {
1826 	struct inet6_dev *idev;
1827 	int err = -EADDRNOTAVAIL;
1828 
1829 	rcu_read_lock();
1830 	idev = __in6_dev_get(dev);
1831 	if (idev) {
1832 		read_lock_bh(&idev->lock);
1833 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1834 		read_unlock_bh(&idev->lock);
1835 	}
1836 	rcu_read_unlock();
1837 	return err;
1838 }
1839 
1840 static int ipv6_count_addresses(const struct inet6_dev *idev)
1841 {
1842 	const struct inet6_ifaddr *ifp;
1843 	int cnt = 0;
1844 
1845 	rcu_read_lock();
1846 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1847 		cnt++;
1848 	rcu_read_unlock();
1849 	return cnt;
1850 }
1851 
1852 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1853 		  const struct net_device *dev, int strict)
1854 {
1855 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1856 				       strict, IFA_F_TENTATIVE);
1857 }
1858 EXPORT_SYMBOL(ipv6_chk_addr);
1859 
1860 /* device argument is used to find the L3 domain of interest. If
1861  * skip_dev_check is set, then the ifp device is not checked against
1862  * the passed in dev argument. So the 2 cases for addresses checks are:
1863  *   1. does the address exist in the L3 domain that dev is part of
1864  *      (skip_dev_check = true), or
1865  *
1866  *   2. does the address exist on the specific device
1867  *      (skip_dev_check = false)
1868  */
1869 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1870 			    const struct net_device *dev, bool skip_dev_check,
1871 			    int strict, u32 banned_flags)
1872 {
1873 	unsigned int hash = inet6_addr_hash(net, addr);
1874 	const struct net_device *l3mdev;
1875 	struct inet6_ifaddr *ifp;
1876 	u32 ifp_flags;
1877 
1878 	rcu_read_lock();
1879 
1880 	l3mdev = l3mdev_master_dev_rcu(dev);
1881 	if (skip_dev_check)
1882 		dev = NULL;
1883 
1884 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1885 		if (!net_eq(dev_net(ifp->idev->dev), net))
1886 			continue;
1887 
1888 		if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1889 			continue;
1890 
1891 		/* Decouple optimistic from tentative for evaluation here.
1892 		 * Ban optimistic addresses explicitly, when required.
1893 		 */
1894 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1895 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1896 			    : ifp->flags;
1897 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1898 		    !(ifp_flags&banned_flags) &&
1899 		    (!dev || ifp->idev->dev == dev ||
1900 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1901 			rcu_read_unlock();
1902 			return 1;
1903 		}
1904 	}
1905 
1906 	rcu_read_unlock();
1907 	return 0;
1908 }
1909 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1910 
1911 
1912 /* Compares an address/prefix_len with addresses on device @dev.
1913  * If one is found it returns true.
1914  */
1915 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1916 	const unsigned int prefix_len, struct net_device *dev)
1917 {
1918 	const struct inet6_ifaddr *ifa;
1919 	const struct inet6_dev *idev;
1920 	bool ret = false;
1921 
1922 	rcu_read_lock();
1923 	idev = __in6_dev_get(dev);
1924 	if (idev) {
1925 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1926 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1927 			if (ret)
1928 				break;
1929 		}
1930 	}
1931 	rcu_read_unlock();
1932 
1933 	return ret;
1934 }
1935 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1936 
1937 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1938 {
1939 	const struct inet6_ifaddr *ifa;
1940 	const struct inet6_dev *idev;
1941 	int	onlink;
1942 
1943 	onlink = 0;
1944 	rcu_read_lock();
1945 	idev = __in6_dev_get(dev);
1946 	if (idev) {
1947 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1948 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1949 						   ifa->prefix_len);
1950 			if (onlink)
1951 				break;
1952 		}
1953 	}
1954 	rcu_read_unlock();
1955 	return onlink;
1956 }
1957 EXPORT_SYMBOL(ipv6_chk_prefix);
1958 
1959 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1960 				     struct net_device *dev, int strict)
1961 {
1962 	unsigned int hash = inet6_addr_hash(net, addr);
1963 	struct inet6_ifaddr *ifp, *result = NULL;
1964 
1965 	rcu_read_lock();
1966 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1967 		if (!net_eq(dev_net(ifp->idev->dev), net))
1968 			continue;
1969 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1970 			if (!dev || ifp->idev->dev == dev ||
1971 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1972 				result = ifp;
1973 				in6_ifa_hold(ifp);
1974 				break;
1975 			}
1976 		}
1977 	}
1978 	rcu_read_unlock();
1979 
1980 	return result;
1981 }
1982 
1983 /* Gets referenced address, destroys ifaddr */
1984 
1985 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1986 {
1987 	if (dad_failed)
1988 		ifp->flags |= IFA_F_DADFAILED;
1989 
1990 	if (ifp->flags&IFA_F_TEMPORARY) {
1991 		struct inet6_ifaddr *ifpub;
1992 		spin_lock_bh(&ifp->lock);
1993 		ifpub = ifp->ifpub;
1994 		if (ifpub) {
1995 			in6_ifa_hold(ifpub);
1996 			spin_unlock_bh(&ifp->lock);
1997 			ipv6_create_tempaddr(ifpub, ifp, true);
1998 			in6_ifa_put(ifpub);
1999 		} else {
2000 			spin_unlock_bh(&ifp->lock);
2001 		}
2002 		ipv6_del_addr(ifp);
2003 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2004 		spin_lock_bh(&ifp->lock);
2005 		addrconf_del_dad_work(ifp);
2006 		ifp->flags |= IFA_F_TENTATIVE;
2007 		if (dad_failed)
2008 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2009 		spin_unlock_bh(&ifp->lock);
2010 		if (dad_failed)
2011 			ipv6_ifa_notify(0, ifp);
2012 		in6_ifa_put(ifp);
2013 	} else {
2014 		ipv6_del_addr(ifp);
2015 	}
2016 }
2017 
2018 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2019 {
2020 	int err = -ENOENT;
2021 
2022 	spin_lock_bh(&ifp->lock);
2023 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2024 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2025 		err = 0;
2026 	}
2027 	spin_unlock_bh(&ifp->lock);
2028 
2029 	return err;
2030 }
2031 
2032 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2033 {
2034 	struct inet6_dev *idev = ifp->idev;
2035 	struct net *net = dev_net(ifp->idev->dev);
2036 
2037 	if (addrconf_dad_end(ifp)) {
2038 		in6_ifa_put(ifp);
2039 		return;
2040 	}
2041 
2042 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2043 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2044 
2045 	spin_lock_bh(&ifp->lock);
2046 
2047 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2048 		struct in6_addr new_addr;
2049 		struct inet6_ifaddr *ifp2;
2050 		int retries = ifp->stable_privacy_retry + 1;
2051 		struct ifa6_config cfg = {
2052 			.pfx = &new_addr,
2053 			.plen = ifp->prefix_len,
2054 			.ifa_flags = ifp->flags,
2055 			.valid_lft = ifp->valid_lft,
2056 			.preferred_lft = ifp->prefered_lft,
2057 			.scope = ifp->scope,
2058 		};
2059 
2060 		if (retries > net->ipv6.sysctl.idgen_retries) {
2061 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2062 					     ifp->idev->dev->name);
2063 			goto errdad;
2064 		}
2065 
2066 		new_addr = ifp->addr;
2067 		if (ipv6_generate_stable_address(&new_addr, retries,
2068 						 idev))
2069 			goto errdad;
2070 
2071 		spin_unlock_bh(&ifp->lock);
2072 
2073 		if (idev->cnf.max_addresses &&
2074 		    ipv6_count_addresses(idev) >=
2075 		    idev->cnf.max_addresses)
2076 			goto lock_errdad;
2077 
2078 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2079 				     ifp->idev->dev->name);
2080 
2081 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2082 		if (IS_ERR(ifp2))
2083 			goto lock_errdad;
2084 
2085 		spin_lock_bh(&ifp2->lock);
2086 		ifp2->stable_privacy_retry = retries;
2087 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2088 		spin_unlock_bh(&ifp2->lock);
2089 
2090 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2091 		in6_ifa_put(ifp2);
2092 lock_errdad:
2093 		spin_lock_bh(&ifp->lock);
2094 	}
2095 
2096 errdad:
2097 	/* transition from _POSTDAD to _ERRDAD */
2098 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2099 	spin_unlock_bh(&ifp->lock);
2100 
2101 	addrconf_mod_dad_work(ifp, 0);
2102 	in6_ifa_put(ifp);
2103 }
2104 
2105 /* Join to solicited addr multicast group.
2106  * caller must hold RTNL */
2107 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2108 {
2109 	struct in6_addr maddr;
2110 
2111 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2112 		return;
2113 
2114 	addrconf_addr_solict_mult(addr, &maddr);
2115 	ipv6_dev_mc_inc(dev, &maddr);
2116 }
2117 
2118 /* caller must hold RTNL */
2119 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2120 {
2121 	struct in6_addr maddr;
2122 
2123 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2124 		return;
2125 
2126 	addrconf_addr_solict_mult(addr, &maddr);
2127 	__ipv6_dev_mc_dec(idev, &maddr);
2128 }
2129 
2130 /* caller must hold RTNL */
2131 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2132 {
2133 	struct in6_addr addr;
2134 
2135 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2136 		return;
2137 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2138 	if (ipv6_addr_any(&addr))
2139 		return;
2140 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2141 }
2142 
2143 /* caller must hold RTNL */
2144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2145 {
2146 	struct in6_addr addr;
2147 
2148 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2149 		return;
2150 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2151 	if (ipv6_addr_any(&addr))
2152 		return;
2153 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2154 }
2155 
2156 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2157 {
2158 	switch (dev->addr_len) {
2159 	case ETH_ALEN:
2160 		memcpy(eui, dev->dev_addr, 3);
2161 		eui[3] = 0xFF;
2162 		eui[4] = 0xFE;
2163 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2164 		break;
2165 	case EUI64_ADDR_LEN:
2166 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2167 		eui[0] ^= 2;
2168 		break;
2169 	default:
2170 		return -1;
2171 	}
2172 
2173 	return 0;
2174 }
2175 
2176 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2177 {
2178 	union fwnet_hwaddr *ha;
2179 
2180 	if (dev->addr_len != FWNET_ALEN)
2181 		return -1;
2182 
2183 	ha = (union fwnet_hwaddr *)dev->dev_addr;
2184 
2185 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2186 	eui[0] ^= 2;
2187 	return 0;
2188 }
2189 
2190 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2191 {
2192 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2193 	if (dev->addr_len != ARCNET_ALEN)
2194 		return -1;
2195 	memset(eui, 0, 7);
2196 	eui[7] = *(u8 *)dev->dev_addr;
2197 	return 0;
2198 }
2199 
2200 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2201 {
2202 	if (dev->addr_len != INFINIBAND_ALEN)
2203 		return -1;
2204 	memcpy(eui, dev->dev_addr + 12, 8);
2205 	eui[0] |= 2;
2206 	return 0;
2207 }
2208 
2209 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2210 {
2211 	if (addr == 0)
2212 		return -1;
2213 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2214 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2215 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2216 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2217 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2218 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2219 	eui[1] = 0;
2220 	eui[2] = 0x5E;
2221 	eui[3] = 0xFE;
2222 	memcpy(eui + 4, &addr, 4);
2223 	return 0;
2224 }
2225 
2226 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2227 {
2228 	if (dev->priv_flags & IFF_ISATAP)
2229 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2230 	return -1;
2231 }
2232 
2233 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2234 {
2235 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2236 }
2237 
2238 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2239 {
2240 	memcpy(eui, dev->perm_addr, 3);
2241 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2242 	eui[3] = 0xFF;
2243 	eui[4] = 0xFE;
2244 	eui[0] ^= 2;
2245 	return 0;
2246 }
2247 
2248 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2249 {
2250 	switch (dev->type) {
2251 	case ARPHRD_ETHER:
2252 	case ARPHRD_FDDI:
2253 		return addrconf_ifid_eui48(eui, dev);
2254 	case ARPHRD_ARCNET:
2255 		return addrconf_ifid_arcnet(eui, dev);
2256 	case ARPHRD_INFINIBAND:
2257 		return addrconf_ifid_infiniband(eui, dev);
2258 	case ARPHRD_SIT:
2259 		return addrconf_ifid_sit(eui, dev);
2260 	case ARPHRD_IPGRE:
2261 	case ARPHRD_TUNNEL:
2262 		return addrconf_ifid_gre(eui, dev);
2263 	case ARPHRD_6LOWPAN:
2264 		return addrconf_ifid_6lowpan(eui, dev);
2265 	case ARPHRD_IEEE1394:
2266 		return addrconf_ifid_ieee1394(eui, dev);
2267 	case ARPHRD_TUNNEL6:
2268 	case ARPHRD_IP6GRE:
2269 	case ARPHRD_RAWIP:
2270 		return addrconf_ifid_ip6tnl(eui, dev);
2271 	}
2272 	return -1;
2273 }
2274 
2275 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2276 {
2277 	int err = -1;
2278 	struct inet6_ifaddr *ifp;
2279 
2280 	read_lock_bh(&idev->lock);
2281 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2282 		if (ifp->scope > IFA_LINK)
2283 			break;
2284 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2285 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2286 			err = 0;
2287 			break;
2288 		}
2289 	}
2290 	read_unlock_bh(&idev->lock);
2291 	return err;
2292 }
2293 
2294 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2295 static void ipv6_regen_rndid(struct inet6_dev *idev)
2296 {
2297 regen:
2298 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
2299 	idev->rndid[0] &= ~0x02;
2300 
2301 	/*
2302 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2303 	 * check if generated address is not inappropriate
2304 	 *
2305 	 *  - Reserved subnet anycast (RFC 2526)
2306 	 *	11111101 11....11 1xxxxxxx
2307 	 *  - ISATAP (RFC4214) 6.1
2308 	 *	00-00-5E-FE-xx-xx-xx-xx
2309 	 *  - value 0
2310 	 *  - XXX: already assigned to an address on the device
2311 	 */
2312 	if (idev->rndid[0] == 0xfd &&
2313 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2314 	    (idev->rndid[7]&0x80))
2315 		goto regen;
2316 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2317 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2318 			goto regen;
2319 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2320 			goto regen;
2321 	}
2322 }
2323 
2324 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2325 {
2326 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2327 		ipv6_regen_rndid(idev);
2328 }
2329 
2330 /*
2331  *	Add prefix route.
2332  */
2333 
2334 static void
2335 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2336 		      struct net_device *dev, unsigned long expires,
2337 		      u32 flags, gfp_t gfp_flags)
2338 {
2339 	struct fib6_config cfg = {
2340 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2341 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2342 		.fc_ifindex = dev->ifindex,
2343 		.fc_expires = expires,
2344 		.fc_dst_len = plen,
2345 		.fc_flags = RTF_UP | flags,
2346 		.fc_nlinfo.nl_net = dev_net(dev),
2347 		.fc_protocol = RTPROT_KERNEL,
2348 		.fc_type = RTN_UNICAST,
2349 	};
2350 
2351 	cfg.fc_dst = *pfx;
2352 
2353 	/* Prevent useless cloning on PtP SIT.
2354 	   This thing is done here expecting that the whole
2355 	   class of non-broadcast devices need not cloning.
2356 	 */
2357 #if IS_ENABLED(CONFIG_IPV6_SIT)
2358 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2359 		cfg.fc_flags |= RTF_NONEXTHOP;
2360 #endif
2361 
2362 	ip6_route_add(&cfg, gfp_flags, NULL);
2363 }
2364 
2365 
2366 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2367 						  int plen,
2368 						  const struct net_device *dev,
2369 						  u32 flags, u32 noflags)
2370 {
2371 	struct fib6_node *fn;
2372 	struct fib6_info *rt = NULL;
2373 	struct fib6_table *table;
2374 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2375 
2376 	table = fib6_get_table(dev_net(dev), tb_id);
2377 	if (!table)
2378 		return NULL;
2379 
2380 	rcu_read_lock();
2381 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2382 	if (!fn)
2383 		goto out;
2384 
2385 	for_each_fib6_node_rt_rcu(fn) {
2386 		if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2387 			continue;
2388 		if ((rt->fib6_flags & flags) != flags)
2389 			continue;
2390 		if ((rt->fib6_flags & noflags) != 0)
2391 			continue;
2392 		if (!fib6_info_hold_safe(rt))
2393 			continue;
2394 		break;
2395 	}
2396 out:
2397 	rcu_read_unlock();
2398 	return rt;
2399 }
2400 
2401 
2402 /* Create "default" multicast route to the interface */
2403 
2404 static void addrconf_add_mroute(struct net_device *dev)
2405 {
2406 	struct fib6_config cfg = {
2407 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2408 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2409 		.fc_ifindex = dev->ifindex,
2410 		.fc_dst_len = 8,
2411 		.fc_flags = RTF_UP,
2412 		.fc_type = RTN_UNICAST,
2413 		.fc_nlinfo.nl_net = dev_net(dev),
2414 	};
2415 
2416 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2417 
2418 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2419 }
2420 
2421 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2422 {
2423 	struct inet6_dev *idev;
2424 
2425 	ASSERT_RTNL();
2426 
2427 	idev = ipv6_find_idev(dev);
2428 	if (!idev)
2429 		return ERR_PTR(-ENOBUFS);
2430 
2431 	if (idev->cnf.disable_ipv6)
2432 		return ERR_PTR(-EACCES);
2433 
2434 	/* Add default multicast route */
2435 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2436 		addrconf_add_mroute(dev);
2437 
2438 	return idev;
2439 }
2440 
2441 static void manage_tempaddrs(struct inet6_dev *idev,
2442 			     struct inet6_ifaddr *ifp,
2443 			     __u32 valid_lft, __u32 prefered_lft,
2444 			     bool create, unsigned long now)
2445 {
2446 	u32 flags;
2447 	struct inet6_ifaddr *ift;
2448 
2449 	read_lock_bh(&idev->lock);
2450 	/* update all temporary addresses in the list */
2451 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2452 		int age, max_valid, max_prefered;
2453 
2454 		if (ifp != ift->ifpub)
2455 			continue;
2456 
2457 		/* RFC 4941 section 3.3:
2458 		 * If a received option will extend the lifetime of a public
2459 		 * address, the lifetimes of temporary addresses should
2460 		 * be extended, subject to the overall constraint that no
2461 		 * temporary addresses should ever remain "valid" or "preferred"
2462 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2463 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2464 		 */
2465 		age = (now - ift->cstamp) / HZ;
2466 		max_valid = idev->cnf.temp_valid_lft - age;
2467 		if (max_valid < 0)
2468 			max_valid = 0;
2469 
2470 		max_prefered = idev->cnf.temp_prefered_lft -
2471 			       idev->desync_factor - age;
2472 		if (max_prefered < 0)
2473 			max_prefered = 0;
2474 
2475 		if (valid_lft > max_valid)
2476 			valid_lft = max_valid;
2477 
2478 		if (prefered_lft > max_prefered)
2479 			prefered_lft = max_prefered;
2480 
2481 		spin_lock(&ift->lock);
2482 		flags = ift->flags;
2483 		ift->valid_lft = valid_lft;
2484 		ift->prefered_lft = prefered_lft;
2485 		ift->tstamp = now;
2486 		if (prefered_lft > 0)
2487 			ift->flags &= ~IFA_F_DEPRECATED;
2488 
2489 		spin_unlock(&ift->lock);
2490 		if (!(flags&IFA_F_TENTATIVE))
2491 			ipv6_ifa_notify(0, ift);
2492 	}
2493 
2494 	if ((create || list_empty(&idev->tempaddr_list)) &&
2495 	    idev->cnf.use_tempaddr > 0) {
2496 		/* When a new public address is created as described
2497 		 * in [ADDRCONF], also create a new temporary address.
2498 		 * Also create a temporary address if it's enabled but
2499 		 * no temporary address currently exists.
2500 		 */
2501 		read_unlock_bh(&idev->lock);
2502 		ipv6_create_tempaddr(ifp, NULL, false);
2503 	} else {
2504 		read_unlock_bh(&idev->lock);
2505 	}
2506 }
2507 
2508 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2509 {
2510 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2511 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2512 }
2513 
2514 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2515 				 const struct prefix_info *pinfo,
2516 				 struct inet6_dev *in6_dev,
2517 				 const struct in6_addr *addr, int addr_type,
2518 				 u32 addr_flags, bool sllao, bool tokenized,
2519 				 __u32 valid_lft, u32 prefered_lft)
2520 {
2521 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2522 	int create = 0, update_lft = 0;
2523 
2524 	if (!ifp && valid_lft) {
2525 		int max_addresses = in6_dev->cnf.max_addresses;
2526 		struct ifa6_config cfg = {
2527 			.pfx = addr,
2528 			.plen = pinfo->prefix_len,
2529 			.ifa_flags = addr_flags,
2530 			.valid_lft = valid_lft,
2531 			.preferred_lft = prefered_lft,
2532 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2533 		};
2534 
2535 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2536 		if ((net->ipv6.devconf_all->optimistic_dad ||
2537 		     in6_dev->cnf.optimistic_dad) &&
2538 		    !net->ipv6.devconf_all->forwarding && sllao)
2539 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2540 #endif
2541 
2542 		/* Do not allow to create too much of autoconfigured
2543 		 * addresses; this would be too easy way to crash kernel.
2544 		 */
2545 		if (!max_addresses ||
2546 		    ipv6_count_addresses(in6_dev) < max_addresses)
2547 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2548 
2549 		if (IS_ERR_OR_NULL(ifp))
2550 			return -1;
2551 
2552 		create = 1;
2553 		spin_lock_bh(&ifp->lock);
2554 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2555 		ifp->cstamp = jiffies;
2556 		ifp->tokenized = tokenized;
2557 		spin_unlock_bh(&ifp->lock);
2558 		addrconf_dad_start(ifp);
2559 	}
2560 
2561 	if (ifp) {
2562 		u32 flags;
2563 		unsigned long now;
2564 		u32 stored_lft;
2565 
2566 		/* update lifetime (RFC2462 5.5.3 e) */
2567 		spin_lock_bh(&ifp->lock);
2568 		now = jiffies;
2569 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2570 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2571 		else
2572 			stored_lft = 0;
2573 		if (!create && stored_lft) {
2574 			const u32 minimum_lft = min_t(u32,
2575 				stored_lft, MIN_VALID_LIFETIME);
2576 			valid_lft = max(valid_lft, minimum_lft);
2577 
2578 			/* RFC4862 Section 5.5.3e:
2579 			 * "Note that the preferred lifetime of the
2580 			 *  corresponding address is always reset to
2581 			 *  the Preferred Lifetime in the received
2582 			 *  Prefix Information option, regardless of
2583 			 *  whether the valid lifetime is also reset or
2584 			 *  ignored."
2585 			 *
2586 			 * So we should always update prefered_lft here.
2587 			 */
2588 			update_lft = 1;
2589 		}
2590 
2591 		if (update_lft) {
2592 			ifp->valid_lft = valid_lft;
2593 			ifp->prefered_lft = prefered_lft;
2594 			ifp->tstamp = now;
2595 			flags = ifp->flags;
2596 			ifp->flags &= ~IFA_F_DEPRECATED;
2597 			spin_unlock_bh(&ifp->lock);
2598 
2599 			if (!(flags&IFA_F_TENTATIVE))
2600 				ipv6_ifa_notify(0, ifp);
2601 		} else
2602 			spin_unlock_bh(&ifp->lock);
2603 
2604 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2605 				 create, now);
2606 
2607 		in6_ifa_put(ifp);
2608 		addrconf_verify();
2609 	}
2610 
2611 	return 0;
2612 }
2613 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2614 
2615 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2616 {
2617 	struct prefix_info *pinfo;
2618 	__u32 valid_lft;
2619 	__u32 prefered_lft;
2620 	int addr_type, err;
2621 	u32 addr_flags = 0;
2622 	struct inet6_dev *in6_dev;
2623 	struct net *net = dev_net(dev);
2624 
2625 	pinfo = (struct prefix_info *) opt;
2626 
2627 	if (len < sizeof(struct prefix_info)) {
2628 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2629 		return;
2630 	}
2631 
2632 	/*
2633 	 *	Validation checks ([ADDRCONF], page 19)
2634 	 */
2635 
2636 	addr_type = ipv6_addr_type(&pinfo->prefix);
2637 
2638 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2639 		return;
2640 
2641 	valid_lft = ntohl(pinfo->valid);
2642 	prefered_lft = ntohl(pinfo->prefered);
2643 
2644 	if (prefered_lft > valid_lft) {
2645 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2646 		return;
2647 	}
2648 
2649 	in6_dev = in6_dev_get(dev);
2650 
2651 	if (!in6_dev) {
2652 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2653 				    dev->name);
2654 		return;
2655 	}
2656 
2657 	/*
2658 	 *	Two things going on here:
2659 	 *	1) Add routes for on-link prefixes
2660 	 *	2) Configure prefixes with the auto flag set
2661 	 */
2662 
2663 	if (pinfo->onlink) {
2664 		struct fib6_info *rt;
2665 		unsigned long rt_expires;
2666 
2667 		/* Avoid arithmetic overflow. Really, we could
2668 		 * save rt_expires in seconds, likely valid_lft,
2669 		 * but it would require division in fib gc, that it
2670 		 * not good.
2671 		 */
2672 		if (HZ > USER_HZ)
2673 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2674 		else
2675 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2676 
2677 		if (addrconf_finite_timeout(rt_expires))
2678 			rt_expires *= HZ;
2679 
2680 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2681 					       pinfo->prefix_len,
2682 					       dev,
2683 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2684 					       RTF_GATEWAY | RTF_DEFAULT);
2685 
2686 		if (rt) {
2687 			/* Autoconf prefix route */
2688 			if (valid_lft == 0) {
2689 				ip6_del_rt(net, rt);
2690 				rt = NULL;
2691 			} else if (addrconf_finite_timeout(rt_expires)) {
2692 				/* not infinity */
2693 				fib6_set_expires(rt, jiffies + rt_expires);
2694 			} else {
2695 				fib6_clean_expires(rt);
2696 			}
2697 		} else if (valid_lft) {
2698 			clock_t expires = 0;
2699 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2700 			if (addrconf_finite_timeout(rt_expires)) {
2701 				/* not infinity */
2702 				flags |= RTF_EXPIRES;
2703 				expires = jiffies_to_clock_t(rt_expires);
2704 			}
2705 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2706 					      0, dev, expires, flags,
2707 					      GFP_ATOMIC);
2708 		}
2709 		fib6_info_release(rt);
2710 	}
2711 
2712 	/* Try to figure out our local address for this prefix */
2713 
2714 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2715 		struct in6_addr addr;
2716 		bool tokenized = false, dev_addr_generated = false;
2717 
2718 		if (pinfo->prefix_len == 64) {
2719 			memcpy(&addr, &pinfo->prefix, 8);
2720 
2721 			if (!ipv6_addr_any(&in6_dev->token)) {
2722 				read_lock_bh(&in6_dev->lock);
2723 				memcpy(addr.s6_addr + 8,
2724 				       in6_dev->token.s6_addr + 8, 8);
2725 				read_unlock_bh(&in6_dev->lock);
2726 				tokenized = true;
2727 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2728 				   !ipv6_generate_stable_address(&addr, 0,
2729 								 in6_dev)) {
2730 				addr_flags |= IFA_F_STABLE_PRIVACY;
2731 				goto ok;
2732 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2733 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2734 				goto put;
2735 			} else {
2736 				dev_addr_generated = true;
2737 			}
2738 			goto ok;
2739 		}
2740 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2741 				    pinfo->prefix_len);
2742 		goto put;
2743 
2744 ok:
2745 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2746 						   &addr, addr_type,
2747 						   addr_flags, sllao,
2748 						   tokenized, valid_lft,
2749 						   prefered_lft);
2750 		if (err)
2751 			goto put;
2752 
2753 		/* Ignore error case here because previous prefix add addr was
2754 		 * successful which will be notified.
2755 		 */
2756 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2757 					      addr_type, addr_flags, sllao,
2758 					      tokenized, valid_lft,
2759 					      prefered_lft,
2760 					      dev_addr_generated);
2761 	}
2762 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2763 put:
2764 	in6_dev_put(in6_dev);
2765 }
2766 
2767 /*
2768  *	Set destination address.
2769  *	Special case for SIT interfaces where we create a new "virtual"
2770  *	device.
2771  */
2772 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2773 {
2774 	struct in6_ifreq ireq;
2775 	struct net_device *dev;
2776 	int err = -EINVAL;
2777 
2778 	rtnl_lock();
2779 
2780 	err = -EFAULT;
2781 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2782 		goto err_exit;
2783 
2784 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2785 
2786 	err = -ENODEV;
2787 	if (!dev)
2788 		goto err_exit;
2789 
2790 #if IS_ENABLED(CONFIG_IPV6_SIT)
2791 	if (dev->type == ARPHRD_SIT) {
2792 		const struct net_device_ops *ops = dev->netdev_ops;
2793 		struct ifreq ifr;
2794 		struct ip_tunnel_parm p;
2795 
2796 		err = -EADDRNOTAVAIL;
2797 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2798 			goto err_exit;
2799 
2800 		memset(&p, 0, sizeof(p));
2801 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2802 		p.iph.saddr = 0;
2803 		p.iph.version = 4;
2804 		p.iph.ihl = 5;
2805 		p.iph.protocol = IPPROTO_IPV6;
2806 		p.iph.ttl = 64;
2807 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2808 
2809 		if (ops->ndo_do_ioctl) {
2810 			mm_segment_t oldfs = get_fs();
2811 
2812 			set_fs(KERNEL_DS);
2813 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2814 			set_fs(oldfs);
2815 		} else
2816 			err = -EOPNOTSUPP;
2817 
2818 		if (err == 0) {
2819 			err = -ENOBUFS;
2820 			dev = __dev_get_by_name(net, p.name);
2821 			if (!dev)
2822 				goto err_exit;
2823 			err = dev_open(dev);
2824 		}
2825 	}
2826 #endif
2827 
2828 err_exit:
2829 	rtnl_unlock();
2830 	return err;
2831 }
2832 
2833 static int ipv6_mc_config(struct sock *sk, bool join,
2834 			  const struct in6_addr *addr, int ifindex)
2835 {
2836 	int ret;
2837 
2838 	ASSERT_RTNL();
2839 
2840 	lock_sock(sk);
2841 	if (join)
2842 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2843 	else
2844 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2845 	release_sock(sk);
2846 
2847 	return ret;
2848 }
2849 
2850 /*
2851  *	Manual configuration of address on an interface
2852  */
2853 static int inet6_addr_add(struct net *net, int ifindex,
2854 			  struct ifa6_config *cfg,
2855 			  struct netlink_ext_ack *extack)
2856 {
2857 	struct inet6_ifaddr *ifp;
2858 	struct inet6_dev *idev;
2859 	struct net_device *dev;
2860 	unsigned long timeout;
2861 	clock_t expires;
2862 	u32 flags;
2863 
2864 	ASSERT_RTNL();
2865 
2866 	if (cfg->plen > 128)
2867 		return -EINVAL;
2868 
2869 	/* check the lifetime */
2870 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2871 		return -EINVAL;
2872 
2873 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2874 		return -EINVAL;
2875 
2876 	dev = __dev_get_by_index(net, ifindex);
2877 	if (!dev)
2878 		return -ENODEV;
2879 
2880 	idev = addrconf_add_dev(dev);
2881 	if (IS_ERR(idev))
2882 		return PTR_ERR(idev);
2883 
2884 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2885 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2886 					 true, cfg->pfx, ifindex);
2887 
2888 		if (ret < 0)
2889 			return ret;
2890 	}
2891 
2892 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2893 
2894 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2895 	if (addrconf_finite_timeout(timeout)) {
2896 		expires = jiffies_to_clock_t(timeout * HZ);
2897 		cfg->valid_lft = timeout;
2898 		flags = RTF_EXPIRES;
2899 	} else {
2900 		expires = 0;
2901 		flags = 0;
2902 		cfg->ifa_flags |= IFA_F_PERMANENT;
2903 	}
2904 
2905 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2906 	if (addrconf_finite_timeout(timeout)) {
2907 		if (timeout == 0)
2908 			cfg->ifa_flags |= IFA_F_DEPRECATED;
2909 		cfg->preferred_lft = timeout;
2910 	}
2911 
2912 	ifp = ipv6_add_addr(idev, cfg, true, extack);
2913 	if (!IS_ERR(ifp)) {
2914 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2915 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2916 					      ifp->rt_priority, dev, expires,
2917 					      flags, GFP_KERNEL);
2918 		}
2919 
2920 		/* Send a netlink notification if DAD is enabled and
2921 		 * optimistic flag is not set
2922 		 */
2923 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2924 			ipv6_ifa_notify(0, ifp);
2925 		/*
2926 		 * Note that section 3.1 of RFC 4429 indicates
2927 		 * that the Optimistic flag should not be set for
2928 		 * manually configured addresses
2929 		 */
2930 		addrconf_dad_start(ifp);
2931 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2932 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
2933 					 cfg->preferred_lft, true, jiffies);
2934 		in6_ifa_put(ifp);
2935 		addrconf_verify_rtnl();
2936 		return 0;
2937 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2938 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2939 			       cfg->pfx, ifindex);
2940 	}
2941 
2942 	return PTR_ERR(ifp);
2943 }
2944 
2945 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2946 			  const struct in6_addr *pfx, unsigned int plen)
2947 {
2948 	struct inet6_ifaddr *ifp;
2949 	struct inet6_dev *idev;
2950 	struct net_device *dev;
2951 
2952 	if (plen > 128)
2953 		return -EINVAL;
2954 
2955 	dev = __dev_get_by_index(net, ifindex);
2956 	if (!dev)
2957 		return -ENODEV;
2958 
2959 	idev = __in6_dev_get(dev);
2960 	if (!idev)
2961 		return -ENXIO;
2962 
2963 	read_lock_bh(&idev->lock);
2964 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2965 		if (ifp->prefix_len == plen &&
2966 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2967 			in6_ifa_hold(ifp);
2968 			read_unlock_bh(&idev->lock);
2969 
2970 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
2971 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
2972 				manage_tempaddrs(idev, ifp, 0, 0, false,
2973 						 jiffies);
2974 			ipv6_del_addr(ifp);
2975 			addrconf_verify_rtnl();
2976 			if (ipv6_addr_is_multicast(pfx)) {
2977 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2978 					       false, pfx, dev->ifindex);
2979 			}
2980 			return 0;
2981 		}
2982 	}
2983 	read_unlock_bh(&idev->lock);
2984 	return -EADDRNOTAVAIL;
2985 }
2986 
2987 
2988 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2989 {
2990 	struct ifa6_config cfg = {
2991 		.ifa_flags = IFA_F_PERMANENT,
2992 		.preferred_lft = INFINITY_LIFE_TIME,
2993 		.valid_lft = INFINITY_LIFE_TIME,
2994 	};
2995 	struct in6_ifreq ireq;
2996 	int err;
2997 
2998 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2999 		return -EPERM;
3000 
3001 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3002 		return -EFAULT;
3003 
3004 	cfg.pfx = &ireq.ifr6_addr;
3005 	cfg.plen = ireq.ifr6_prefixlen;
3006 
3007 	rtnl_lock();
3008 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3009 	rtnl_unlock();
3010 	return err;
3011 }
3012 
3013 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3014 {
3015 	struct in6_ifreq ireq;
3016 	int err;
3017 
3018 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3019 		return -EPERM;
3020 
3021 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3022 		return -EFAULT;
3023 
3024 	rtnl_lock();
3025 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3026 			     ireq.ifr6_prefixlen);
3027 	rtnl_unlock();
3028 	return err;
3029 }
3030 
3031 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3032 		     int plen, int scope)
3033 {
3034 	struct inet6_ifaddr *ifp;
3035 	struct ifa6_config cfg = {
3036 		.pfx = addr,
3037 		.plen = plen,
3038 		.ifa_flags = IFA_F_PERMANENT,
3039 		.valid_lft = INFINITY_LIFE_TIME,
3040 		.preferred_lft = INFINITY_LIFE_TIME,
3041 		.scope = scope
3042 	};
3043 
3044 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3045 	if (!IS_ERR(ifp)) {
3046 		spin_lock_bh(&ifp->lock);
3047 		ifp->flags &= ~IFA_F_TENTATIVE;
3048 		spin_unlock_bh(&ifp->lock);
3049 		rt_genid_bump_ipv6(dev_net(idev->dev));
3050 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3051 		in6_ifa_put(ifp);
3052 	}
3053 }
3054 
3055 #if IS_ENABLED(CONFIG_IPV6_SIT)
3056 static void sit_add_v4_addrs(struct inet6_dev *idev)
3057 {
3058 	struct in6_addr addr;
3059 	struct net_device *dev;
3060 	struct net *net = dev_net(idev->dev);
3061 	int scope, plen;
3062 	u32 pflags = 0;
3063 
3064 	ASSERT_RTNL();
3065 
3066 	memset(&addr, 0, sizeof(struct in6_addr));
3067 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3068 
3069 	if (idev->dev->flags&IFF_POINTOPOINT) {
3070 		addr.s6_addr32[0] = htonl(0xfe800000);
3071 		scope = IFA_LINK;
3072 		plen = 64;
3073 	} else {
3074 		scope = IPV6_ADDR_COMPATv4;
3075 		plen = 96;
3076 		pflags |= RTF_NONEXTHOP;
3077 	}
3078 
3079 	if (addr.s6_addr32[3]) {
3080 		add_addr(idev, &addr, plen, scope);
3081 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3082 				      GFP_KERNEL);
3083 		return;
3084 	}
3085 
3086 	for_each_netdev(net, dev) {
3087 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3088 		if (in_dev && (dev->flags & IFF_UP)) {
3089 			struct in_ifaddr *ifa;
3090 
3091 			int flag = scope;
3092 
3093 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3094 
3095 				addr.s6_addr32[3] = ifa->ifa_local;
3096 
3097 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3098 					continue;
3099 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3100 					if (idev->dev->flags&IFF_POINTOPOINT)
3101 						continue;
3102 					flag |= IFA_HOST;
3103 				}
3104 
3105 				add_addr(idev, &addr, plen, flag);
3106 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3107 						      0, pflags, GFP_KERNEL);
3108 			}
3109 		}
3110 	}
3111 }
3112 #endif
3113 
3114 static void init_loopback(struct net_device *dev)
3115 {
3116 	struct inet6_dev  *idev;
3117 
3118 	/* ::1 */
3119 
3120 	ASSERT_RTNL();
3121 
3122 	idev = ipv6_find_idev(dev);
3123 	if (!idev) {
3124 		pr_debug("%s: add_dev failed\n", __func__);
3125 		return;
3126 	}
3127 
3128 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3129 }
3130 
3131 void addrconf_add_linklocal(struct inet6_dev *idev,
3132 			    const struct in6_addr *addr, u32 flags)
3133 {
3134 	struct ifa6_config cfg = {
3135 		.pfx = addr,
3136 		.plen = 64,
3137 		.ifa_flags = flags | IFA_F_PERMANENT,
3138 		.valid_lft = INFINITY_LIFE_TIME,
3139 		.preferred_lft = INFINITY_LIFE_TIME,
3140 		.scope = IFA_LINK
3141 	};
3142 	struct inet6_ifaddr *ifp;
3143 
3144 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3145 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3146 	     idev->cnf.optimistic_dad) &&
3147 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3148 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3149 #endif
3150 
3151 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3152 	if (!IS_ERR(ifp)) {
3153 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3154 				      0, 0, GFP_ATOMIC);
3155 		addrconf_dad_start(ifp);
3156 		in6_ifa_put(ifp);
3157 	}
3158 }
3159 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3160 
3161 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3162 {
3163 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3164 		return true;
3165 
3166 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3167 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3168 		return true;
3169 
3170 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3171 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3172 		return true;
3173 
3174 	return false;
3175 }
3176 
3177 static int ipv6_generate_stable_address(struct in6_addr *address,
3178 					u8 dad_count,
3179 					const struct inet6_dev *idev)
3180 {
3181 	static DEFINE_SPINLOCK(lock);
3182 	static __u32 digest[SHA_DIGEST_WORDS];
3183 	static __u32 workspace[SHA_WORKSPACE_WORDS];
3184 
3185 	static union {
3186 		char __data[SHA_MESSAGE_BYTES];
3187 		struct {
3188 			struct in6_addr secret;
3189 			__be32 prefix[2];
3190 			unsigned char hwaddr[MAX_ADDR_LEN];
3191 			u8 dad_count;
3192 		} __packed;
3193 	} data;
3194 
3195 	struct in6_addr secret;
3196 	struct in6_addr temp;
3197 	struct net *net = dev_net(idev->dev);
3198 
3199 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3200 
3201 	if (idev->cnf.stable_secret.initialized)
3202 		secret = idev->cnf.stable_secret.secret;
3203 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3204 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3205 	else
3206 		return -1;
3207 
3208 retry:
3209 	spin_lock_bh(&lock);
3210 
3211 	sha_init(digest);
3212 	memset(&data, 0, sizeof(data));
3213 	memset(workspace, 0, sizeof(workspace));
3214 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3215 	data.prefix[0] = address->s6_addr32[0];
3216 	data.prefix[1] = address->s6_addr32[1];
3217 	data.secret = secret;
3218 	data.dad_count = dad_count;
3219 
3220 	sha_transform(digest, data.__data, workspace);
3221 
3222 	temp = *address;
3223 	temp.s6_addr32[2] = (__force __be32)digest[0];
3224 	temp.s6_addr32[3] = (__force __be32)digest[1];
3225 
3226 	spin_unlock_bh(&lock);
3227 
3228 	if (ipv6_reserved_interfaceid(temp)) {
3229 		dad_count++;
3230 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3231 			return -1;
3232 		goto retry;
3233 	}
3234 
3235 	*address = temp;
3236 	return 0;
3237 }
3238 
3239 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3240 {
3241 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3242 
3243 	if (s->initialized)
3244 		return;
3245 	s = &idev->cnf.stable_secret;
3246 	get_random_bytes(&s->secret, sizeof(s->secret));
3247 	s->initialized = true;
3248 }
3249 
3250 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3251 {
3252 	struct in6_addr addr;
3253 
3254 	/* no link local addresses on L3 master devices */
3255 	if (netif_is_l3_master(idev->dev))
3256 		return;
3257 
3258 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3259 
3260 	switch (idev->cnf.addr_gen_mode) {
3261 	case IN6_ADDR_GEN_MODE_RANDOM:
3262 		ipv6_gen_mode_random_init(idev);
3263 		/* fallthrough */
3264 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3265 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3266 			addrconf_add_linklocal(idev, &addr,
3267 					       IFA_F_STABLE_PRIVACY);
3268 		else if (prefix_route)
3269 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3270 					      0, 0, GFP_KERNEL);
3271 		break;
3272 	case IN6_ADDR_GEN_MODE_EUI64:
3273 		/* addrconf_add_linklocal also adds a prefix_route and we
3274 		 * only need to care about prefix routes if ipv6_generate_eui64
3275 		 * couldn't generate one.
3276 		 */
3277 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3278 			addrconf_add_linklocal(idev, &addr, 0);
3279 		else if (prefix_route)
3280 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3281 					      0, 0, GFP_KERNEL);
3282 		break;
3283 	case IN6_ADDR_GEN_MODE_NONE:
3284 	default:
3285 		/* will not add any link local address */
3286 		break;
3287 	}
3288 }
3289 
3290 static void addrconf_dev_config(struct net_device *dev)
3291 {
3292 	struct inet6_dev *idev;
3293 
3294 	ASSERT_RTNL();
3295 
3296 	if ((dev->type != ARPHRD_ETHER) &&
3297 	    (dev->type != ARPHRD_FDDI) &&
3298 	    (dev->type != ARPHRD_ARCNET) &&
3299 	    (dev->type != ARPHRD_INFINIBAND) &&
3300 	    (dev->type != ARPHRD_IEEE1394) &&
3301 	    (dev->type != ARPHRD_TUNNEL6) &&
3302 	    (dev->type != ARPHRD_6LOWPAN) &&
3303 	    (dev->type != ARPHRD_IP6GRE) &&
3304 	    (dev->type != ARPHRD_IPGRE) &&
3305 	    (dev->type != ARPHRD_TUNNEL) &&
3306 	    (dev->type != ARPHRD_NONE) &&
3307 	    (dev->type != ARPHRD_RAWIP)) {
3308 		/* Alas, we support only Ethernet autoconfiguration. */
3309 		return;
3310 	}
3311 
3312 	idev = addrconf_add_dev(dev);
3313 	if (IS_ERR(idev))
3314 		return;
3315 
3316 	/* this device type has no EUI support */
3317 	if (dev->type == ARPHRD_NONE &&
3318 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3319 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3320 
3321 	addrconf_addr_gen(idev, false);
3322 }
3323 
3324 #if IS_ENABLED(CONFIG_IPV6_SIT)
3325 static void addrconf_sit_config(struct net_device *dev)
3326 {
3327 	struct inet6_dev *idev;
3328 
3329 	ASSERT_RTNL();
3330 
3331 	/*
3332 	 * Configure the tunnel with one of our IPv4
3333 	 * addresses... we should configure all of
3334 	 * our v4 addrs in the tunnel
3335 	 */
3336 
3337 	idev = ipv6_find_idev(dev);
3338 	if (!idev) {
3339 		pr_debug("%s: add_dev failed\n", __func__);
3340 		return;
3341 	}
3342 
3343 	if (dev->priv_flags & IFF_ISATAP) {
3344 		addrconf_addr_gen(idev, false);
3345 		return;
3346 	}
3347 
3348 	sit_add_v4_addrs(idev);
3349 
3350 	if (dev->flags&IFF_POINTOPOINT)
3351 		addrconf_add_mroute(dev);
3352 }
3353 #endif
3354 
3355 #if IS_ENABLED(CONFIG_NET_IPGRE)
3356 static void addrconf_gre_config(struct net_device *dev)
3357 {
3358 	struct inet6_dev *idev;
3359 
3360 	ASSERT_RTNL();
3361 
3362 	idev = ipv6_find_idev(dev);
3363 	if (!idev) {
3364 		pr_debug("%s: add_dev failed\n", __func__);
3365 		return;
3366 	}
3367 
3368 	addrconf_addr_gen(idev, true);
3369 	if (dev->flags & IFF_POINTOPOINT)
3370 		addrconf_add_mroute(dev);
3371 }
3372 #endif
3373 
3374 static int fixup_permanent_addr(struct net *net,
3375 				struct inet6_dev *idev,
3376 				struct inet6_ifaddr *ifp)
3377 {
3378 	/* !fib6_node means the host route was removed from the
3379 	 * FIB, for example, if 'lo' device is taken down. In that
3380 	 * case regenerate the host route.
3381 	 */
3382 	if (!ifp->rt || !ifp->rt->fib6_node) {
3383 		struct fib6_info *f6i, *prev;
3384 
3385 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3386 					 GFP_ATOMIC);
3387 		if (IS_ERR(f6i))
3388 			return PTR_ERR(f6i);
3389 
3390 		/* ifp->rt can be accessed outside of rtnl */
3391 		spin_lock(&ifp->lock);
3392 		prev = ifp->rt;
3393 		ifp->rt = f6i;
3394 		spin_unlock(&ifp->lock);
3395 
3396 		fib6_info_release(prev);
3397 	}
3398 
3399 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3400 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3401 				      ifp->rt_priority, idev->dev, 0, 0,
3402 				      GFP_ATOMIC);
3403 	}
3404 
3405 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3406 		addrconf_dad_start(ifp);
3407 
3408 	return 0;
3409 }
3410 
3411 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3412 {
3413 	struct inet6_ifaddr *ifp, *tmp;
3414 	struct inet6_dev *idev;
3415 
3416 	idev = __in6_dev_get(dev);
3417 	if (!idev)
3418 		return;
3419 
3420 	write_lock_bh(&idev->lock);
3421 
3422 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3423 		if ((ifp->flags & IFA_F_PERMANENT) &&
3424 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3425 			write_unlock_bh(&idev->lock);
3426 			in6_ifa_hold(ifp);
3427 			ipv6_del_addr(ifp);
3428 			write_lock_bh(&idev->lock);
3429 
3430 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3431 					     idev->dev->name, &ifp->addr);
3432 		}
3433 	}
3434 
3435 	write_unlock_bh(&idev->lock);
3436 }
3437 
3438 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3439 			   void *ptr)
3440 {
3441 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3442 	struct netdev_notifier_changeupper_info *info;
3443 	struct inet6_dev *idev = __in6_dev_get(dev);
3444 	struct net *net = dev_net(dev);
3445 	int run_pending = 0;
3446 	int err;
3447 
3448 	switch (event) {
3449 	case NETDEV_REGISTER:
3450 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3451 			idev = ipv6_add_dev(dev);
3452 			if (IS_ERR(idev))
3453 				return notifier_from_errno(PTR_ERR(idev));
3454 		}
3455 		break;
3456 
3457 	case NETDEV_CHANGEMTU:
3458 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3459 		if (dev->mtu < IPV6_MIN_MTU) {
3460 			addrconf_ifdown(dev, dev != net->loopback_dev);
3461 			break;
3462 		}
3463 
3464 		if (idev) {
3465 			rt6_mtu_change(dev, dev->mtu);
3466 			idev->cnf.mtu6 = dev->mtu;
3467 			break;
3468 		}
3469 
3470 		/* allocate new idev */
3471 		idev = ipv6_add_dev(dev);
3472 		if (IS_ERR(idev))
3473 			break;
3474 
3475 		/* device is still not ready */
3476 		if (!(idev->if_flags & IF_READY))
3477 			break;
3478 
3479 		run_pending = 1;
3480 
3481 		/* fall through */
3482 
3483 	case NETDEV_UP:
3484 	case NETDEV_CHANGE:
3485 		if (dev->flags & IFF_SLAVE)
3486 			break;
3487 
3488 		if (idev && idev->cnf.disable_ipv6)
3489 			break;
3490 
3491 		if (event == NETDEV_UP) {
3492 			/* restore routes for permanent addresses */
3493 			addrconf_permanent_addr(net, dev);
3494 
3495 			if (!addrconf_link_ready(dev)) {
3496 				/* device is not ready yet. */
3497 				pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3498 					dev->name);
3499 				break;
3500 			}
3501 
3502 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3503 				idev = ipv6_add_dev(dev);
3504 
3505 			if (!IS_ERR_OR_NULL(idev)) {
3506 				idev->if_flags |= IF_READY;
3507 				run_pending = 1;
3508 			}
3509 		} else if (event == NETDEV_CHANGE) {
3510 			if (!addrconf_link_ready(dev)) {
3511 				/* device is still not ready. */
3512 				rt6_sync_down_dev(dev, event);
3513 				break;
3514 			}
3515 
3516 			if (idev) {
3517 				if (idev->if_flags & IF_READY) {
3518 					/* device is already configured -
3519 					 * but resend MLD reports, we might
3520 					 * have roamed and need to update
3521 					 * multicast snooping switches
3522 					 */
3523 					ipv6_mc_up(idev);
3524 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3525 					break;
3526 				}
3527 				idev->if_flags |= IF_READY;
3528 			}
3529 
3530 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3531 				dev->name);
3532 
3533 			run_pending = 1;
3534 		}
3535 
3536 		switch (dev->type) {
3537 #if IS_ENABLED(CONFIG_IPV6_SIT)
3538 		case ARPHRD_SIT:
3539 			addrconf_sit_config(dev);
3540 			break;
3541 #endif
3542 #if IS_ENABLED(CONFIG_NET_IPGRE)
3543 		case ARPHRD_IPGRE:
3544 			addrconf_gre_config(dev);
3545 			break;
3546 #endif
3547 		case ARPHRD_LOOPBACK:
3548 			init_loopback(dev);
3549 			break;
3550 
3551 		default:
3552 			addrconf_dev_config(dev);
3553 			break;
3554 		}
3555 
3556 		if (!IS_ERR_OR_NULL(idev)) {
3557 			if (run_pending)
3558 				addrconf_dad_run(idev);
3559 
3560 			/* Device has an address by now */
3561 			rt6_sync_up(dev, RTNH_F_DEAD);
3562 
3563 			/*
3564 			 * If the MTU changed during the interface down,
3565 			 * when the interface up, the changed MTU must be
3566 			 * reflected in the idev as well as routers.
3567 			 */
3568 			if (idev->cnf.mtu6 != dev->mtu &&
3569 			    dev->mtu >= IPV6_MIN_MTU) {
3570 				rt6_mtu_change(dev, dev->mtu);
3571 				idev->cnf.mtu6 = dev->mtu;
3572 			}
3573 			idev->tstamp = jiffies;
3574 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3575 
3576 			/*
3577 			 * If the changed mtu during down is lower than
3578 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3579 			 */
3580 			if (dev->mtu < IPV6_MIN_MTU)
3581 				addrconf_ifdown(dev, dev != net->loopback_dev);
3582 		}
3583 		break;
3584 
3585 	case NETDEV_DOWN:
3586 	case NETDEV_UNREGISTER:
3587 		/*
3588 		 *	Remove all addresses from this interface.
3589 		 */
3590 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3591 		break;
3592 
3593 	case NETDEV_CHANGENAME:
3594 		if (idev) {
3595 			snmp6_unregister_dev(idev);
3596 			addrconf_sysctl_unregister(idev);
3597 			err = addrconf_sysctl_register(idev);
3598 			if (err)
3599 				return notifier_from_errno(err);
3600 			err = snmp6_register_dev(idev);
3601 			if (err) {
3602 				addrconf_sysctl_unregister(idev);
3603 				return notifier_from_errno(err);
3604 			}
3605 		}
3606 		break;
3607 
3608 	case NETDEV_PRE_TYPE_CHANGE:
3609 	case NETDEV_POST_TYPE_CHANGE:
3610 		if (idev)
3611 			addrconf_type_change(dev, event);
3612 		break;
3613 
3614 	case NETDEV_CHANGEUPPER:
3615 		info = ptr;
3616 
3617 		/* flush all routes if dev is linked to or unlinked from
3618 		 * an L3 master device (e.g., VRF)
3619 		 */
3620 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3621 			addrconf_ifdown(dev, 0);
3622 	}
3623 
3624 	return NOTIFY_OK;
3625 }
3626 
3627 /*
3628  *	addrconf module should be notified of a device going up
3629  */
3630 static struct notifier_block ipv6_dev_notf = {
3631 	.notifier_call = addrconf_notify,
3632 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3633 };
3634 
3635 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3636 {
3637 	struct inet6_dev *idev;
3638 	ASSERT_RTNL();
3639 
3640 	idev = __in6_dev_get(dev);
3641 
3642 	if (event == NETDEV_POST_TYPE_CHANGE)
3643 		ipv6_mc_remap(idev);
3644 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3645 		ipv6_mc_unmap(idev);
3646 }
3647 
3648 static bool addr_is_local(const struct in6_addr *addr)
3649 {
3650 	return ipv6_addr_type(addr) &
3651 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3652 }
3653 
3654 static int addrconf_ifdown(struct net_device *dev, int how)
3655 {
3656 	unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3657 	struct net *net = dev_net(dev);
3658 	struct inet6_dev *idev;
3659 	struct inet6_ifaddr *ifa, *tmp;
3660 	bool keep_addr = false;
3661 	int state, i;
3662 
3663 	ASSERT_RTNL();
3664 
3665 	rt6_disable_ip(dev, event);
3666 
3667 	idev = __in6_dev_get(dev);
3668 	if (!idev)
3669 		return -ENODEV;
3670 
3671 	/*
3672 	 * Step 1: remove reference to ipv6 device from parent device.
3673 	 *	   Do not dev_put!
3674 	 */
3675 	if (how) {
3676 		idev->dead = 1;
3677 
3678 		/* protected by rtnl_lock */
3679 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3680 
3681 		/* Step 1.5: remove snmp6 entry */
3682 		snmp6_unregister_dev(idev);
3683 
3684 	}
3685 
3686 	/* combine the user config with event to determine if permanent
3687 	 * addresses are to be removed from address hash table
3688 	 */
3689 	if (!how && !idev->cnf.disable_ipv6) {
3690 		/* aggregate the system setting and interface setting */
3691 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3692 
3693 		if (!_keep_addr)
3694 			_keep_addr = idev->cnf.keep_addr_on_down;
3695 
3696 		keep_addr = (_keep_addr > 0);
3697 	}
3698 
3699 	/* Step 2: clear hash table */
3700 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3701 		struct hlist_head *h = &inet6_addr_lst[i];
3702 
3703 		spin_lock_bh(&addrconf_hash_lock);
3704 restart:
3705 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3706 			if (ifa->idev == idev) {
3707 				addrconf_del_dad_work(ifa);
3708 				/* combined flag + permanent flag decide if
3709 				 * address is retained on a down event
3710 				 */
3711 				if (!keep_addr ||
3712 				    !(ifa->flags & IFA_F_PERMANENT) ||
3713 				    addr_is_local(&ifa->addr)) {
3714 					hlist_del_init_rcu(&ifa->addr_lst);
3715 					goto restart;
3716 				}
3717 			}
3718 		}
3719 		spin_unlock_bh(&addrconf_hash_lock);
3720 	}
3721 
3722 	write_lock_bh(&idev->lock);
3723 
3724 	addrconf_del_rs_timer(idev);
3725 
3726 	/* Step 2: clear flags for stateless addrconf */
3727 	if (!how)
3728 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3729 
3730 	/* Step 3: clear tempaddr list */
3731 	while (!list_empty(&idev->tempaddr_list)) {
3732 		ifa = list_first_entry(&idev->tempaddr_list,
3733 				       struct inet6_ifaddr, tmp_list);
3734 		list_del(&ifa->tmp_list);
3735 		write_unlock_bh(&idev->lock);
3736 		spin_lock_bh(&ifa->lock);
3737 
3738 		if (ifa->ifpub) {
3739 			in6_ifa_put(ifa->ifpub);
3740 			ifa->ifpub = NULL;
3741 		}
3742 		spin_unlock_bh(&ifa->lock);
3743 		in6_ifa_put(ifa);
3744 		write_lock_bh(&idev->lock);
3745 	}
3746 
3747 	list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3748 		struct fib6_info *rt = NULL;
3749 		bool keep;
3750 
3751 		addrconf_del_dad_work(ifa);
3752 
3753 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3754 			!addr_is_local(&ifa->addr);
3755 
3756 		write_unlock_bh(&idev->lock);
3757 		spin_lock_bh(&ifa->lock);
3758 
3759 		if (keep) {
3760 			/* set state to skip the notifier below */
3761 			state = INET6_IFADDR_STATE_DEAD;
3762 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3763 			if (!(ifa->flags & IFA_F_NODAD))
3764 				ifa->flags |= IFA_F_TENTATIVE;
3765 
3766 			rt = ifa->rt;
3767 			ifa->rt = NULL;
3768 		} else {
3769 			state = ifa->state;
3770 			ifa->state = INET6_IFADDR_STATE_DEAD;
3771 		}
3772 
3773 		spin_unlock_bh(&ifa->lock);
3774 
3775 		if (rt)
3776 			ip6_del_rt(net, rt);
3777 
3778 		if (state != INET6_IFADDR_STATE_DEAD) {
3779 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3780 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3781 		} else {
3782 			if (idev->cnf.forwarding)
3783 				addrconf_leave_anycast(ifa);
3784 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3785 		}
3786 
3787 		write_lock_bh(&idev->lock);
3788 		if (!keep) {
3789 			list_del_rcu(&ifa->if_list);
3790 			in6_ifa_put(ifa);
3791 		}
3792 	}
3793 
3794 	write_unlock_bh(&idev->lock);
3795 
3796 	/* Step 5: Discard anycast and multicast list */
3797 	if (how) {
3798 		ipv6_ac_destroy_dev(idev);
3799 		ipv6_mc_destroy_dev(idev);
3800 	} else {
3801 		ipv6_mc_down(idev);
3802 	}
3803 
3804 	idev->tstamp = jiffies;
3805 
3806 	/* Last: Shot the device (if unregistered) */
3807 	if (how) {
3808 		addrconf_sysctl_unregister(idev);
3809 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3810 		neigh_ifdown(&nd_tbl, dev);
3811 		in6_dev_put(idev);
3812 	}
3813 	return 0;
3814 }
3815 
3816 static void addrconf_rs_timer(struct timer_list *t)
3817 {
3818 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3819 	struct net_device *dev = idev->dev;
3820 	struct in6_addr lladdr;
3821 
3822 	write_lock(&idev->lock);
3823 	if (idev->dead || !(idev->if_flags & IF_READY))
3824 		goto out;
3825 
3826 	if (!ipv6_accept_ra(idev))
3827 		goto out;
3828 
3829 	/* Announcement received after solicitation was sent */
3830 	if (idev->if_flags & IF_RA_RCVD)
3831 		goto out;
3832 
3833 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3834 		write_unlock(&idev->lock);
3835 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3836 			ndisc_send_rs(dev, &lladdr,
3837 				      &in6addr_linklocal_allrouters);
3838 		else
3839 			goto put;
3840 
3841 		write_lock(&idev->lock);
3842 		idev->rs_interval = rfc3315_s14_backoff_update(
3843 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3844 		/* The wait after the last probe can be shorter */
3845 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3846 					     idev->cnf.rtr_solicits) ?
3847 				      idev->cnf.rtr_solicit_delay :
3848 				      idev->rs_interval);
3849 	} else {
3850 		/*
3851 		 * Note: we do not support deprecated "all on-link"
3852 		 * assumption any longer.
3853 		 */
3854 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3855 	}
3856 
3857 out:
3858 	write_unlock(&idev->lock);
3859 put:
3860 	in6_dev_put(idev);
3861 }
3862 
3863 /*
3864  *	Duplicate Address Detection
3865  */
3866 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3867 {
3868 	unsigned long rand_num;
3869 	struct inet6_dev *idev = ifp->idev;
3870 	u64 nonce;
3871 
3872 	if (ifp->flags & IFA_F_OPTIMISTIC)
3873 		rand_num = 0;
3874 	else
3875 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3876 
3877 	nonce = 0;
3878 	if (idev->cnf.enhanced_dad ||
3879 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3880 		do
3881 			get_random_bytes(&nonce, 6);
3882 		while (nonce == 0);
3883 	}
3884 	ifp->dad_nonce = nonce;
3885 	ifp->dad_probes = idev->cnf.dad_transmits;
3886 	addrconf_mod_dad_work(ifp, rand_num);
3887 }
3888 
3889 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3890 {
3891 	struct inet6_dev *idev = ifp->idev;
3892 	struct net_device *dev = idev->dev;
3893 	bool bump_id, notify = false;
3894 	struct net *net;
3895 
3896 	addrconf_join_solict(dev, &ifp->addr);
3897 
3898 	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3899 
3900 	read_lock_bh(&idev->lock);
3901 	spin_lock(&ifp->lock);
3902 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3903 		goto out;
3904 
3905 	net = dev_net(dev);
3906 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3907 	    (net->ipv6.devconf_all->accept_dad < 1 &&
3908 	     idev->cnf.accept_dad < 1) ||
3909 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3910 	    ifp->flags & IFA_F_NODAD) {
3911 		bool send_na = false;
3912 
3913 		if (ifp->flags & IFA_F_TENTATIVE &&
3914 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3915 			send_na = true;
3916 		bump_id = ifp->flags & IFA_F_TENTATIVE;
3917 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3918 		spin_unlock(&ifp->lock);
3919 		read_unlock_bh(&idev->lock);
3920 
3921 		addrconf_dad_completed(ifp, bump_id, send_na);
3922 		return;
3923 	}
3924 
3925 	if (!(idev->if_flags & IF_READY)) {
3926 		spin_unlock(&ifp->lock);
3927 		read_unlock_bh(&idev->lock);
3928 		/*
3929 		 * If the device is not ready:
3930 		 * - keep it tentative if it is a permanent address.
3931 		 * - otherwise, kill it.
3932 		 */
3933 		in6_ifa_hold(ifp);
3934 		addrconf_dad_stop(ifp, 0);
3935 		return;
3936 	}
3937 
3938 	/*
3939 	 * Optimistic nodes can start receiving
3940 	 * Frames right away
3941 	 */
3942 	if (ifp->flags & IFA_F_OPTIMISTIC) {
3943 		ip6_ins_rt(net, ifp->rt);
3944 		if (ipv6_use_optimistic_addr(net, idev)) {
3945 			/* Because optimistic nodes can use this address,
3946 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3947 			 */
3948 			notify = true;
3949 		}
3950 	}
3951 
3952 	addrconf_dad_kick(ifp);
3953 out:
3954 	spin_unlock(&ifp->lock);
3955 	read_unlock_bh(&idev->lock);
3956 	if (notify)
3957 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3958 }
3959 
3960 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3961 {
3962 	bool begin_dad = false;
3963 
3964 	spin_lock_bh(&ifp->lock);
3965 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3966 		ifp->state = INET6_IFADDR_STATE_PREDAD;
3967 		begin_dad = true;
3968 	}
3969 	spin_unlock_bh(&ifp->lock);
3970 
3971 	if (begin_dad)
3972 		addrconf_mod_dad_work(ifp, 0);
3973 }
3974 
3975 static void addrconf_dad_work(struct work_struct *w)
3976 {
3977 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3978 						struct inet6_ifaddr,
3979 						dad_work);
3980 	struct inet6_dev *idev = ifp->idev;
3981 	bool bump_id, disable_ipv6 = false;
3982 	struct in6_addr mcaddr;
3983 
3984 	enum {
3985 		DAD_PROCESS,
3986 		DAD_BEGIN,
3987 		DAD_ABORT,
3988 	} action = DAD_PROCESS;
3989 
3990 	rtnl_lock();
3991 
3992 	spin_lock_bh(&ifp->lock);
3993 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3994 		action = DAD_BEGIN;
3995 		ifp->state = INET6_IFADDR_STATE_DAD;
3996 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3997 		action = DAD_ABORT;
3998 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
3999 
4000 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4001 		     idev->cnf.accept_dad > 1) &&
4002 		    !idev->cnf.disable_ipv6 &&
4003 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4004 			struct in6_addr addr;
4005 
4006 			addr.s6_addr32[0] = htonl(0xfe800000);
4007 			addr.s6_addr32[1] = 0;
4008 
4009 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4010 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4011 				/* DAD failed for link-local based on MAC */
4012 				idev->cnf.disable_ipv6 = 1;
4013 
4014 				pr_info("%s: IPv6 being disabled!\n",
4015 					ifp->idev->dev->name);
4016 				disable_ipv6 = true;
4017 			}
4018 		}
4019 	}
4020 	spin_unlock_bh(&ifp->lock);
4021 
4022 	if (action == DAD_BEGIN) {
4023 		addrconf_dad_begin(ifp);
4024 		goto out;
4025 	} else if (action == DAD_ABORT) {
4026 		in6_ifa_hold(ifp);
4027 		addrconf_dad_stop(ifp, 1);
4028 		if (disable_ipv6)
4029 			addrconf_ifdown(idev->dev, 0);
4030 		goto out;
4031 	}
4032 
4033 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4034 		goto out;
4035 
4036 	write_lock_bh(&idev->lock);
4037 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4038 		write_unlock_bh(&idev->lock);
4039 		goto out;
4040 	}
4041 
4042 	spin_lock(&ifp->lock);
4043 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4044 		spin_unlock(&ifp->lock);
4045 		write_unlock_bh(&idev->lock);
4046 		goto out;
4047 	}
4048 
4049 	if (ifp->dad_probes == 0) {
4050 		bool send_na = false;
4051 
4052 		/*
4053 		 * DAD was successful
4054 		 */
4055 
4056 		if (ifp->flags & IFA_F_TENTATIVE &&
4057 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4058 			send_na = true;
4059 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4060 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4061 		spin_unlock(&ifp->lock);
4062 		write_unlock_bh(&idev->lock);
4063 
4064 		addrconf_dad_completed(ifp, bump_id, send_na);
4065 
4066 		goto out;
4067 	}
4068 
4069 	ifp->dad_probes--;
4070 	addrconf_mod_dad_work(ifp,
4071 			      NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4072 	spin_unlock(&ifp->lock);
4073 	write_unlock_bh(&idev->lock);
4074 
4075 	/* send a neighbour solicitation for our addr */
4076 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4077 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4078 		      ifp->dad_nonce);
4079 out:
4080 	in6_ifa_put(ifp);
4081 	rtnl_unlock();
4082 }
4083 
4084 /* ifp->idev must be at least read locked */
4085 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4086 {
4087 	struct inet6_ifaddr *ifpiter;
4088 	struct inet6_dev *idev = ifp->idev;
4089 
4090 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4091 		if (ifpiter->scope > IFA_LINK)
4092 			break;
4093 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4094 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4095 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4096 		    IFA_F_PERMANENT)
4097 			return false;
4098 	}
4099 	return true;
4100 }
4101 
4102 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4103 				   bool send_na)
4104 {
4105 	struct net_device *dev = ifp->idev->dev;
4106 	struct in6_addr lladdr;
4107 	bool send_rs, send_mld;
4108 
4109 	addrconf_del_dad_work(ifp);
4110 
4111 	/*
4112 	 *	Configure the address for reception. Now it is valid.
4113 	 */
4114 
4115 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4116 
4117 	/* If added prefix is link local and we are prepared to process
4118 	   router advertisements, start sending router solicitations.
4119 	 */
4120 
4121 	read_lock_bh(&ifp->idev->lock);
4122 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4123 	send_rs = send_mld &&
4124 		  ipv6_accept_ra(ifp->idev) &&
4125 		  ifp->idev->cnf.rtr_solicits != 0 &&
4126 		  (dev->flags&IFF_LOOPBACK) == 0;
4127 	read_unlock_bh(&ifp->idev->lock);
4128 
4129 	/* While dad is in progress mld report's source address is in6_addrany.
4130 	 * Resend with proper ll now.
4131 	 */
4132 	if (send_mld)
4133 		ipv6_mc_dad_complete(ifp->idev);
4134 
4135 	/* send unsolicited NA if enabled */
4136 	if (send_na &&
4137 	    (ifp->idev->cnf.ndisc_notify ||
4138 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4139 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4140 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4141 			      /*solicited=*/ false, /*override=*/ true,
4142 			      /*inc_opt=*/ true);
4143 	}
4144 
4145 	if (send_rs) {
4146 		/*
4147 		 *	If a host as already performed a random delay
4148 		 *	[...] as part of DAD [...] there is no need
4149 		 *	to delay again before sending the first RS
4150 		 */
4151 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4152 			return;
4153 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4154 
4155 		write_lock_bh(&ifp->idev->lock);
4156 		spin_lock(&ifp->lock);
4157 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4158 			ifp->idev->cnf.rtr_solicit_interval);
4159 		ifp->idev->rs_probes = 1;
4160 		ifp->idev->if_flags |= IF_RS_SENT;
4161 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4162 		spin_unlock(&ifp->lock);
4163 		write_unlock_bh(&ifp->idev->lock);
4164 	}
4165 
4166 	if (bump_id)
4167 		rt_genid_bump_ipv6(dev_net(dev));
4168 
4169 	/* Make sure that a new temporary address will be created
4170 	 * before this temporary address becomes deprecated.
4171 	 */
4172 	if (ifp->flags & IFA_F_TEMPORARY)
4173 		addrconf_verify_rtnl();
4174 }
4175 
4176 static void addrconf_dad_run(struct inet6_dev *idev)
4177 {
4178 	struct inet6_ifaddr *ifp;
4179 
4180 	read_lock_bh(&idev->lock);
4181 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4182 		spin_lock(&ifp->lock);
4183 		if (ifp->flags & IFA_F_TENTATIVE &&
4184 		    ifp->state == INET6_IFADDR_STATE_DAD)
4185 			addrconf_dad_kick(ifp);
4186 		spin_unlock(&ifp->lock);
4187 	}
4188 	read_unlock_bh(&idev->lock);
4189 }
4190 
4191 #ifdef CONFIG_PROC_FS
4192 struct if6_iter_state {
4193 	struct seq_net_private p;
4194 	int bucket;
4195 	int offset;
4196 };
4197 
4198 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4199 {
4200 	struct if6_iter_state *state = seq->private;
4201 	struct net *net = seq_file_net(seq);
4202 	struct inet6_ifaddr *ifa = NULL;
4203 	int p = 0;
4204 
4205 	/* initial bucket if pos is 0 */
4206 	if (pos == 0) {
4207 		state->bucket = 0;
4208 		state->offset = 0;
4209 	}
4210 
4211 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4212 		hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4213 					 addr_lst) {
4214 			if (!net_eq(dev_net(ifa->idev->dev), net))
4215 				continue;
4216 			/* sync with offset */
4217 			if (p < state->offset) {
4218 				p++;
4219 				continue;
4220 			}
4221 			return ifa;
4222 		}
4223 
4224 		/* prepare for next bucket */
4225 		state->offset = 0;
4226 		p = 0;
4227 	}
4228 	return NULL;
4229 }
4230 
4231 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4232 					 struct inet6_ifaddr *ifa)
4233 {
4234 	struct if6_iter_state *state = seq->private;
4235 	struct net *net = seq_file_net(seq);
4236 
4237 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4238 		if (!net_eq(dev_net(ifa->idev->dev), net))
4239 			continue;
4240 		state->offset++;
4241 		return ifa;
4242 	}
4243 
4244 	state->offset = 0;
4245 	while (++state->bucket < IN6_ADDR_HSIZE) {
4246 		hlist_for_each_entry_rcu(ifa,
4247 				     &inet6_addr_lst[state->bucket], addr_lst) {
4248 			if (!net_eq(dev_net(ifa->idev->dev), net))
4249 				continue;
4250 			return ifa;
4251 		}
4252 	}
4253 
4254 	return NULL;
4255 }
4256 
4257 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4258 	__acquires(rcu)
4259 {
4260 	rcu_read_lock();
4261 	return if6_get_first(seq, *pos);
4262 }
4263 
4264 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4265 {
4266 	struct inet6_ifaddr *ifa;
4267 
4268 	ifa = if6_get_next(seq, v);
4269 	++*pos;
4270 	return ifa;
4271 }
4272 
4273 static void if6_seq_stop(struct seq_file *seq, void *v)
4274 	__releases(rcu)
4275 {
4276 	rcu_read_unlock();
4277 }
4278 
4279 static int if6_seq_show(struct seq_file *seq, void *v)
4280 {
4281 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4282 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4283 		   &ifp->addr,
4284 		   ifp->idev->dev->ifindex,
4285 		   ifp->prefix_len,
4286 		   ifp->scope,
4287 		   (u8) ifp->flags,
4288 		   ifp->idev->dev->name);
4289 	return 0;
4290 }
4291 
4292 static const struct seq_operations if6_seq_ops = {
4293 	.start	= if6_seq_start,
4294 	.next	= if6_seq_next,
4295 	.show	= if6_seq_show,
4296 	.stop	= if6_seq_stop,
4297 };
4298 
4299 static int __net_init if6_proc_net_init(struct net *net)
4300 {
4301 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4302 			sizeof(struct if6_iter_state)))
4303 		return -ENOMEM;
4304 	return 0;
4305 }
4306 
4307 static void __net_exit if6_proc_net_exit(struct net *net)
4308 {
4309 	remove_proc_entry("if_inet6", net->proc_net);
4310 }
4311 
4312 static struct pernet_operations if6_proc_net_ops = {
4313 	.init = if6_proc_net_init,
4314 	.exit = if6_proc_net_exit,
4315 };
4316 
4317 int __init if6_proc_init(void)
4318 {
4319 	return register_pernet_subsys(&if6_proc_net_ops);
4320 }
4321 
4322 void if6_proc_exit(void)
4323 {
4324 	unregister_pernet_subsys(&if6_proc_net_ops);
4325 }
4326 #endif	/* CONFIG_PROC_FS */
4327 
4328 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4329 /* Check if address is a home address configured on any interface. */
4330 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4331 {
4332 	unsigned int hash = inet6_addr_hash(net, addr);
4333 	struct inet6_ifaddr *ifp = NULL;
4334 	int ret = 0;
4335 
4336 	rcu_read_lock();
4337 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4338 		if (!net_eq(dev_net(ifp->idev->dev), net))
4339 			continue;
4340 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4341 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4342 			ret = 1;
4343 			break;
4344 		}
4345 	}
4346 	rcu_read_unlock();
4347 	return ret;
4348 }
4349 #endif
4350 
4351 /*
4352  *	Periodic address status verification
4353  */
4354 
4355 static void addrconf_verify_rtnl(void)
4356 {
4357 	unsigned long now, next, next_sec, next_sched;
4358 	struct inet6_ifaddr *ifp;
4359 	int i;
4360 
4361 	ASSERT_RTNL();
4362 
4363 	rcu_read_lock_bh();
4364 	now = jiffies;
4365 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4366 
4367 	cancel_delayed_work(&addr_chk_work);
4368 
4369 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4370 restart:
4371 		hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4372 			unsigned long age;
4373 
4374 			/* When setting preferred_lft to a value not zero or
4375 			 * infinity, while valid_lft is infinity
4376 			 * IFA_F_PERMANENT has a non-infinity life time.
4377 			 */
4378 			if ((ifp->flags & IFA_F_PERMANENT) &&
4379 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4380 				continue;
4381 
4382 			spin_lock(&ifp->lock);
4383 			/* We try to batch several events at once. */
4384 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4385 
4386 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4387 			    age >= ifp->valid_lft) {
4388 				spin_unlock(&ifp->lock);
4389 				in6_ifa_hold(ifp);
4390 				ipv6_del_addr(ifp);
4391 				goto restart;
4392 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4393 				spin_unlock(&ifp->lock);
4394 				continue;
4395 			} else if (age >= ifp->prefered_lft) {
4396 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4397 				int deprecate = 0;
4398 
4399 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4400 					deprecate = 1;
4401 					ifp->flags |= IFA_F_DEPRECATED;
4402 				}
4403 
4404 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4405 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4406 					next = ifp->tstamp + ifp->valid_lft * HZ;
4407 
4408 				spin_unlock(&ifp->lock);
4409 
4410 				if (deprecate) {
4411 					in6_ifa_hold(ifp);
4412 
4413 					ipv6_ifa_notify(0, ifp);
4414 					in6_ifa_put(ifp);
4415 					goto restart;
4416 				}
4417 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4418 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4419 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4420 					ifp->idev->cnf.dad_transmits *
4421 					NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4422 
4423 				if (age >= ifp->prefered_lft - regen_advance) {
4424 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4425 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4426 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4427 					if (!ifp->regen_count && ifpub) {
4428 						ifp->regen_count++;
4429 						in6_ifa_hold(ifp);
4430 						in6_ifa_hold(ifpub);
4431 						spin_unlock(&ifp->lock);
4432 
4433 						spin_lock(&ifpub->lock);
4434 						ifpub->regen_count = 0;
4435 						spin_unlock(&ifpub->lock);
4436 						rcu_read_unlock_bh();
4437 						ipv6_create_tempaddr(ifpub, ifp, true);
4438 						in6_ifa_put(ifpub);
4439 						in6_ifa_put(ifp);
4440 						rcu_read_lock_bh();
4441 						goto restart;
4442 					}
4443 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4444 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4445 				spin_unlock(&ifp->lock);
4446 			} else {
4447 				/* ifp->prefered_lft <= ifp->valid_lft */
4448 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4449 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4450 				spin_unlock(&ifp->lock);
4451 			}
4452 		}
4453 	}
4454 
4455 	next_sec = round_jiffies_up(next);
4456 	next_sched = next;
4457 
4458 	/* If rounded timeout is accurate enough, accept it. */
4459 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4460 		next_sched = next_sec;
4461 
4462 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4463 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4464 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4465 
4466 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4467 		 now, next, next_sec, next_sched);
4468 	mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4469 	rcu_read_unlock_bh();
4470 }
4471 
4472 static void addrconf_verify_work(struct work_struct *w)
4473 {
4474 	rtnl_lock();
4475 	addrconf_verify_rtnl();
4476 	rtnl_unlock();
4477 }
4478 
4479 static void addrconf_verify(void)
4480 {
4481 	mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4482 }
4483 
4484 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4485 				     struct in6_addr **peer_pfx)
4486 {
4487 	struct in6_addr *pfx = NULL;
4488 
4489 	*peer_pfx = NULL;
4490 
4491 	if (addr)
4492 		pfx = nla_data(addr);
4493 
4494 	if (local) {
4495 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4496 			*peer_pfx = pfx;
4497 		pfx = nla_data(local);
4498 	}
4499 
4500 	return pfx;
4501 }
4502 
4503 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4504 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4505 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4506 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4507 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4508 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4509 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4510 };
4511 
4512 static int
4513 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4514 		  struct netlink_ext_ack *extack)
4515 {
4516 	struct net *net = sock_net(skb->sk);
4517 	struct ifaddrmsg *ifm;
4518 	struct nlattr *tb[IFA_MAX+1];
4519 	struct in6_addr *pfx, *peer_pfx;
4520 	u32 ifa_flags;
4521 	int err;
4522 
4523 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4524 			  extack);
4525 	if (err < 0)
4526 		return err;
4527 
4528 	ifm = nlmsg_data(nlh);
4529 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4530 	if (!pfx)
4531 		return -EINVAL;
4532 
4533 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4534 
4535 	/* We ignore other flags so far. */
4536 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4537 
4538 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4539 			      ifm->ifa_prefixlen);
4540 }
4541 
4542 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4543 			       unsigned long expires, u32 flags)
4544 {
4545 	struct fib6_info *f6i;
4546 	u32 prio;
4547 
4548 	f6i = addrconf_get_prefix_route(&ifp->addr,
4549 					ifp->prefix_len,
4550 					ifp->idev->dev,
4551 					0, RTF_GATEWAY | RTF_DEFAULT);
4552 	if (!f6i)
4553 		return -ENOENT;
4554 
4555 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4556 	if (f6i->fib6_metric != prio) {
4557 		/* delete old one */
4558 		ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4559 
4560 		/* add new one */
4561 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4562 				      ifp->rt_priority, ifp->idev->dev,
4563 				      expires, flags, GFP_KERNEL);
4564 	} else {
4565 		if (!expires)
4566 			fib6_clean_expires(f6i);
4567 		else
4568 			fib6_set_expires(f6i, expires);
4569 
4570 		fib6_info_release(f6i);
4571 	}
4572 
4573 	return 0;
4574 }
4575 
4576 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4577 {
4578 	u32 flags;
4579 	clock_t expires;
4580 	unsigned long timeout;
4581 	bool was_managetempaddr;
4582 	bool had_prefixroute;
4583 
4584 	ASSERT_RTNL();
4585 
4586 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4587 		return -EINVAL;
4588 
4589 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4590 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4591 		return -EINVAL;
4592 
4593 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4594 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4595 
4596 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4597 	if (addrconf_finite_timeout(timeout)) {
4598 		expires = jiffies_to_clock_t(timeout * HZ);
4599 		cfg->valid_lft = timeout;
4600 		flags = RTF_EXPIRES;
4601 	} else {
4602 		expires = 0;
4603 		flags = 0;
4604 		cfg->ifa_flags |= IFA_F_PERMANENT;
4605 	}
4606 
4607 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4608 	if (addrconf_finite_timeout(timeout)) {
4609 		if (timeout == 0)
4610 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4611 		cfg->preferred_lft = timeout;
4612 	}
4613 
4614 	spin_lock_bh(&ifp->lock);
4615 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4616 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4617 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4618 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4619 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4620 			IFA_F_NOPREFIXROUTE);
4621 	ifp->flags |= cfg->ifa_flags;
4622 	ifp->tstamp = jiffies;
4623 	ifp->valid_lft = cfg->valid_lft;
4624 	ifp->prefered_lft = cfg->preferred_lft;
4625 
4626 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4627 		ifp->rt_priority = cfg->rt_priority;
4628 
4629 	spin_unlock_bh(&ifp->lock);
4630 	if (!(ifp->flags&IFA_F_TENTATIVE))
4631 		ipv6_ifa_notify(0, ifp);
4632 
4633 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4634 		int rc = -ENOENT;
4635 
4636 		if (had_prefixroute)
4637 			rc = modify_prefix_route(ifp, expires, flags);
4638 
4639 		/* prefix route could have been deleted; if so restore it */
4640 		if (rc == -ENOENT) {
4641 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4642 					      ifp->rt_priority, ifp->idev->dev,
4643 					      expires, flags, GFP_KERNEL);
4644 		}
4645 	} else if (had_prefixroute) {
4646 		enum cleanup_prefix_rt_t action;
4647 		unsigned long rt_expires;
4648 
4649 		write_lock_bh(&ifp->idev->lock);
4650 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4651 		write_unlock_bh(&ifp->idev->lock);
4652 
4653 		if (action != CLEANUP_PREFIX_RT_NOP) {
4654 			cleanup_prefix_route(ifp, rt_expires,
4655 				action == CLEANUP_PREFIX_RT_DEL);
4656 		}
4657 	}
4658 
4659 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4660 		if (was_managetempaddr &&
4661 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4662 			cfg->valid_lft = 0;
4663 			cfg->preferred_lft = 0;
4664 		}
4665 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4666 				 cfg->preferred_lft, !was_managetempaddr,
4667 				 jiffies);
4668 	}
4669 
4670 	addrconf_verify_rtnl();
4671 
4672 	return 0;
4673 }
4674 
4675 static int
4676 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4677 		  struct netlink_ext_ack *extack)
4678 {
4679 	struct net *net = sock_net(skb->sk);
4680 	struct ifaddrmsg *ifm;
4681 	struct nlattr *tb[IFA_MAX+1];
4682 	struct in6_addr *peer_pfx;
4683 	struct inet6_ifaddr *ifa;
4684 	struct net_device *dev;
4685 	struct inet6_dev *idev;
4686 	struct ifa6_config cfg;
4687 	int err;
4688 
4689 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4690 			  extack);
4691 	if (err < 0)
4692 		return err;
4693 
4694 	memset(&cfg, 0, sizeof(cfg));
4695 
4696 	ifm = nlmsg_data(nlh);
4697 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4698 	if (!cfg.pfx)
4699 		return -EINVAL;
4700 
4701 	cfg.peer_pfx = peer_pfx;
4702 	cfg.plen = ifm->ifa_prefixlen;
4703 	if (tb[IFA_RT_PRIORITY])
4704 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4705 
4706 	cfg.valid_lft = INFINITY_LIFE_TIME;
4707 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4708 
4709 	if (tb[IFA_CACHEINFO]) {
4710 		struct ifa_cacheinfo *ci;
4711 
4712 		ci = nla_data(tb[IFA_CACHEINFO]);
4713 		cfg.valid_lft = ci->ifa_valid;
4714 		cfg.preferred_lft = ci->ifa_prefered;
4715 	}
4716 
4717 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4718 	if (!dev)
4719 		return -ENODEV;
4720 
4721 	if (tb[IFA_FLAGS])
4722 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4723 	else
4724 		cfg.ifa_flags = ifm->ifa_flags;
4725 
4726 	/* We ignore other flags so far. */
4727 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4728 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4729 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4730 
4731 	idev = ipv6_find_idev(dev);
4732 	if (IS_ERR(idev))
4733 		return PTR_ERR(idev);
4734 
4735 	if (!ipv6_allow_optimistic_dad(net, idev))
4736 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4737 
4738 	if (cfg.ifa_flags & IFA_F_NODAD &&
4739 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4740 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4741 		return -EINVAL;
4742 	}
4743 
4744 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4745 	if (!ifa) {
4746 		/*
4747 		 * It would be best to check for !NLM_F_CREATE here but
4748 		 * userspace already relies on not having to provide this.
4749 		 */
4750 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4751 	}
4752 
4753 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4754 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4755 		err = -EEXIST;
4756 	else
4757 		err = inet6_addr_modify(ifa, &cfg);
4758 
4759 	in6_ifa_put(ifa);
4760 
4761 	return err;
4762 }
4763 
4764 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4765 			  u8 scope, int ifindex)
4766 {
4767 	struct ifaddrmsg *ifm;
4768 
4769 	ifm = nlmsg_data(nlh);
4770 	ifm->ifa_family = AF_INET6;
4771 	ifm->ifa_prefixlen = prefixlen;
4772 	ifm->ifa_flags = flags;
4773 	ifm->ifa_scope = scope;
4774 	ifm->ifa_index = ifindex;
4775 }
4776 
4777 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4778 			 unsigned long tstamp, u32 preferred, u32 valid)
4779 {
4780 	struct ifa_cacheinfo ci;
4781 
4782 	ci.cstamp = cstamp_delta(cstamp);
4783 	ci.tstamp = cstamp_delta(tstamp);
4784 	ci.ifa_prefered = preferred;
4785 	ci.ifa_valid = valid;
4786 
4787 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4788 }
4789 
4790 static inline int rt_scope(int ifa_scope)
4791 {
4792 	if (ifa_scope & IFA_HOST)
4793 		return RT_SCOPE_HOST;
4794 	else if (ifa_scope & IFA_LINK)
4795 		return RT_SCOPE_LINK;
4796 	else if (ifa_scope & IFA_SITE)
4797 		return RT_SCOPE_SITE;
4798 	else
4799 		return RT_SCOPE_UNIVERSE;
4800 }
4801 
4802 static inline int inet6_ifaddr_msgsize(void)
4803 {
4804 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4805 	       + nla_total_size(16) /* IFA_LOCAL */
4806 	       + nla_total_size(16) /* IFA_ADDRESS */
4807 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4808 	       + nla_total_size(4)  /* IFA_FLAGS */
4809 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4810 }
4811 
4812 enum addr_type_t {
4813 	UNICAST_ADDR,
4814 	MULTICAST_ADDR,
4815 	ANYCAST_ADDR,
4816 };
4817 
4818 struct inet6_fill_args {
4819 	u32 portid;
4820 	u32 seq;
4821 	int event;
4822 	unsigned int flags;
4823 	int netnsid;
4824 	int ifindex;
4825 	enum addr_type_t type;
4826 };
4827 
4828 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4829 			     struct inet6_fill_args *args)
4830 {
4831 	struct nlmsghdr  *nlh;
4832 	u32 preferred, valid;
4833 
4834 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4835 			sizeof(struct ifaddrmsg), args->flags);
4836 	if (!nlh)
4837 		return -EMSGSIZE;
4838 
4839 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4840 		      ifa->idev->dev->ifindex);
4841 
4842 	if (args->netnsid >= 0 &&
4843 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4844 		goto error;
4845 
4846 	if (!((ifa->flags&IFA_F_PERMANENT) &&
4847 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4848 		preferred = ifa->prefered_lft;
4849 		valid = ifa->valid_lft;
4850 		if (preferred != INFINITY_LIFE_TIME) {
4851 			long tval = (jiffies - ifa->tstamp)/HZ;
4852 			if (preferred > tval)
4853 				preferred -= tval;
4854 			else
4855 				preferred = 0;
4856 			if (valid != INFINITY_LIFE_TIME) {
4857 				if (valid > tval)
4858 					valid -= tval;
4859 				else
4860 					valid = 0;
4861 			}
4862 		}
4863 	} else {
4864 		preferred = INFINITY_LIFE_TIME;
4865 		valid = INFINITY_LIFE_TIME;
4866 	}
4867 
4868 	if (!ipv6_addr_any(&ifa->peer_addr)) {
4869 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4870 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4871 			goto error;
4872 	} else
4873 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4874 			goto error;
4875 
4876 	if (ifa->rt_priority &&
4877 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4878 		goto error;
4879 
4880 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4881 		goto error;
4882 
4883 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4884 		goto error;
4885 
4886 	nlmsg_end(skb, nlh);
4887 	return 0;
4888 
4889 error:
4890 	nlmsg_cancel(skb, nlh);
4891 	return -EMSGSIZE;
4892 }
4893 
4894 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4895 			       struct inet6_fill_args *args)
4896 {
4897 	struct nlmsghdr  *nlh;
4898 	u8 scope = RT_SCOPE_UNIVERSE;
4899 	int ifindex = ifmca->idev->dev->ifindex;
4900 
4901 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4902 		scope = RT_SCOPE_SITE;
4903 
4904 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4905 			sizeof(struct ifaddrmsg), args->flags);
4906 	if (!nlh)
4907 		return -EMSGSIZE;
4908 
4909 	if (args->netnsid >= 0 &&
4910 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4911 		return -EMSGSIZE;
4912 
4913 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4914 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4915 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4916 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4917 		nlmsg_cancel(skb, nlh);
4918 		return -EMSGSIZE;
4919 	}
4920 
4921 	nlmsg_end(skb, nlh);
4922 	return 0;
4923 }
4924 
4925 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4926 			       struct inet6_fill_args *args)
4927 {
4928 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4929 	int ifindex = dev ? dev->ifindex : 1;
4930 	struct nlmsghdr  *nlh;
4931 	u8 scope = RT_SCOPE_UNIVERSE;
4932 
4933 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4934 		scope = RT_SCOPE_SITE;
4935 
4936 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4937 			sizeof(struct ifaddrmsg), args->flags);
4938 	if (!nlh)
4939 		return -EMSGSIZE;
4940 
4941 	if (args->netnsid >= 0 &&
4942 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4943 		return -EMSGSIZE;
4944 
4945 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4946 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4947 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4948 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4949 		nlmsg_cancel(skb, nlh);
4950 		return -EMSGSIZE;
4951 	}
4952 
4953 	nlmsg_end(skb, nlh);
4954 	return 0;
4955 }
4956 
4957 /* called with rcu_read_lock() */
4958 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4959 			  struct netlink_callback *cb, int s_ip_idx,
4960 			  struct inet6_fill_args *fillargs)
4961 {
4962 	struct ifmcaddr6 *ifmca;
4963 	struct ifacaddr6 *ifaca;
4964 	int ip_idx = 0;
4965 	int err = 1;
4966 
4967 	read_lock_bh(&idev->lock);
4968 	switch (fillargs->type) {
4969 	case UNICAST_ADDR: {
4970 		struct inet6_ifaddr *ifa;
4971 		fillargs->event = RTM_NEWADDR;
4972 
4973 		/* unicast address incl. temp addr */
4974 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
4975 			if (ip_idx < s_ip_idx)
4976 				goto next;
4977 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
4978 			if (err < 0)
4979 				break;
4980 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4981 next:
4982 			ip_idx++;
4983 		}
4984 		break;
4985 	}
4986 	case MULTICAST_ADDR:
4987 		fillargs->event = RTM_GETMULTICAST;
4988 
4989 		/* multicast address */
4990 		for (ifmca = idev->mc_list; ifmca;
4991 		     ifmca = ifmca->next, ip_idx++) {
4992 			if (ip_idx < s_ip_idx)
4993 				continue;
4994 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
4995 			if (err < 0)
4996 				break;
4997 		}
4998 		break;
4999 	case ANYCAST_ADDR:
5000 		fillargs->event = RTM_GETANYCAST;
5001 		/* anycast address */
5002 		for (ifaca = idev->ac_list; ifaca;
5003 		     ifaca = ifaca->aca_next, ip_idx++) {
5004 			if (ip_idx < s_ip_idx)
5005 				continue;
5006 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5007 			if (err < 0)
5008 				break;
5009 		}
5010 		break;
5011 	default:
5012 		break;
5013 	}
5014 	read_unlock_bh(&idev->lock);
5015 	cb->args[2] = ip_idx;
5016 	return err;
5017 }
5018 
5019 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5020 				       struct inet6_fill_args *fillargs,
5021 				       struct net **tgt_net, struct sock *sk,
5022 				       struct netlink_callback *cb)
5023 {
5024 	struct netlink_ext_ack *extack = cb->extack;
5025 	struct nlattr *tb[IFA_MAX+1];
5026 	struct ifaddrmsg *ifm;
5027 	int err, i;
5028 
5029 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5030 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5031 		return -EINVAL;
5032 	}
5033 
5034 	ifm = nlmsg_data(nlh);
5035 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5036 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5037 		return -EINVAL;
5038 	}
5039 
5040 	fillargs->ifindex = ifm->ifa_index;
5041 	if (fillargs->ifindex) {
5042 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5043 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5044 	}
5045 
5046 	err = nlmsg_parse_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5047 				 ifa_ipv6_policy, extack);
5048 	if (err < 0)
5049 		return err;
5050 
5051 	for (i = 0; i <= IFA_MAX; ++i) {
5052 		if (!tb[i])
5053 			continue;
5054 
5055 		if (i == IFA_TARGET_NETNSID) {
5056 			struct net *net;
5057 
5058 			fillargs->netnsid = nla_get_s32(tb[i]);
5059 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5060 			if (IS_ERR(net)) {
5061 				fillargs->netnsid = -1;
5062 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5063 				return PTR_ERR(net);
5064 			}
5065 			*tgt_net = net;
5066 		} else {
5067 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5068 			return -EINVAL;
5069 		}
5070 	}
5071 
5072 	return 0;
5073 }
5074 
5075 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5076 			   enum addr_type_t type)
5077 {
5078 	const struct nlmsghdr *nlh = cb->nlh;
5079 	struct inet6_fill_args fillargs = {
5080 		.portid = NETLINK_CB(cb->skb).portid,
5081 		.seq = cb->nlh->nlmsg_seq,
5082 		.flags = NLM_F_MULTI,
5083 		.netnsid = -1,
5084 		.type = type,
5085 	};
5086 	struct net *net = sock_net(skb->sk);
5087 	struct net *tgt_net = net;
5088 	int idx, s_idx, s_ip_idx;
5089 	int h, s_h;
5090 	struct net_device *dev;
5091 	struct inet6_dev *idev;
5092 	struct hlist_head *head;
5093 	int err = 0;
5094 
5095 	s_h = cb->args[0];
5096 	s_idx = idx = cb->args[1];
5097 	s_ip_idx = cb->args[2];
5098 
5099 	if (cb->strict_check) {
5100 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5101 						  skb->sk, cb);
5102 		if (err < 0)
5103 			goto put_tgt_net;
5104 
5105 		err = 0;
5106 		if (fillargs.ifindex) {
5107 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5108 			if (!dev) {
5109 				err = -ENODEV;
5110 				goto put_tgt_net;
5111 			}
5112 			idev = __in6_dev_get(dev);
5113 			if (idev) {
5114 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5115 						     &fillargs);
5116 			}
5117 			goto put_tgt_net;
5118 		}
5119 	}
5120 
5121 	rcu_read_lock();
5122 	cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5123 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5124 		idx = 0;
5125 		head = &tgt_net->dev_index_head[h];
5126 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5127 			if (idx < s_idx)
5128 				goto cont;
5129 			if (h > s_h || idx > s_idx)
5130 				s_ip_idx = 0;
5131 			idev = __in6_dev_get(dev);
5132 			if (!idev)
5133 				goto cont;
5134 
5135 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5136 					   &fillargs) < 0)
5137 				goto done;
5138 cont:
5139 			idx++;
5140 		}
5141 	}
5142 done:
5143 	rcu_read_unlock();
5144 	cb->args[0] = h;
5145 	cb->args[1] = idx;
5146 put_tgt_net:
5147 	if (fillargs.netnsid >= 0)
5148 		put_net(tgt_net);
5149 
5150 	return err < 0 ? err : skb->len;
5151 }
5152 
5153 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5154 {
5155 	enum addr_type_t type = UNICAST_ADDR;
5156 
5157 	return inet6_dump_addr(skb, cb, type);
5158 }
5159 
5160 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5161 {
5162 	enum addr_type_t type = MULTICAST_ADDR;
5163 
5164 	return inet6_dump_addr(skb, cb, type);
5165 }
5166 
5167 
5168 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5169 {
5170 	enum addr_type_t type = ANYCAST_ADDR;
5171 
5172 	return inet6_dump_addr(skb, cb, type);
5173 }
5174 
5175 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5176 			     struct netlink_ext_ack *extack)
5177 {
5178 	struct net *net = sock_net(in_skb->sk);
5179 	struct inet6_fill_args fillargs = {
5180 		.portid = NETLINK_CB(in_skb).portid,
5181 		.seq = nlh->nlmsg_seq,
5182 		.event = RTM_NEWADDR,
5183 		.flags = 0,
5184 		.netnsid = -1,
5185 	};
5186 	struct net *tgt_net = net;
5187 	struct ifaddrmsg *ifm;
5188 	struct nlattr *tb[IFA_MAX+1];
5189 	struct in6_addr *addr = NULL, *peer;
5190 	struct net_device *dev = NULL;
5191 	struct inet6_ifaddr *ifa;
5192 	struct sk_buff *skb;
5193 	int err;
5194 
5195 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5196 			  extack);
5197 	if (err < 0)
5198 		return err;
5199 
5200 	if (tb[IFA_TARGET_NETNSID]) {
5201 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5202 
5203 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5204 						  fillargs.netnsid);
5205 		if (IS_ERR(tgt_net))
5206 			return PTR_ERR(tgt_net);
5207 	}
5208 
5209 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5210 	if (!addr)
5211 		return -EINVAL;
5212 
5213 	ifm = nlmsg_data(nlh);
5214 	if (ifm->ifa_index)
5215 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5216 
5217 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5218 	if (!ifa) {
5219 		err = -EADDRNOTAVAIL;
5220 		goto errout;
5221 	}
5222 
5223 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5224 	if (!skb) {
5225 		err = -ENOBUFS;
5226 		goto errout_ifa;
5227 	}
5228 
5229 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5230 	if (err < 0) {
5231 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5232 		WARN_ON(err == -EMSGSIZE);
5233 		kfree_skb(skb);
5234 		goto errout_ifa;
5235 	}
5236 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5237 errout_ifa:
5238 	in6_ifa_put(ifa);
5239 errout:
5240 	if (dev)
5241 		dev_put(dev);
5242 	if (fillargs.netnsid >= 0)
5243 		put_net(tgt_net);
5244 
5245 	return err;
5246 }
5247 
5248 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5249 {
5250 	struct sk_buff *skb;
5251 	struct net *net = dev_net(ifa->idev->dev);
5252 	struct inet6_fill_args fillargs = {
5253 		.portid = 0,
5254 		.seq = 0,
5255 		.event = event,
5256 		.flags = 0,
5257 		.netnsid = -1,
5258 	};
5259 	int err = -ENOBUFS;
5260 
5261 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5262 	if (!skb)
5263 		goto errout;
5264 
5265 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5266 	if (err < 0) {
5267 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5268 		WARN_ON(err == -EMSGSIZE);
5269 		kfree_skb(skb);
5270 		goto errout;
5271 	}
5272 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5273 	return;
5274 errout:
5275 	if (err < 0)
5276 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5277 }
5278 
5279 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5280 				__s32 *array, int bytes)
5281 {
5282 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5283 
5284 	memset(array, 0, bytes);
5285 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5286 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5287 	array[DEVCONF_MTU6] = cnf->mtu6;
5288 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5289 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5290 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5291 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5292 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5293 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5294 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5295 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5296 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5297 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5298 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5299 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5300 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5301 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5302 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5303 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5304 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5305 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5306 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5307 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5308 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5309 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5310 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5311 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5312 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5313 #ifdef CONFIG_IPV6_ROUTER_PREF
5314 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5315 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5316 		jiffies_to_msecs(cnf->rtr_probe_interval);
5317 #ifdef CONFIG_IPV6_ROUTE_INFO
5318 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5319 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5320 #endif
5321 #endif
5322 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5323 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5324 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5325 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5326 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5327 #endif
5328 #ifdef CONFIG_IPV6_MROUTE
5329 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5330 #endif
5331 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5332 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5333 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5334 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5335 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5336 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5337 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5338 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5339 	/* we omit DEVCONF_STABLE_SECRET for now */
5340 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5341 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5342 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5343 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5344 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5345 #ifdef CONFIG_IPV6_SEG6_HMAC
5346 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5347 #endif
5348 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5349 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5350 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5351 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5352 }
5353 
5354 static inline size_t inet6_ifla6_size(void)
5355 {
5356 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5357 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5358 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5359 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5360 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5361 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5362 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5363 	     + 0;
5364 }
5365 
5366 static inline size_t inet6_if_nlmsg_size(void)
5367 {
5368 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5369 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5370 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5371 	       + nla_total_size(4) /* IFLA_MTU */
5372 	       + nla_total_size(4) /* IFLA_LINK */
5373 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5374 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5375 }
5376 
5377 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5378 					int bytes)
5379 {
5380 	int i;
5381 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5382 	BUG_ON(pad < 0);
5383 
5384 	/* Use put_unaligned() because stats may not be aligned for u64. */
5385 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5386 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5387 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5388 
5389 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5390 }
5391 
5392 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5393 					int bytes, size_t syncpoff)
5394 {
5395 	int i, c;
5396 	u64 buff[IPSTATS_MIB_MAX];
5397 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5398 
5399 	BUG_ON(pad < 0);
5400 
5401 	memset(buff, 0, sizeof(buff));
5402 	buff[0] = IPSTATS_MIB_MAX;
5403 
5404 	for_each_possible_cpu(c) {
5405 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5406 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5407 	}
5408 
5409 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5410 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5411 }
5412 
5413 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5414 			     int bytes)
5415 {
5416 	switch (attrtype) {
5417 	case IFLA_INET6_STATS:
5418 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5419 				     offsetof(struct ipstats_mib, syncp));
5420 		break;
5421 	case IFLA_INET6_ICMP6STATS:
5422 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5423 		break;
5424 	}
5425 }
5426 
5427 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5428 				  u32 ext_filter_mask)
5429 {
5430 	struct nlattr *nla;
5431 	struct ifla_cacheinfo ci;
5432 
5433 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5434 		goto nla_put_failure;
5435 	ci.max_reasm_len = IPV6_MAXPLEN;
5436 	ci.tstamp = cstamp_delta(idev->tstamp);
5437 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5438 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5439 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5440 		goto nla_put_failure;
5441 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5442 	if (!nla)
5443 		goto nla_put_failure;
5444 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5445 
5446 	/* XXX - MC not implemented */
5447 
5448 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5449 		return 0;
5450 
5451 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5452 	if (!nla)
5453 		goto nla_put_failure;
5454 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5455 
5456 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5457 	if (!nla)
5458 		goto nla_put_failure;
5459 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5460 
5461 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5462 	if (!nla)
5463 		goto nla_put_failure;
5464 
5465 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5466 		goto nla_put_failure;
5467 
5468 	read_lock_bh(&idev->lock);
5469 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5470 	read_unlock_bh(&idev->lock);
5471 
5472 	return 0;
5473 
5474 nla_put_failure:
5475 	return -EMSGSIZE;
5476 }
5477 
5478 static size_t inet6_get_link_af_size(const struct net_device *dev,
5479 				     u32 ext_filter_mask)
5480 {
5481 	if (!__in6_dev_get(dev))
5482 		return 0;
5483 
5484 	return inet6_ifla6_size();
5485 }
5486 
5487 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5488 			      u32 ext_filter_mask)
5489 {
5490 	struct inet6_dev *idev = __in6_dev_get(dev);
5491 
5492 	if (!idev)
5493 		return -ENODATA;
5494 
5495 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5496 		return -EMSGSIZE;
5497 
5498 	return 0;
5499 }
5500 
5501 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5502 {
5503 	struct inet6_ifaddr *ifp;
5504 	struct net_device *dev = idev->dev;
5505 	bool clear_token, update_rs = false;
5506 	struct in6_addr ll_addr;
5507 
5508 	ASSERT_RTNL();
5509 
5510 	if (!token)
5511 		return -EINVAL;
5512 	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5513 		return -EINVAL;
5514 	if (!ipv6_accept_ra(idev))
5515 		return -EINVAL;
5516 	if (idev->cnf.rtr_solicits == 0)
5517 		return -EINVAL;
5518 
5519 	write_lock_bh(&idev->lock);
5520 
5521 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5522 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5523 
5524 	write_unlock_bh(&idev->lock);
5525 
5526 	clear_token = ipv6_addr_any(token);
5527 	if (clear_token)
5528 		goto update_lft;
5529 
5530 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5531 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5532 			     IFA_F_OPTIMISTIC)) {
5533 		/* If we're not ready, then normal ifup will take care
5534 		 * of this. Otherwise, we need to request our rs here.
5535 		 */
5536 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5537 		update_rs = true;
5538 	}
5539 
5540 update_lft:
5541 	write_lock_bh(&idev->lock);
5542 
5543 	if (update_rs) {
5544 		idev->if_flags |= IF_RS_SENT;
5545 		idev->rs_interval = rfc3315_s14_backoff_init(
5546 			idev->cnf.rtr_solicit_interval);
5547 		idev->rs_probes = 1;
5548 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5549 	}
5550 
5551 	/* Well, that's kinda nasty ... */
5552 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5553 		spin_lock(&ifp->lock);
5554 		if (ifp->tokenized) {
5555 			ifp->valid_lft = 0;
5556 			ifp->prefered_lft = 0;
5557 		}
5558 		spin_unlock(&ifp->lock);
5559 	}
5560 
5561 	write_unlock_bh(&idev->lock);
5562 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5563 	addrconf_verify_rtnl();
5564 	return 0;
5565 }
5566 
5567 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5568 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5569 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5570 };
5571 
5572 static int inet6_validate_link_af(const struct net_device *dev,
5573 				  const struct nlattr *nla)
5574 {
5575 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5576 
5577 	if (dev && !__in6_dev_get(dev))
5578 		return -EAFNOSUPPORT;
5579 
5580 	return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5581 				NULL);
5582 }
5583 
5584 static int check_addr_gen_mode(int mode)
5585 {
5586 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5587 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5588 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5589 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5590 		return -EINVAL;
5591 	return 1;
5592 }
5593 
5594 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5595 				int mode)
5596 {
5597 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5598 	    !idev->cnf.stable_secret.initialized &&
5599 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5600 		return -EINVAL;
5601 	return 1;
5602 }
5603 
5604 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5605 {
5606 	int err = -EINVAL;
5607 	struct inet6_dev *idev = __in6_dev_get(dev);
5608 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5609 
5610 	if (!idev)
5611 		return -EAFNOSUPPORT;
5612 
5613 	if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5614 		BUG();
5615 
5616 	if (tb[IFLA_INET6_TOKEN]) {
5617 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5618 		if (err)
5619 			return err;
5620 	}
5621 
5622 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5623 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5624 
5625 		if (check_addr_gen_mode(mode) < 0 ||
5626 		    check_stable_privacy(idev, dev_net(dev), mode) < 0)
5627 			return -EINVAL;
5628 
5629 		idev->cnf.addr_gen_mode = mode;
5630 		err = 0;
5631 	}
5632 
5633 	return err;
5634 }
5635 
5636 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5637 			     u32 portid, u32 seq, int event, unsigned int flags)
5638 {
5639 	struct net_device *dev = idev->dev;
5640 	struct ifinfomsg *hdr;
5641 	struct nlmsghdr *nlh;
5642 	void *protoinfo;
5643 
5644 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5645 	if (!nlh)
5646 		return -EMSGSIZE;
5647 
5648 	hdr = nlmsg_data(nlh);
5649 	hdr->ifi_family = AF_INET6;
5650 	hdr->__ifi_pad = 0;
5651 	hdr->ifi_type = dev->type;
5652 	hdr->ifi_index = dev->ifindex;
5653 	hdr->ifi_flags = dev_get_flags(dev);
5654 	hdr->ifi_change = 0;
5655 
5656 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5657 	    (dev->addr_len &&
5658 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5659 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5660 	    (dev->ifindex != dev_get_iflink(dev) &&
5661 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5662 	    nla_put_u8(skb, IFLA_OPERSTATE,
5663 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5664 		goto nla_put_failure;
5665 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5666 	if (!protoinfo)
5667 		goto nla_put_failure;
5668 
5669 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5670 		goto nla_put_failure;
5671 
5672 	nla_nest_end(skb, protoinfo);
5673 	nlmsg_end(skb, nlh);
5674 	return 0;
5675 
5676 nla_put_failure:
5677 	nlmsg_cancel(skb, nlh);
5678 	return -EMSGSIZE;
5679 }
5680 
5681 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5682 				   struct netlink_ext_ack *extack)
5683 {
5684 	struct ifinfomsg *ifm;
5685 
5686 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5687 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5688 		return -EINVAL;
5689 	}
5690 
5691 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5692 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5693 		return -EINVAL;
5694 	}
5695 
5696 	ifm = nlmsg_data(nlh);
5697 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5698 	    ifm->ifi_change || ifm->ifi_index) {
5699 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5700 		return -EINVAL;
5701 	}
5702 
5703 	return 0;
5704 }
5705 
5706 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5707 {
5708 	struct net *net = sock_net(skb->sk);
5709 	int h, s_h;
5710 	int idx = 0, s_idx;
5711 	struct net_device *dev;
5712 	struct inet6_dev *idev;
5713 	struct hlist_head *head;
5714 
5715 	/* only requests using strict checking can pass data to
5716 	 * influence the dump
5717 	 */
5718 	if (cb->strict_check) {
5719 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5720 
5721 		if (err < 0)
5722 			return err;
5723 	}
5724 
5725 	s_h = cb->args[0];
5726 	s_idx = cb->args[1];
5727 
5728 	rcu_read_lock();
5729 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5730 		idx = 0;
5731 		head = &net->dev_index_head[h];
5732 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5733 			if (idx < s_idx)
5734 				goto cont;
5735 			idev = __in6_dev_get(dev);
5736 			if (!idev)
5737 				goto cont;
5738 			if (inet6_fill_ifinfo(skb, idev,
5739 					      NETLINK_CB(cb->skb).portid,
5740 					      cb->nlh->nlmsg_seq,
5741 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
5742 				goto out;
5743 cont:
5744 			idx++;
5745 		}
5746 	}
5747 out:
5748 	rcu_read_unlock();
5749 	cb->args[1] = idx;
5750 	cb->args[0] = h;
5751 
5752 	return skb->len;
5753 }
5754 
5755 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5756 {
5757 	struct sk_buff *skb;
5758 	struct net *net = dev_net(idev->dev);
5759 	int err = -ENOBUFS;
5760 
5761 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5762 	if (!skb)
5763 		goto errout;
5764 
5765 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5766 	if (err < 0) {
5767 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5768 		WARN_ON(err == -EMSGSIZE);
5769 		kfree_skb(skb);
5770 		goto errout;
5771 	}
5772 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5773 	return;
5774 errout:
5775 	if (err < 0)
5776 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5777 }
5778 
5779 static inline size_t inet6_prefix_nlmsg_size(void)
5780 {
5781 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
5782 	       + nla_total_size(sizeof(struct in6_addr))
5783 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
5784 }
5785 
5786 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5787 			     struct prefix_info *pinfo, u32 portid, u32 seq,
5788 			     int event, unsigned int flags)
5789 {
5790 	struct prefixmsg *pmsg;
5791 	struct nlmsghdr *nlh;
5792 	struct prefix_cacheinfo	ci;
5793 
5794 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5795 	if (!nlh)
5796 		return -EMSGSIZE;
5797 
5798 	pmsg = nlmsg_data(nlh);
5799 	pmsg->prefix_family = AF_INET6;
5800 	pmsg->prefix_pad1 = 0;
5801 	pmsg->prefix_pad2 = 0;
5802 	pmsg->prefix_ifindex = idev->dev->ifindex;
5803 	pmsg->prefix_len = pinfo->prefix_len;
5804 	pmsg->prefix_type = pinfo->type;
5805 	pmsg->prefix_pad3 = 0;
5806 	pmsg->prefix_flags = 0;
5807 	if (pinfo->onlink)
5808 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5809 	if (pinfo->autoconf)
5810 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5811 
5812 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5813 		goto nla_put_failure;
5814 	ci.preferred_time = ntohl(pinfo->prefered);
5815 	ci.valid_time = ntohl(pinfo->valid);
5816 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5817 		goto nla_put_failure;
5818 	nlmsg_end(skb, nlh);
5819 	return 0;
5820 
5821 nla_put_failure:
5822 	nlmsg_cancel(skb, nlh);
5823 	return -EMSGSIZE;
5824 }
5825 
5826 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5827 			 struct prefix_info *pinfo)
5828 {
5829 	struct sk_buff *skb;
5830 	struct net *net = dev_net(idev->dev);
5831 	int err = -ENOBUFS;
5832 
5833 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5834 	if (!skb)
5835 		goto errout;
5836 
5837 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5838 	if (err < 0) {
5839 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5840 		WARN_ON(err == -EMSGSIZE);
5841 		kfree_skb(skb);
5842 		goto errout;
5843 	}
5844 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5845 	return;
5846 errout:
5847 	if (err < 0)
5848 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5849 }
5850 
5851 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5852 {
5853 	struct net *net = dev_net(ifp->idev->dev);
5854 
5855 	if (event)
5856 		ASSERT_RTNL();
5857 
5858 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5859 
5860 	switch (event) {
5861 	case RTM_NEWADDR:
5862 		/*
5863 		 * If the address was optimistic
5864 		 * we inserted the route at the start of
5865 		 * our DAD process, so we don't need
5866 		 * to do it again
5867 		 */
5868 		if (!rcu_access_pointer(ifp->rt->fib6_node))
5869 			ip6_ins_rt(net, ifp->rt);
5870 		if (ifp->idev->cnf.forwarding)
5871 			addrconf_join_anycast(ifp);
5872 		if (!ipv6_addr_any(&ifp->peer_addr))
5873 			addrconf_prefix_route(&ifp->peer_addr, 128, 0,
5874 					      ifp->idev->dev, 0, 0,
5875 					      GFP_ATOMIC);
5876 		break;
5877 	case RTM_DELADDR:
5878 		if (ifp->idev->cnf.forwarding)
5879 			addrconf_leave_anycast(ifp);
5880 		addrconf_leave_solict(ifp->idev, &ifp->addr);
5881 		if (!ipv6_addr_any(&ifp->peer_addr)) {
5882 			struct fib6_info *rt;
5883 
5884 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5885 						       ifp->idev->dev, 0, 0);
5886 			if (rt)
5887 				ip6_del_rt(net, rt);
5888 		}
5889 		if (ifp->rt) {
5890 			ip6_del_rt(net, ifp->rt);
5891 			ifp->rt = NULL;
5892 		}
5893 		rt_genid_bump_ipv6(net);
5894 		break;
5895 	}
5896 	atomic_inc(&net->ipv6.dev_addr_genid);
5897 }
5898 
5899 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5900 {
5901 	rcu_read_lock_bh();
5902 	if (likely(ifp->idev->dead == 0))
5903 		__ipv6_ifa_notify(event, ifp);
5904 	rcu_read_unlock_bh();
5905 }
5906 
5907 #ifdef CONFIG_SYSCTL
5908 
5909 static
5910 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5911 			   void __user *buffer, size_t *lenp, loff_t *ppos)
5912 {
5913 	int *valp = ctl->data;
5914 	int val = *valp;
5915 	loff_t pos = *ppos;
5916 	struct ctl_table lctl;
5917 	int ret;
5918 
5919 	/*
5920 	 * ctl->data points to idev->cnf.forwarding, we should
5921 	 * not modify it until we get the rtnl lock.
5922 	 */
5923 	lctl = *ctl;
5924 	lctl.data = &val;
5925 
5926 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5927 
5928 	if (write)
5929 		ret = addrconf_fixup_forwarding(ctl, valp, val);
5930 	if (ret)
5931 		*ppos = pos;
5932 	return ret;
5933 }
5934 
5935 static
5936 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5937 			void __user *buffer, size_t *lenp, loff_t *ppos)
5938 {
5939 	struct inet6_dev *idev = ctl->extra1;
5940 	int min_mtu = IPV6_MIN_MTU;
5941 	struct ctl_table lctl;
5942 
5943 	lctl = *ctl;
5944 	lctl.extra1 = &min_mtu;
5945 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5946 
5947 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5948 }
5949 
5950 static void dev_disable_change(struct inet6_dev *idev)
5951 {
5952 	struct netdev_notifier_info info;
5953 
5954 	if (!idev || !idev->dev)
5955 		return;
5956 
5957 	netdev_notifier_info_init(&info, idev->dev);
5958 	if (idev->cnf.disable_ipv6)
5959 		addrconf_notify(NULL, NETDEV_DOWN, &info);
5960 	else
5961 		addrconf_notify(NULL, NETDEV_UP, &info);
5962 }
5963 
5964 static void addrconf_disable_change(struct net *net, __s32 newf)
5965 {
5966 	struct net_device *dev;
5967 	struct inet6_dev *idev;
5968 
5969 	for_each_netdev(net, dev) {
5970 		idev = __in6_dev_get(dev);
5971 		if (idev) {
5972 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5973 			idev->cnf.disable_ipv6 = newf;
5974 			if (changed)
5975 				dev_disable_change(idev);
5976 		}
5977 	}
5978 }
5979 
5980 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5981 {
5982 	struct net *net;
5983 	int old;
5984 
5985 	if (!rtnl_trylock())
5986 		return restart_syscall();
5987 
5988 	net = (struct net *)table->extra2;
5989 	old = *p;
5990 	*p = newf;
5991 
5992 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5993 		rtnl_unlock();
5994 		return 0;
5995 	}
5996 
5997 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
5998 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
5999 		addrconf_disable_change(net, newf);
6000 	} else if ((!newf) ^ (!old))
6001 		dev_disable_change((struct inet6_dev *)table->extra1);
6002 
6003 	rtnl_unlock();
6004 	return 0;
6005 }
6006 
6007 static
6008 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6009 			    void __user *buffer, size_t *lenp, loff_t *ppos)
6010 {
6011 	int *valp = ctl->data;
6012 	int val = *valp;
6013 	loff_t pos = *ppos;
6014 	struct ctl_table lctl;
6015 	int ret;
6016 
6017 	/*
6018 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6019 	 * not modify it until we get the rtnl lock.
6020 	 */
6021 	lctl = *ctl;
6022 	lctl.data = &val;
6023 
6024 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6025 
6026 	if (write)
6027 		ret = addrconf_disable_ipv6(ctl, valp, val);
6028 	if (ret)
6029 		*ppos = pos;
6030 	return ret;
6031 }
6032 
6033 static
6034 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6035 			      void __user *buffer, size_t *lenp, loff_t *ppos)
6036 {
6037 	int *valp = ctl->data;
6038 	int ret;
6039 	int old, new;
6040 
6041 	old = *valp;
6042 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6043 	new = *valp;
6044 
6045 	if (write && old != new) {
6046 		struct net *net = ctl->extra2;
6047 
6048 		if (!rtnl_trylock())
6049 			return restart_syscall();
6050 
6051 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6052 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6053 						     NETCONFA_PROXY_NEIGH,
6054 						     NETCONFA_IFINDEX_DEFAULT,
6055 						     net->ipv6.devconf_dflt);
6056 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6057 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6058 						     NETCONFA_PROXY_NEIGH,
6059 						     NETCONFA_IFINDEX_ALL,
6060 						     net->ipv6.devconf_all);
6061 		else {
6062 			struct inet6_dev *idev = ctl->extra1;
6063 
6064 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6065 						     NETCONFA_PROXY_NEIGH,
6066 						     idev->dev->ifindex,
6067 						     &idev->cnf);
6068 		}
6069 		rtnl_unlock();
6070 	}
6071 
6072 	return ret;
6073 }
6074 
6075 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6076 					 void __user *buffer, size_t *lenp,
6077 					 loff_t *ppos)
6078 {
6079 	int ret = 0;
6080 	u32 new_val;
6081 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6082 	struct net *net = (struct net *)ctl->extra2;
6083 	struct ctl_table tmp = {
6084 		.data = &new_val,
6085 		.maxlen = sizeof(new_val),
6086 		.mode = ctl->mode,
6087 	};
6088 
6089 	if (!rtnl_trylock())
6090 		return restart_syscall();
6091 
6092 	new_val = *((u32 *)ctl->data);
6093 
6094 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6095 	if (ret != 0)
6096 		goto out;
6097 
6098 	if (write) {
6099 		if (check_addr_gen_mode(new_val) < 0) {
6100 			ret = -EINVAL;
6101 			goto out;
6102 		}
6103 
6104 		if (idev) {
6105 			if (check_stable_privacy(idev, net, new_val) < 0) {
6106 				ret = -EINVAL;
6107 				goto out;
6108 			}
6109 
6110 			if (idev->cnf.addr_gen_mode != new_val) {
6111 				idev->cnf.addr_gen_mode = new_val;
6112 				addrconf_dev_config(idev->dev);
6113 			}
6114 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6115 			struct net_device *dev;
6116 
6117 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6118 			for_each_netdev(net, dev) {
6119 				idev = __in6_dev_get(dev);
6120 				if (idev &&
6121 				    idev->cnf.addr_gen_mode != new_val) {
6122 					idev->cnf.addr_gen_mode = new_val;
6123 					addrconf_dev_config(idev->dev);
6124 				}
6125 			}
6126 		}
6127 
6128 		*((u32 *)ctl->data) = new_val;
6129 	}
6130 
6131 out:
6132 	rtnl_unlock();
6133 
6134 	return ret;
6135 }
6136 
6137 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6138 					 void __user *buffer, size_t *lenp,
6139 					 loff_t *ppos)
6140 {
6141 	int err;
6142 	struct in6_addr addr;
6143 	char str[IPV6_MAX_STRLEN];
6144 	struct ctl_table lctl = *ctl;
6145 	struct net *net = ctl->extra2;
6146 	struct ipv6_stable_secret *secret = ctl->data;
6147 
6148 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6149 		return -EIO;
6150 
6151 	lctl.maxlen = IPV6_MAX_STRLEN;
6152 	lctl.data = str;
6153 
6154 	if (!rtnl_trylock())
6155 		return restart_syscall();
6156 
6157 	if (!write && !secret->initialized) {
6158 		err = -EIO;
6159 		goto out;
6160 	}
6161 
6162 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6163 	if (err >= sizeof(str)) {
6164 		err = -EIO;
6165 		goto out;
6166 	}
6167 
6168 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6169 	if (err || !write)
6170 		goto out;
6171 
6172 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6173 		err = -EIO;
6174 		goto out;
6175 	}
6176 
6177 	secret->initialized = true;
6178 	secret->secret = addr;
6179 
6180 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6181 		struct net_device *dev;
6182 
6183 		for_each_netdev(net, dev) {
6184 			struct inet6_dev *idev = __in6_dev_get(dev);
6185 
6186 			if (idev) {
6187 				idev->cnf.addr_gen_mode =
6188 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6189 			}
6190 		}
6191 	} else {
6192 		struct inet6_dev *idev = ctl->extra1;
6193 
6194 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6195 	}
6196 
6197 out:
6198 	rtnl_unlock();
6199 
6200 	return err;
6201 }
6202 
6203 static
6204 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6205 						int write,
6206 						void __user *buffer,
6207 						size_t *lenp,
6208 						loff_t *ppos)
6209 {
6210 	int *valp = ctl->data;
6211 	int val = *valp;
6212 	loff_t pos = *ppos;
6213 	struct ctl_table lctl;
6214 	int ret;
6215 
6216 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6217 	 * we should not modify it until we get the rtnl lock.
6218 	 */
6219 	lctl = *ctl;
6220 	lctl.data = &val;
6221 
6222 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6223 
6224 	if (write)
6225 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6226 	if (ret)
6227 		*ppos = pos;
6228 	return ret;
6229 }
6230 
6231 static
6232 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6233 {
6234 	if (rt) {
6235 		if (action)
6236 			rt->dst.flags |= DST_NOPOLICY;
6237 		else
6238 			rt->dst.flags &= ~DST_NOPOLICY;
6239 	}
6240 }
6241 
6242 static
6243 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6244 {
6245 	struct inet6_ifaddr *ifa;
6246 
6247 	read_lock_bh(&idev->lock);
6248 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6249 		spin_lock(&ifa->lock);
6250 		if (ifa->rt) {
6251 			struct fib6_info *rt = ifa->rt;
6252 			int cpu;
6253 
6254 			rcu_read_lock();
6255 			ifa->rt->dst_nopolicy = val ? true : false;
6256 			if (rt->rt6i_pcpu) {
6257 				for_each_possible_cpu(cpu) {
6258 					struct rt6_info **rtp;
6259 
6260 					rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6261 					addrconf_set_nopolicy(*rtp, val);
6262 				}
6263 			}
6264 			rcu_read_unlock();
6265 		}
6266 		spin_unlock(&ifa->lock);
6267 	}
6268 	read_unlock_bh(&idev->lock);
6269 }
6270 
6271 static
6272 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6273 {
6274 	struct inet6_dev *idev;
6275 	struct net *net;
6276 
6277 	if (!rtnl_trylock())
6278 		return restart_syscall();
6279 
6280 	*valp = val;
6281 
6282 	net = (struct net *)ctl->extra2;
6283 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6284 		rtnl_unlock();
6285 		return 0;
6286 	}
6287 
6288 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6289 		struct net_device *dev;
6290 
6291 		for_each_netdev(net, dev) {
6292 			idev = __in6_dev_get(dev);
6293 			if (idev)
6294 				addrconf_disable_policy_idev(idev, val);
6295 		}
6296 	} else {
6297 		idev = (struct inet6_dev *)ctl->extra1;
6298 		addrconf_disable_policy_idev(idev, val);
6299 	}
6300 
6301 	rtnl_unlock();
6302 	return 0;
6303 }
6304 
6305 static
6306 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6307 				   void __user *buffer, size_t *lenp,
6308 				   loff_t *ppos)
6309 {
6310 	int *valp = ctl->data;
6311 	int val = *valp;
6312 	loff_t pos = *ppos;
6313 	struct ctl_table lctl;
6314 	int ret;
6315 
6316 	lctl = *ctl;
6317 	lctl.data = &val;
6318 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6319 
6320 	if (write && (*valp != val))
6321 		ret = addrconf_disable_policy(ctl, valp, val);
6322 
6323 	if (ret)
6324 		*ppos = pos;
6325 
6326 	return ret;
6327 }
6328 
6329 static int minus_one = -1;
6330 static const int zero = 0;
6331 static const int one = 1;
6332 static const int two_five_five = 255;
6333 
6334 static const struct ctl_table addrconf_sysctl[] = {
6335 	{
6336 		.procname	= "forwarding",
6337 		.data		= &ipv6_devconf.forwarding,
6338 		.maxlen		= sizeof(int),
6339 		.mode		= 0644,
6340 		.proc_handler	= addrconf_sysctl_forward,
6341 	},
6342 	{
6343 		.procname	= "hop_limit",
6344 		.data		= &ipv6_devconf.hop_limit,
6345 		.maxlen		= sizeof(int),
6346 		.mode		= 0644,
6347 		.proc_handler	= proc_dointvec_minmax,
6348 		.extra1		= (void *)&one,
6349 		.extra2		= (void *)&two_five_five,
6350 	},
6351 	{
6352 		.procname	= "mtu",
6353 		.data		= &ipv6_devconf.mtu6,
6354 		.maxlen		= sizeof(int),
6355 		.mode		= 0644,
6356 		.proc_handler	= addrconf_sysctl_mtu,
6357 	},
6358 	{
6359 		.procname	= "accept_ra",
6360 		.data		= &ipv6_devconf.accept_ra,
6361 		.maxlen		= sizeof(int),
6362 		.mode		= 0644,
6363 		.proc_handler	= proc_dointvec,
6364 	},
6365 	{
6366 		.procname	= "accept_redirects",
6367 		.data		= &ipv6_devconf.accept_redirects,
6368 		.maxlen		= sizeof(int),
6369 		.mode		= 0644,
6370 		.proc_handler	= proc_dointvec,
6371 	},
6372 	{
6373 		.procname	= "autoconf",
6374 		.data		= &ipv6_devconf.autoconf,
6375 		.maxlen		= sizeof(int),
6376 		.mode		= 0644,
6377 		.proc_handler	= proc_dointvec,
6378 	},
6379 	{
6380 		.procname	= "dad_transmits",
6381 		.data		= &ipv6_devconf.dad_transmits,
6382 		.maxlen		= sizeof(int),
6383 		.mode		= 0644,
6384 		.proc_handler	= proc_dointvec,
6385 	},
6386 	{
6387 		.procname	= "router_solicitations",
6388 		.data		= &ipv6_devconf.rtr_solicits,
6389 		.maxlen		= sizeof(int),
6390 		.mode		= 0644,
6391 		.proc_handler	= proc_dointvec_minmax,
6392 		.extra1		= &minus_one,
6393 	},
6394 	{
6395 		.procname	= "router_solicitation_interval",
6396 		.data		= &ipv6_devconf.rtr_solicit_interval,
6397 		.maxlen		= sizeof(int),
6398 		.mode		= 0644,
6399 		.proc_handler	= proc_dointvec_jiffies,
6400 	},
6401 	{
6402 		.procname	= "router_solicitation_max_interval",
6403 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6404 		.maxlen		= sizeof(int),
6405 		.mode		= 0644,
6406 		.proc_handler	= proc_dointvec_jiffies,
6407 	},
6408 	{
6409 		.procname	= "router_solicitation_delay",
6410 		.data		= &ipv6_devconf.rtr_solicit_delay,
6411 		.maxlen		= sizeof(int),
6412 		.mode		= 0644,
6413 		.proc_handler	= proc_dointvec_jiffies,
6414 	},
6415 	{
6416 		.procname	= "force_mld_version",
6417 		.data		= &ipv6_devconf.force_mld_version,
6418 		.maxlen		= sizeof(int),
6419 		.mode		= 0644,
6420 		.proc_handler	= proc_dointvec,
6421 	},
6422 	{
6423 		.procname	= "mldv1_unsolicited_report_interval",
6424 		.data		=
6425 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6426 		.maxlen		= sizeof(int),
6427 		.mode		= 0644,
6428 		.proc_handler	= proc_dointvec_ms_jiffies,
6429 	},
6430 	{
6431 		.procname	= "mldv2_unsolicited_report_interval",
6432 		.data		=
6433 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6434 		.maxlen		= sizeof(int),
6435 		.mode		= 0644,
6436 		.proc_handler	= proc_dointvec_ms_jiffies,
6437 	},
6438 	{
6439 		.procname	= "use_tempaddr",
6440 		.data		= &ipv6_devconf.use_tempaddr,
6441 		.maxlen		= sizeof(int),
6442 		.mode		= 0644,
6443 		.proc_handler	= proc_dointvec,
6444 	},
6445 	{
6446 		.procname	= "temp_valid_lft",
6447 		.data		= &ipv6_devconf.temp_valid_lft,
6448 		.maxlen		= sizeof(int),
6449 		.mode		= 0644,
6450 		.proc_handler	= proc_dointvec,
6451 	},
6452 	{
6453 		.procname	= "temp_prefered_lft",
6454 		.data		= &ipv6_devconf.temp_prefered_lft,
6455 		.maxlen		= sizeof(int),
6456 		.mode		= 0644,
6457 		.proc_handler	= proc_dointvec,
6458 	},
6459 	{
6460 		.procname	= "regen_max_retry",
6461 		.data		= &ipv6_devconf.regen_max_retry,
6462 		.maxlen		= sizeof(int),
6463 		.mode		= 0644,
6464 		.proc_handler	= proc_dointvec,
6465 	},
6466 	{
6467 		.procname	= "max_desync_factor",
6468 		.data		= &ipv6_devconf.max_desync_factor,
6469 		.maxlen		= sizeof(int),
6470 		.mode		= 0644,
6471 		.proc_handler	= proc_dointvec,
6472 	},
6473 	{
6474 		.procname	= "max_addresses",
6475 		.data		= &ipv6_devconf.max_addresses,
6476 		.maxlen		= sizeof(int),
6477 		.mode		= 0644,
6478 		.proc_handler	= proc_dointvec,
6479 	},
6480 	{
6481 		.procname	= "accept_ra_defrtr",
6482 		.data		= &ipv6_devconf.accept_ra_defrtr,
6483 		.maxlen		= sizeof(int),
6484 		.mode		= 0644,
6485 		.proc_handler	= proc_dointvec,
6486 	},
6487 	{
6488 		.procname	= "accept_ra_min_hop_limit",
6489 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6490 		.maxlen		= sizeof(int),
6491 		.mode		= 0644,
6492 		.proc_handler	= proc_dointvec,
6493 	},
6494 	{
6495 		.procname	= "accept_ra_pinfo",
6496 		.data		= &ipv6_devconf.accept_ra_pinfo,
6497 		.maxlen		= sizeof(int),
6498 		.mode		= 0644,
6499 		.proc_handler	= proc_dointvec,
6500 	},
6501 #ifdef CONFIG_IPV6_ROUTER_PREF
6502 	{
6503 		.procname	= "accept_ra_rtr_pref",
6504 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6505 		.maxlen		= sizeof(int),
6506 		.mode		= 0644,
6507 		.proc_handler	= proc_dointvec,
6508 	},
6509 	{
6510 		.procname	= "router_probe_interval",
6511 		.data		= &ipv6_devconf.rtr_probe_interval,
6512 		.maxlen		= sizeof(int),
6513 		.mode		= 0644,
6514 		.proc_handler	= proc_dointvec_jiffies,
6515 	},
6516 #ifdef CONFIG_IPV6_ROUTE_INFO
6517 	{
6518 		.procname	= "accept_ra_rt_info_min_plen",
6519 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6520 		.maxlen		= sizeof(int),
6521 		.mode		= 0644,
6522 		.proc_handler	= proc_dointvec,
6523 	},
6524 	{
6525 		.procname	= "accept_ra_rt_info_max_plen",
6526 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6527 		.maxlen		= sizeof(int),
6528 		.mode		= 0644,
6529 		.proc_handler	= proc_dointvec,
6530 	},
6531 #endif
6532 #endif
6533 	{
6534 		.procname	= "proxy_ndp",
6535 		.data		= &ipv6_devconf.proxy_ndp,
6536 		.maxlen		= sizeof(int),
6537 		.mode		= 0644,
6538 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6539 	},
6540 	{
6541 		.procname	= "accept_source_route",
6542 		.data		= &ipv6_devconf.accept_source_route,
6543 		.maxlen		= sizeof(int),
6544 		.mode		= 0644,
6545 		.proc_handler	= proc_dointvec,
6546 	},
6547 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6548 	{
6549 		.procname	= "optimistic_dad",
6550 		.data		= &ipv6_devconf.optimistic_dad,
6551 		.maxlen		= sizeof(int),
6552 		.mode		= 0644,
6553 		.proc_handler   = proc_dointvec,
6554 	},
6555 	{
6556 		.procname	= "use_optimistic",
6557 		.data		= &ipv6_devconf.use_optimistic,
6558 		.maxlen		= sizeof(int),
6559 		.mode		= 0644,
6560 		.proc_handler	= proc_dointvec,
6561 	},
6562 #endif
6563 #ifdef CONFIG_IPV6_MROUTE
6564 	{
6565 		.procname	= "mc_forwarding",
6566 		.data		= &ipv6_devconf.mc_forwarding,
6567 		.maxlen		= sizeof(int),
6568 		.mode		= 0444,
6569 		.proc_handler	= proc_dointvec,
6570 	},
6571 #endif
6572 	{
6573 		.procname	= "disable_ipv6",
6574 		.data		= &ipv6_devconf.disable_ipv6,
6575 		.maxlen		= sizeof(int),
6576 		.mode		= 0644,
6577 		.proc_handler	= addrconf_sysctl_disable,
6578 	},
6579 	{
6580 		.procname	= "accept_dad",
6581 		.data		= &ipv6_devconf.accept_dad,
6582 		.maxlen		= sizeof(int),
6583 		.mode		= 0644,
6584 		.proc_handler	= proc_dointvec,
6585 	},
6586 	{
6587 		.procname	= "force_tllao",
6588 		.data		= &ipv6_devconf.force_tllao,
6589 		.maxlen		= sizeof(int),
6590 		.mode		= 0644,
6591 		.proc_handler	= proc_dointvec
6592 	},
6593 	{
6594 		.procname	= "ndisc_notify",
6595 		.data		= &ipv6_devconf.ndisc_notify,
6596 		.maxlen		= sizeof(int),
6597 		.mode		= 0644,
6598 		.proc_handler	= proc_dointvec
6599 	},
6600 	{
6601 		.procname	= "suppress_frag_ndisc",
6602 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6603 		.maxlen		= sizeof(int),
6604 		.mode		= 0644,
6605 		.proc_handler	= proc_dointvec
6606 	},
6607 	{
6608 		.procname	= "accept_ra_from_local",
6609 		.data		= &ipv6_devconf.accept_ra_from_local,
6610 		.maxlen		= sizeof(int),
6611 		.mode		= 0644,
6612 		.proc_handler	= proc_dointvec,
6613 	},
6614 	{
6615 		.procname	= "accept_ra_mtu",
6616 		.data		= &ipv6_devconf.accept_ra_mtu,
6617 		.maxlen		= sizeof(int),
6618 		.mode		= 0644,
6619 		.proc_handler	= proc_dointvec,
6620 	},
6621 	{
6622 		.procname	= "stable_secret",
6623 		.data		= &ipv6_devconf.stable_secret,
6624 		.maxlen		= IPV6_MAX_STRLEN,
6625 		.mode		= 0600,
6626 		.proc_handler	= addrconf_sysctl_stable_secret,
6627 	},
6628 	{
6629 		.procname	= "use_oif_addrs_only",
6630 		.data		= &ipv6_devconf.use_oif_addrs_only,
6631 		.maxlen		= sizeof(int),
6632 		.mode		= 0644,
6633 		.proc_handler	= proc_dointvec,
6634 	},
6635 	{
6636 		.procname	= "ignore_routes_with_linkdown",
6637 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6638 		.maxlen		= sizeof(int),
6639 		.mode		= 0644,
6640 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6641 	},
6642 	{
6643 		.procname	= "drop_unicast_in_l2_multicast",
6644 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6645 		.maxlen		= sizeof(int),
6646 		.mode		= 0644,
6647 		.proc_handler	= proc_dointvec,
6648 	},
6649 	{
6650 		.procname	= "drop_unsolicited_na",
6651 		.data		= &ipv6_devconf.drop_unsolicited_na,
6652 		.maxlen		= sizeof(int),
6653 		.mode		= 0644,
6654 		.proc_handler	= proc_dointvec,
6655 	},
6656 	{
6657 		.procname	= "keep_addr_on_down",
6658 		.data		= &ipv6_devconf.keep_addr_on_down,
6659 		.maxlen		= sizeof(int),
6660 		.mode		= 0644,
6661 		.proc_handler	= proc_dointvec,
6662 
6663 	},
6664 	{
6665 		.procname	= "seg6_enabled",
6666 		.data		= &ipv6_devconf.seg6_enabled,
6667 		.maxlen		= sizeof(int),
6668 		.mode		= 0644,
6669 		.proc_handler	= proc_dointvec,
6670 	},
6671 #ifdef CONFIG_IPV6_SEG6_HMAC
6672 	{
6673 		.procname	= "seg6_require_hmac",
6674 		.data		= &ipv6_devconf.seg6_require_hmac,
6675 		.maxlen		= sizeof(int),
6676 		.mode		= 0644,
6677 		.proc_handler	= proc_dointvec,
6678 	},
6679 #endif
6680 	{
6681 		.procname       = "enhanced_dad",
6682 		.data           = &ipv6_devconf.enhanced_dad,
6683 		.maxlen         = sizeof(int),
6684 		.mode           = 0644,
6685 		.proc_handler   = proc_dointvec,
6686 	},
6687 	{
6688 		.procname		= "addr_gen_mode",
6689 		.data			= &ipv6_devconf.addr_gen_mode,
6690 		.maxlen			= sizeof(int),
6691 		.mode			= 0644,
6692 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6693 	},
6694 	{
6695 		.procname       = "disable_policy",
6696 		.data           = &ipv6_devconf.disable_policy,
6697 		.maxlen         = sizeof(int),
6698 		.mode           = 0644,
6699 		.proc_handler   = addrconf_sysctl_disable_policy,
6700 	},
6701 	{
6702 		.procname	= "ndisc_tclass",
6703 		.data		= &ipv6_devconf.ndisc_tclass,
6704 		.maxlen		= sizeof(int),
6705 		.mode		= 0644,
6706 		.proc_handler	= proc_dointvec_minmax,
6707 		.extra1		= (void *)&zero,
6708 		.extra2		= (void *)&two_five_five,
6709 	},
6710 	{
6711 		/* sentinel */
6712 	}
6713 };
6714 
6715 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6716 		struct inet6_dev *idev, struct ipv6_devconf *p)
6717 {
6718 	int i, ifindex;
6719 	struct ctl_table *table;
6720 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6721 
6722 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6723 	if (!table)
6724 		goto out;
6725 
6726 	for (i = 0; table[i].data; i++) {
6727 		table[i].data += (char *)p - (char *)&ipv6_devconf;
6728 		/* If one of these is already set, then it is not safe to
6729 		 * overwrite either of them: this makes proc_dointvec_minmax
6730 		 * usable.
6731 		 */
6732 		if (!table[i].extra1 && !table[i].extra2) {
6733 			table[i].extra1 = idev; /* embedded; no ref */
6734 			table[i].extra2 = net;
6735 		}
6736 	}
6737 
6738 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6739 
6740 	p->sysctl_header = register_net_sysctl(net, path, table);
6741 	if (!p->sysctl_header)
6742 		goto free;
6743 
6744 	if (!strcmp(dev_name, "all"))
6745 		ifindex = NETCONFA_IFINDEX_ALL;
6746 	else if (!strcmp(dev_name, "default"))
6747 		ifindex = NETCONFA_IFINDEX_DEFAULT;
6748 	else
6749 		ifindex = idev->dev->ifindex;
6750 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6751 				     ifindex, p);
6752 	return 0;
6753 
6754 free:
6755 	kfree(table);
6756 out:
6757 	return -ENOBUFS;
6758 }
6759 
6760 static void __addrconf_sysctl_unregister(struct net *net,
6761 					 struct ipv6_devconf *p, int ifindex)
6762 {
6763 	struct ctl_table *table;
6764 
6765 	if (!p->sysctl_header)
6766 		return;
6767 
6768 	table = p->sysctl_header->ctl_table_arg;
6769 	unregister_net_sysctl_table(p->sysctl_header);
6770 	p->sysctl_header = NULL;
6771 	kfree(table);
6772 
6773 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6774 }
6775 
6776 static int addrconf_sysctl_register(struct inet6_dev *idev)
6777 {
6778 	int err;
6779 
6780 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
6781 		return -EINVAL;
6782 
6783 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6784 				    &ndisc_ifinfo_sysctl_change);
6785 	if (err)
6786 		return err;
6787 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6788 					 idev, &idev->cnf);
6789 	if (err)
6790 		neigh_sysctl_unregister(idev->nd_parms);
6791 
6792 	return err;
6793 }
6794 
6795 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6796 {
6797 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6798 				     idev->dev->ifindex);
6799 	neigh_sysctl_unregister(idev->nd_parms);
6800 }
6801 
6802 
6803 #endif
6804 
6805 static int __net_init addrconf_init_net(struct net *net)
6806 {
6807 	int err = -ENOMEM;
6808 	struct ipv6_devconf *all, *dflt;
6809 
6810 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6811 	if (!all)
6812 		goto err_alloc_all;
6813 
6814 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6815 	if (!dflt)
6816 		goto err_alloc_dflt;
6817 
6818 	/* these will be inherited by all namespaces */
6819 	dflt->autoconf = ipv6_defaults.autoconf;
6820 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6821 
6822 	dflt->stable_secret.initialized = false;
6823 	all->stable_secret.initialized = false;
6824 
6825 	net->ipv6.devconf_all = all;
6826 	net->ipv6.devconf_dflt = dflt;
6827 
6828 #ifdef CONFIG_SYSCTL
6829 	err = __addrconf_sysctl_register(net, "all", NULL, all);
6830 	if (err < 0)
6831 		goto err_reg_all;
6832 
6833 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6834 	if (err < 0)
6835 		goto err_reg_dflt;
6836 #endif
6837 	return 0;
6838 
6839 #ifdef CONFIG_SYSCTL
6840 err_reg_dflt:
6841 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6842 err_reg_all:
6843 	kfree(dflt);
6844 #endif
6845 err_alloc_dflt:
6846 	kfree(all);
6847 err_alloc_all:
6848 	return err;
6849 }
6850 
6851 static void __net_exit addrconf_exit_net(struct net *net)
6852 {
6853 #ifdef CONFIG_SYSCTL
6854 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6855 				     NETCONFA_IFINDEX_DEFAULT);
6856 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6857 				     NETCONFA_IFINDEX_ALL);
6858 #endif
6859 	kfree(net->ipv6.devconf_dflt);
6860 	kfree(net->ipv6.devconf_all);
6861 }
6862 
6863 static struct pernet_operations addrconf_ops = {
6864 	.init = addrconf_init_net,
6865 	.exit = addrconf_exit_net,
6866 };
6867 
6868 static struct rtnl_af_ops inet6_ops __read_mostly = {
6869 	.family		  = AF_INET6,
6870 	.fill_link_af	  = inet6_fill_link_af,
6871 	.get_link_af_size = inet6_get_link_af_size,
6872 	.validate_link_af = inet6_validate_link_af,
6873 	.set_link_af	  = inet6_set_link_af,
6874 };
6875 
6876 /*
6877  *	Init / cleanup code
6878  */
6879 
6880 int __init addrconf_init(void)
6881 {
6882 	struct inet6_dev *idev;
6883 	int i, err;
6884 
6885 	err = ipv6_addr_label_init();
6886 	if (err < 0) {
6887 		pr_crit("%s: cannot initialize default policy table: %d\n",
6888 			__func__, err);
6889 		goto out;
6890 	}
6891 
6892 	err = register_pernet_subsys(&addrconf_ops);
6893 	if (err < 0)
6894 		goto out_addrlabel;
6895 
6896 	addrconf_wq = create_workqueue("ipv6_addrconf");
6897 	if (!addrconf_wq) {
6898 		err = -ENOMEM;
6899 		goto out_nowq;
6900 	}
6901 
6902 	/* The addrconf netdev notifier requires that loopback_dev
6903 	 * has it's ipv6 private information allocated and setup
6904 	 * before it can bring up and give link-local addresses
6905 	 * to other devices which are up.
6906 	 *
6907 	 * Unfortunately, loopback_dev is not necessarily the first
6908 	 * entry in the global dev_base list of net devices.  In fact,
6909 	 * it is likely to be the very last entry on that list.
6910 	 * So this causes the notifier registry below to try and
6911 	 * give link-local addresses to all devices besides loopback_dev
6912 	 * first, then loopback_dev, which cases all the non-loopback_dev
6913 	 * devices to fail to get a link-local address.
6914 	 *
6915 	 * So, as a temporary fix, allocate the ipv6 structure for
6916 	 * loopback_dev first by hand.
6917 	 * Longer term, all of the dependencies ipv6 has upon the loopback
6918 	 * device and it being up should be removed.
6919 	 */
6920 	rtnl_lock();
6921 	idev = ipv6_add_dev(init_net.loopback_dev);
6922 	rtnl_unlock();
6923 	if (IS_ERR(idev)) {
6924 		err = PTR_ERR(idev);
6925 		goto errlo;
6926 	}
6927 
6928 	ip6_route_init_special_entries();
6929 
6930 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
6931 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6932 
6933 	register_netdevice_notifier(&ipv6_dev_notf);
6934 
6935 	addrconf_verify();
6936 
6937 	rtnl_af_register(&inet6_ops);
6938 
6939 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6940 				   NULL, inet6_dump_ifinfo, 0);
6941 	if (err < 0)
6942 		goto errout;
6943 
6944 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6945 				   inet6_rtm_newaddr, NULL, 0);
6946 	if (err < 0)
6947 		goto errout;
6948 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6949 				   inet6_rtm_deladdr, NULL, 0);
6950 	if (err < 0)
6951 		goto errout;
6952 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6953 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
6954 				   RTNL_FLAG_DOIT_UNLOCKED);
6955 	if (err < 0)
6956 		goto errout;
6957 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6958 				   NULL, inet6_dump_ifmcaddr, 0);
6959 	if (err < 0)
6960 		goto errout;
6961 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6962 				   NULL, inet6_dump_ifacaddr, 0);
6963 	if (err < 0)
6964 		goto errout;
6965 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6966 				   inet6_netconf_get_devconf,
6967 				   inet6_netconf_dump_devconf,
6968 				   RTNL_FLAG_DOIT_UNLOCKED);
6969 	if (err < 0)
6970 		goto errout;
6971 	err = ipv6_addr_label_rtnl_register();
6972 	if (err < 0)
6973 		goto errout;
6974 
6975 	return 0;
6976 errout:
6977 	rtnl_unregister_all(PF_INET6);
6978 	rtnl_af_unregister(&inet6_ops);
6979 	unregister_netdevice_notifier(&ipv6_dev_notf);
6980 errlo:
6981 	destroy_workqueue(addrconf_wq);
6982 out_nowq:
6983 	unregister_pernet_subsys(&addrconf_ops);
6984 out_addrlabel:
6985 	ipv6_addr_label_cleanup();
6986 out:
6987 	return err;
6988 }
6989 
6990 void addrconf_cleanup(void)
6991 {
6992 	struct net_device *dev;
6993 	int i;
6994 
6995 	unregister_netdevice_notifier(&ipv6_dev_notf);
6996 	unregister_pernet_subsys(&addrconf_ops);
6997 	ipv6_addr_label_cleanup();
6998 
6999 	rtnl_af_unregister(&inet6_ops);
7000 
7001 	rtnl_lock();
7002 
7003 	/* clean dev list */
7004 	for_each_netdev(&init_net, dev) {
7005 		if (__in6_dev_get(dev) == NULL)
7006 			continue;
7007 		addrconf_ifdown(dev, 1);
7008 	}
7009 	addrconf_ifdown(init_net.loopback_dev, 2);
7010 
7011 	/*
7012 	 *	Check hash table.
7013 	 */
7014 	spin_lock_bh(&addrconf_hash_lock);
7015 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7016 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7017 	spin_unlock_bh(&addrconf_hash_lock);
7018 	cancel_delayed_work(&addr_chk_work);
7019 	rtnl_unlock();
7020 
7021 	destroy_workqueue(addrconf_wq);
7022 }
7023