xref: /linux/net/ipv6/addrconf.c (revision fdaf9a5840acaab18694a19e0eb0aa51162eeeed)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 Address [auto]configuration
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  */
10 
11 /*
12  *	Changes:
13  *
14  *	Janos Farkas			:	delete timer on ifdown
15  *	<chexum@bankinf.banki.hu>
16  *	Andi Kleen			:	kill double kfree on module
17  *						unload.
18  *	Maciej W. Rozycki		:	FDDI support
19  *	sekiya@USAGI			:	Don't send too many RS
20  *						packets.
21  *	yoshfuji@USAGI			:       Fixed interval between DAD
22  *						packets.
23  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
24  *						address validation timer.
25  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
26  *						support.
27  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
28  *						address on a same interface.
29  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
30  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
31  *						seq_file.
32  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
33  *						selection; consider scope,
34  *						status etc.
35  */
36 
37 #define pr_fmt(fmt) "IPv6: " fmt
38 
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65 
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
69 
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
88 
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
92 #include <linux/ioam6.h>
93 
94 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
95 
96 #define IPV6_MAX_STRLEN \
97 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98 
99 static inline u32 cstamp_delta(unsigned long cstamp)
100 {
101 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102 }
103 
104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 {
106 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 	u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
108 	do_div(tmp, 1000000);
109 	return (s32)tmp;
110 }
111 
112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 {
114 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 	u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
116 	do_div(tmp, 1000000);
117 	if ((s32)tmp > mrt) {
118 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 		tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
120 		do_div(tmp, 1000000);
121 	}
122 	return (s32)tmp;
123 }
124 
125 #ifdef CONFIG_SYSCTL
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 #else
129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 {
131 	return 0;
132 }
133 
134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135 {
136 }
137 #endif
138 
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140 
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 					u8 dad_count,
145 					const struct inet6_dev *idev);
146 
147 #define IN6_ADDR_HSIZE_SHIFT	8
148 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
149 
150 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
152 
153 static struct workqueue_struct *addrconf_wq;
154 
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157 
158 static void addrconf_type_change(struct net_device *dev,
159 				 unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
161 
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163 						  int plen,
164 						  const struct net_device *dev,
165 						  u32 flags, u32 noflags,
166 						  bool no_gw);
167 
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171 				   bool send_na);
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176 
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 				struct prefix_info *pinfo);
179 
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
181 	.forwarding		= 0,
182 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
183 	.mtu6			= IPV6_MIN_MTU,
184 	.accept_ra		= 1,
185 	.accept_redirects	= 1,
186 	.autoconf		= 1,
187 	.force_mld_version	= 0,
188 	.mldv1_unsolicited_report_interval = 10 * HZ,
189 	.mldv2_unsolicited_report_interval = HZ,
190 	.dad_transmits		= 1,
191 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
192 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
193 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
195 	.use_tempaddr		= 0,
196 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
197 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
198 	.regen_max_retry	= REGEN_MAX_RETRY,
199 	.max_desync_factor	= MAX_DESYNC_FACTOR,
200 	.max_addresses		= IPV6_MAX_ADDRESSES,
201 	.accept_ra_defrtr	= 1,
202 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
203 	.accept_ra_from_local	= 0,
204 	.accept_ra_min_hop_limit= 1,
205 	.accept_ra_pinfo	= 1,
206 #ifdef CONFIG_IPV6_ROUTER_PREF
207 	.accept_ra_rtr_pref	= 1,
208 	.rtr_probe_interval	= 60 * HZ,
209 #ifdef CONFIG_IPV6_ROUTE_INFO
210 	.accept_ra_rt_info_min_plen = 0,
211 	.accept_ra_rt_info_max_plen = 0,
212 #endif
213 #endif
214 	.proxy_ndp		= 0,
215 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
216 	.disable_ipv6		= 0,
217 	.accept_dad		= 0,
218 	.suppress_frag_ndisc	= 1,
219 	.accept_ra_mtu		= 1,
220 	.stable_secret		= {
221 		.initialized = false,
222 	},
223 	.use_oif_addrs_only	= 0,
224 	.ignore_routes_with_linkdown = 0,
225 	.keep_addr_on_down	= 0,
226 	.seg6_enabled		= 0,
227 #ifdef CONFIG_IPV6_SEG6_HMAC
228 	.seg6_require_hmac	= 0,
229 #endif
230 	.enhanced_dad           = 1,
231 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
232 	.disable_policy		= 0,
233 	.rpl_seg_enabled	= 0,
234 	.ioam6_enabled		= 0,
235 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
236 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
237 	.ndisc_evict_nocarrier	= 1,
238 };
239 
240 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
241 	.forwarding		= 0,
242 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
243 	.mtu6			= IPV6_MIN_MTU,
244 	.accept_ra		= 1,
245 	.accept_redirects	= 1,
246 	.autoconf		= 1,
247 	.force_mld_version	= 0,
248 	.mldv1_unsolicited_report_interval = 10 * HZ,
249 	.mldv2_unsolicited_report_interval = HZ,
250 	.dad_transmits		= 1,
251 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
252 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
253 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
254 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
255 	.use_tempaddr		= 0,
256 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
257 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
258 	.regen_max_retry	= REGEN_MAX_RETRY,
259 	.max_desync_factor	= MAX_DESYNC_FACTOR,
260 	.max_addresses		= IPV6_MAX_ADDRESSES,
261 	.accept_ra_defrtr	= 1,
262 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
263 	.accept_ra_from_local	= 0,
264 	.accept_ra_min_hop_limit= 1,
265 	.accept_ra_pinfo	= 1,
266 #ifdef CONFIG_IPV6_ROUTER_PREF
267 	.accept_ra_rtr_pref	= 1,
268 	.rtr_probe_interval	= 60 * HZ,
269 #ifdef CONFIG_IPV6_ROUTE_INFO
270 	.accept_ra_rt_info_min_plen = 0,
271 	.accept_ra_rt_info_max_plen = 0,
272 #endif
273 #endif
274 	.proxy_ndp		= 0,
275 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
276 	.disable_ipv6		= 0,
277 	.accept_dad		= 1,
278 	.suppress_frag_ndisc	= 1,
279 	.accept_ra_mtu		= 1,
280 	.stable_secret		= {
281 		.initialized = false,
282 	},
283 	.use_oif_addrs_only	= 0,
284 	.ignore_routes_with_linkdown = 0,
285 	.keep_addr_on_down	= 0,
286 	.seg6_enabled		= 0,
287 #ifdef CONFIG_IPV6_SEG6_HMAC
288 	.seg6_require_hmac	= 0,
289 #endif
290 	.enhanced_dad           = 1,
291 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
292 	.disable_policy		= 0,
293 	.rpl_seg_enabled	= 0,
294 	.ioam6_enabled		= 0,
295 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
296 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
297 	.ndisc_evict_nocarrier	= 1,
298 };
299 
300 /* Check if link is ready: is it up and is a valid qdisc available */
301 static inline bool addrconf_link_ready(const struct net_device *dev)
302 {
303 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
304 }
305 
306 static void addrconf_del_rs_timer(struct inet6_dev *idev)
307 {
308 	if (del_timer(&idev->rs_timer))
309 		__in6_dev_put(idev);
310 }
311 
312 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
313 {
314 	if (cancel_delayed_work(&ifp->dad_work))
315 		__in6_ifa_put(ifp);
316 }
317 
318 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
319 				  unsigned long when)
320 {
321 	if (!timer_pending(&idev->rs_timer))
322 		in6_dev_hold(idev);
323 	mod_timer(&idev->rs_timer, jiffies + when);
324 }
325 
326 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
327 				   unsigned long delay)
328 {
329 	in6_ifa_hold(ifp);
330 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
331 		in6_ifa_put(ifp);
332 }
333 
334 static int snmp6_alloc_dev(struct inet6_dev *idev)
335 {
336 	int i;
337 
338 	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
339 	if (!idev->stats.ipv6)
340 		goto err_ip;
341 
342 	for_each_possible_cpu(i) {
343 		struct ipstats_mib *addrconf_stats;
344 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
345 		u64_stats_init(&addrconf_stats->syncp);
346 	}
347 
348 
349 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
350 					GFP_KERNEL);
351 	if (!idev->stats.icmpv6dev)
352 		goto err_icmp;
353 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
354 					   GFP_KERNEL);
355 	if (!idev->stats.icmpv6msgdev)
356 		goto err_icmpmsg;
357 
358 	return 0;
359 
360 err_icmpmsg:
361 	kfree(idev->stats.icmpv6dev);
362 err_icmp:
363 	free_percpu(idev->stats.ipv6);
364 err_ip:
365 	return -ENOMEM;
366 }
367 
368 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
369 {
370 	struct inet6_dev *ndev;
371 	int err = -ENOMEM;
372 
373 	ASSERT_RTNL();
374 
375 	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
376 		return ERR_PTR(-EINVAL);
377 
378 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
379 	if (!ndev)
380 		return ERR_PTR(err);
381 
382 	rwlock_init(&ndev->lock);
383 	ndev->dev = dev;
384 	INIT_LIST_HEAD(&ndev->addr_list);
385 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
386 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
387 
388 	if (ndev->cnf.stable_secret.initialized)
389 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
390 
391 	ndev->cnf.mtu6 = dev->mtu;
392 	ndev->ra_mtu = 0;
393 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
394 	if (!ndev->nd_parms) {
395 		kfree(ndev);
396 		return ERR_PTR(err);
397 	}
398 	if (ndev->cnf.forwarding)
399 		dev_disable_lro(dev);
400 	/* We refer to the device */
401 	dev_hold_track(dev, &ndev->dev_tracker, GFP_KERNEL);
402 
403 	if (snmp6_alloc_dev(ndev) < 0) {
404 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
405 			   __func__);
406 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
407 		dev_put_track(dev, &ndev->dev_tracker);
408 		kfree(ndev);
409 		return ERR_PTR(err);
410 	}
411 
412 	if (dev != blackhole_netdev) {
413 		if (snmp6_register_dev(ndev) < 0) {
414 			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
415 				   __func__, dev->name);
416 			goto err_release;
417 		}
418 	}
419 	/* One reference from device. */
420 	refcount_set(&ndev->refcnt, 1);
421 
422 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
423 		ndev->cnf.accept_dad = -1;
424 
425 #if IS_ENABLED(CONFIG_IPV6_SIT)
426 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
427 		pr_info("%s: Disabled Multicast RS\n", dev->name);
428 		ndev->cnf.rtr_solicits = 0;
429 	}
430 #endif
431 
432 	INIT_LIST_HEAD(&ndev->tempaddr_list);
433 	ndev->desync_factor = U32_MAX;
434 	if ((dev->flags&IFF_LOOPBACK) ||
435 	    dev->type == ARPHRD_TUNNEL ||
436 	    dev->type == ARPHRD_TUNNEL6 ||
437 	    dev->type == ARPHRD_SIT ||
438 	    dev->type == ARPHRD_NONE) {
439 		ndev->cnf.use_tempaddr = -1;
440 	}
441 
442 	ndev->token = in6addr_any;
443 
444 	if (netif_running(dev) && addrconf_link_ready(dev))
445 		ndev->if_flags |= IF_READY;
446 
447 	ipv6_mc_init_dev(ndev);
448 	ndev->tstamp = jiffies;
449 	if (dev != blackhole_netdev) {
450 		err = addrconf_sysctl_register(ndev);
451 		if (err) {
452 			ipv6_mc_destroy_dev(ndev);
453 			snmp6_unregister_dev(ndev);
454 			goto err_release;
455 		}
456 	}
457 	/* protected by rtnl_lock */
458 	rcu_assign_pointer(dev->ip6_ptr, ndev);
459 
460 	if (dev != blackhole_netdev) {
461 		/* Join interface-local all-node multicast group */
462 		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
463 
464 		/* Join all-node multicast group */
465 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
466 
467 		/* Join all-router multicast group if forwarding is set */
468 		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
469 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
470 	}
471 	return ndev;
472 
473 err_release:
474 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
475 	ndev->dead = 1;
476 	in6_dev_finish_destroy(ndev);
477 	return ERR_PTR(err);
478 }
479 
480 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
481 {
482 	struct inet6_dev *idev;
483 
484 	ASSERT_RTNL();
485 
486 	idev = __in6_dev_get(dev);
487 	if (!idev) {
488 		idev = ipv6_add_dev(dev);
489 		if (IS_ERR(idev))
490 			return idev;
491 	}
492 
493 	if (dev->flags&IFF_UP)
494 		ipv6_mc_up(idev);
495 	return idev;
496 }
497 
498 static int inet6_netconf_msgsize_devconf(int type)
499 {
500 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
501 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
502 	bool all = false;
503 
504 	if (type == NETCONFA_ALL)
505 		all = true;
506 
507 	if (all || type == NETCONFA_FORWARDING)
508 		size += nla_total_size(4);
509 #ifdef CONFIG_IPV6_MROUTE
510 	if (all || type == NETCONFA_MC_FORWARDING)
511 		size += nla_total_size(4);
512 #endif
513 	if (all || type == NETCONFA_PROXY_NEIGH)
514 		size += nla_total_size(4);
515 
516 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
517 		size += nla_total_size(4);
518 
519 	return size;
520 }
521 
522 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
523 				      struct ipv6_devconf *devconf, u32 portid,
524 				      u32 seq, int event, unsigned int flags,
525 				      int type)
526 {
527 	struct nlmsghdr  *nlh;
528 	struct netconfmsg *ncm;
529 	bool all = false;
530 
531 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
532 			flags);
533 	if (!nlh)
534 		return -EMSGSIZE;
535 
536 	if (type == NETCONFA_ALL)
537 		all = true;
538 
539 	ncm = nlmsg_data(nlh);
540 	ncm->ncm_family = AF_INET6;
541 
542 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
543 		goto nla_put_failure;
544 
545 	if (!devconf)
546 		goto out;
547 
548 	if ((all || type == NETCONFA_FORWARDING) &&
549 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
550 		goto nla_put_failure;
551 #ifdef CONFIG_IPV6_MROUTE
552 	if ((all || type == NETCONFA_MC_FORWARDING) &&
553 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
554 			atomic_read(&devconf->mc_forwarding)) < 0)
555 		goto nla_put_failure;
556 #endif
557 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
558 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
559 		goto nla_put_failure;
560 
561 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
562 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
563 			devconf->ignore_routes_with_linkdown) < 0)
564 		goto nla_put_failure;
565 
566 out:
567 	nlmsg_end(skb, nlh);
568 	return 0;
569 
570 nla_put_failure:
571 	nlmsg_cancel(skb, nlh);
572 	return -EMSGSIZE;
573 }
574 
575 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
576 				  int ifindex, struct ipv6_devconf *devconf)
577 {
578 	struct sk_buff *skb;
579 	int err = -ENOBUFS;
580 
581 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
582 	if (!skb)
583 		goto errout;
584 
585 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
586 					 event, 0, type);
587 	if (err < 0) {
588 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
589 		WARN_ON(err == -EMSGSIZE);
590 		kfree_skb(skb);
591 		goto errout;
592 	}
593 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
594 	return;
595 errout:
596 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
597 }
598 
599 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
600 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
601 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
602 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
603 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
604 };
605 
606 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
607 				       const struct nlmsghdr *nlh,
608 				       struct nlattr **tb,
609 				       struct netlink_ext_ack *extack)
610 {
611 	int i, err;
612 
613 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
614 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
615 		return -EINVAL;
616 	}
617 
618 	if (!netlink_strict_get_check(skb))
619 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
620 					      tb, NETCONFA_MAX,
621 					      devconf_ipv6_policy, extack);
622 
623 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
624 					    tb, NETCONFA_MAX,
625 					    devconf_ipv6_policy, extack);
626 	if (err)
627 		return err;
628 
629 	for (i = 0; i <= NETCONFA_MAX; i++) {
630 		if (!tb[i])
631 			continue;
632 
633 		switch (i) {
634 		case NETCONFA_IFINDEX:
635 			break;
636 		default:
637 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
638 			return -EINVAL;
639 		}
640 	}
641 
642 	return 0;
643 }
644 
645 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
646 				     struct nlmsghdr *nlh,
647 				     struct netlink_ext_ack *extack)
648 {
649 	struct net *net = sock_net(in_skb->sk);
650 	struct nlattr *tb[NETCONFA_MAX+1];
651 	struct inet6_dev *in6_dev = NULL;
652 	struct net_device *dev = NULL;
653 	struct sk_buff *skb;
654 	struct ipv6_devconf *devconf;
655 	int ifindex;
656 	int err;
657 
658 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
659 	if (err < 0)
660 		return err;
661 
662 	if (!tb[NETCONFA_IFINDEX])
663 		return -EINVAL;
664 
665 	err = -EINVAL;
666 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
667 	switch (ifindex) {
668 	case NETCONFA_IFINDEX_ALL:
669 		devconf = net->ipv6.devconf_all;
670 		break;
671 	case NETCONFA_IFINDEX_DEFAULT:
672 		devconf = net->ipv6.devconf_dflt;
673 		break;
674 	default:
675 		dev = dev_get_by_index(net, ifindex);
676 		if (!dev)
677 			return -EINVAL;
678 		in6_dev = in6_dev_get(dev);
679 		if (!in6_dev)
680 			goto errout;
681 		devconf = &in6_dev->cnf;
682 		break;
683 	}
684 
685 	err = -ENOBUFS;
686 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
687 	if (!skb)
688 		goto errout;
689 
690 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
691 					 NETLINK_CB(in_skb).portid,
692 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
693 					 NETCONFA_ALL);
694 	if (err < 0) {
695 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
696 		WARN_ON(err == -EMSGSIZE);
697 		kfree_skb(skb);
698 		goto errout;
699 	}
700 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
701 errout:
702 	if (in6_dev)
703 		in6_dev_put(in6_dev);
704 	dev_put(dev);
705 	return err;
706 }
707 
708 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
709 				      struct netlink_callback *cb)
710 {
711 	const struct nlmsghdr *nlh = cb->nlh;
712 	struct net *net = sock_net(skb->sk);
713 	int h, s_h;
714 	int idx, s_idx;
715 	struct net_device *dev;
716 	struct inet6_dev *idev;
717 	struct hlist_head *head;
718 
719 	if (cb->strict_check) {
720 		struct netlink_ext_ack *extack = cb->extack;
721 		struct netconfmsg *ncm;
722 
723 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
724 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
725 			return -EINVAL;
726 		}
727 
728 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
729 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
730 			return -EINVAL;
731 		}
732 	}
733 
734 	s_h = cb->args[0];
735 	s_idx = idx = cb->args[1];
736 
737 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
738 		idx = 0;
739 		head = &net->dev_index_head[h];
740 		rcu_read_lock();
741 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
742 			  net->dev_base_seq;
743 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
744 			if (idx < s_idx)
745 				goto cont;
746 			idev = __in6_dev_get(dev);
747 			if (!idev)
748 				goto cont;
749 
750 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
751 						       &idev->cnf,
752 						       NETLINK_CB(cb->skb).portid,
753 						       nlh->nlmsg_seq,
754 						       RTM_NEWNETCONF,
755 						       NLM_F_MULTI,
756 						       NETCONFA_ALL) < 0) {
757 				rcu_read_unlock();
758 				goto done;
759 			}
760 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
761 cont:
762 			idx++;
763 		}
764 		rcu_read_unlock();
765 	}
766 	if (h == NETDEV_HASHENTRIES) {
767 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
768 					       net->ipv6.devconf_all,
769 					       NETLINK_CB(cb->skb).portid,
770 					       nlh->nlmsg_seq,
771 					       RTM_NEWNETCONF, NLM_F_MULTI,
772 					       NETCONFA_ALL) < 0)
773 			goto done;
774 		else
775 			h++;
776 	}
777 	if (h == NETDEV_HASHENTRIES + 1) {
778 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
779 					       net->ipv6.devconf_dflt,
780 					       NETLINK_CB(cb->skb).portid,
781 					       nlh->nlmsg_seq,
782 					       RTM_NEWNETCONF, NLM_F_MULTI,
783 					       NETCONFA_ALL) < 0)
784 			goto done;
785 		else
786 			h++;
787 	}
788 done:
789 	cb->args[0] = h;
790 	cb->args[1] = idx;
791 
792 	return skb->len;
793 }
794 
795 #ifdef CONFIG_SYSCTL
796 static void dev_forward_change(struct inet6_dev *idev)
797 {
798 	struct net_device *dev;
799 	struct inet6_ifaddr *ifa;
800 
801 	if (!idev)
802 		return;
803 	dev = idev->dev;
804 	if (idev->cnf.forwarding)
805 		dev_disable_lro(dev);
806 	if (dev->flags & IFF_MULTICAST) {
807 		if (idev->cnf.forwarding) {
808 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
809 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
810 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
811 		} else {
812 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
813 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
814 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
815 		}
816 	}
817 
818 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
819 		if (ifa->flags&IFA_F_TENTATIVE)
820 			continue;
821 		if (idev->cnf.forwarding)
822 			addrconf_join_anycast(ifa);
823 		else
824 			addrconf_leave_anycast(ifa);
825 	}
826 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
827 				     NETCONFA_FORWARDING,
828 				     dev->ifindex, &idev->cnf);
829 }
830 
831 
832 static void addrconf_forward_change(struct net *net, __s32 newf)
833 {
834 	struct net_device *dev;
835 	struct inet6_dev *idev;
836 
837 	for_each_netdev(net, dev) {
838 		idev = __in6_dev_get(dev);
839 		if (idev) {
840 			int changed = (!idev->cnf.forwarding) ^ (!newf);
841 			idev->cnf.forwarding = newf;
842 			if (changed)
843 				dev_forward_change(idev);
844 		}
845 	}
846 }
847 
848 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
849 {
850 	struct net *net;
851 	int old;
852 
853 	if (!rtnl_trylock())
854 		return restart_syscall();
855 
856 	net = (struct net *)table->extra2;
857 	old = *p;
858 	*p = newf;
859 
860 	if (p == &net->ipv6.devconf_dflt->forwarding) {
861 		if ((!newf) ^ (!old))
862 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
863 						     NETCONFA_FORWARDING,
864 						     NETCONFA_IFINDEX_DEFAULT,
865 						     net->ipv6.devconf_dflt);
866 		rtnl_unlock();
867 		return 0;
868 	}
869 
870 	if (p == &net->ipv6.devconf_all->forwarding) {
871 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
872 
873 		net->ipv6.devconf_dflt->forwarding = newf;
874 		if ((!newf) ^ (!old_dflt))
875 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
876 						     NETCONFA_FORWARDING,
877 						     NETCONFA_IFINDEX_DEFAULT,
878 						     net->ipv6.devconf_dflt);
879 
880 		addrconf_forward_change(net, newf);
881 		if ((!newf) ^ (!old))
882 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
883 						     NETCONFA_FORWARDING,
884 						     NETCONFA_IFINDEX_ALL,
885 						     net->ipv6.devconf_all);
886 	} else if ((!newf) ^ (!old))
887 		dev_forward_change((struct inet6_dev *)table->extra1);
888 	rtnl_unlock();
889 
890 	if (newf)
891 		rt6_purge_dflt_routers(net);
892 	return 1;
893 }
894 
895 static void addrconf_linkdown_change(struct net *net, __s32 newf)
896 {
897 	struct net_device *dev;
898 	struct inet6_dev *idev;
899 
900 	for_each_netdev(net, dev) {
901 		idev = __in6_dev_get(dev);
902 		if (idev) {
903 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
904 
905 			idev->cnf.ignore_routes_with_linkdown = newf;
906 			if (changed)
907 				inet6_netconf_notify_devconf(dev_net(dev),
908 							     RTM_NEWNETCONF,
909 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
910 							     dev->ifindex,
911 							     &idev->cnf);
912 		}
913 	}
914 }
915 
916 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
917 {
918 	struct net *net;
919 	int old;
920 
921 	if (!rtnl_trylock())
922 		return restart_syscall();
923 
924 	net = (struct net *)table->extra2;
925 	old = *p;
926 	*p = newf;
927 
928 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
929 		if ((!newf) ^ (!old))
930 			inet6_netconf_notify_devconf(net,
931 						     RTM_NEWNETCONF,
932 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
933 						     NETCONFA_IFINDEX_DEFAULT,
934 						     net->ipv6.devconf_dflt);
935 		rtnl_unlock();
936 		return 0;
937 	}
938 
939 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
940 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
941 		addrconf_linkdown_change(net, newf);
942 		if ((!newf) ^ (!old))
943 			inet6_netconf_notify_devconf(net,
944 						     RTM_NEWNETCONF,
945 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
946 						     NETCONFA_IFINDEX_ALL,
947 						     net->ipv6.devconf_all);
948 	}
949 	rtnl_unlock();
950 
951 	return 1;
952 }
953 
954 #endif
955 
956 /* Nobody refers to this ifaddr, destroy it */
957 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
958 {
959 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
960 
961 #ifdef NET_REFCNT_DEBUG
962 	pr_debug("%s\n", __func__);
963 #endif
964 
965 	in6_dev_put(ifp->idev);
966 
967 	if (cancel_delayed_work(&ifp->dad_work))
968 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
969 			  ifp);
970 
971 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
972 		pr_warn("Freeing alive inet6 address %p\n", ifp);
973 		return;
974 	}
975 
976 	kfree_rcu(ifp, rcu);
977 }
978 
979 static void
980 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
981 {
982 	struct list_head *p;
983 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
984 
985 	/*
986 	 * Each device address list is sorted in order of scope -
987 	 * global before linklocal.
988 	 */
989 	list_for_each(p, &idev->addr_list) {
990 		struct inet6_ifaddr *ifa
991 			= list_entry(p, struct inet6_ifaddr, if_list);
992 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
993 			break;
994 	}
995 
996 	list_add_tail_rcu(&ifp->if_list, p);
997 }
998 
999 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1000 {
1001 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1002 
1003 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1004 }
1005 
1006 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1007 			       struct net_device *dev, unsigned int hash)
1008 {
1009 	struct inet6_ifaddr *ifp;
1010 
1011 	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1012 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1013 			if (!dev || ifp->idev->dev == dev)
1014 				return true;
1015 		}
1016 	}
1017 	return false;
1018 }
1019 
1020 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1021 {
1022 	struct net *net = dev_net(dev);
1023 	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1024 	int err = 0;
1025 
1026 	spin_lock(&net->ipv6.addrconf_hash_lock);
1027 
1028 	/* Ignore adding duplicate addresses on an interface */
1029 	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1030 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1031 		err = -EEXIST;
1032 	} else {
1033 		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1034 	}
1035 
1036 	spin_unlock(&net->ipv6.addrconf_hash_lock);
1037 
1038 	return err;
1039 }
1040 
1041 /* On success it returns ifp with increased reference count */
1042 
1043 static struct inet6_ifaddr *
1044 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1045 	      bool can_block, struct netlink_ext_ack *extack)
1046 {
1047 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1048 	int addr_type = ipv6_addr_type(cfg->pfx);
1049 	struct net *net = dev_net(idev->dev);
1050 	struct inet6_ifaddr *ifa = NULL;
1051 	struct fib6_info *f6i = NULL;
1052 	int err = 0;
1053 
1054 	if (addr_type == IPV6_ADDR_ANY ||
1055 	    (addr_type & IPV6_ADDR_MULTICAST &&
1056 	     !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1057 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1058 	     !netif_is_l3_master(idev->dev) &&
1059 	     addr_type & IPV6_ADDR_LOOPBACK))
1060 		return ERR_PTR(-EADDRNOTAVAIL);
1061 
1062 	if (idev->dead) {
1063 		err = -ENODEV;			/*XXX*/
1064 		goto out;
1065 	}
1066 
1067 	if (idev->cnf.disable_ipv6) {
1068 		err = -EACCES;
1069 		goto out;
1070 	}
1071 
1072 	/* validator notifier needs to be blocking;
1073 	 * do not call in atomic context
1074 	 */
1075 	if (can_block) {
1076 		struct in6_validator_info i6vi = {
1077 			.i6vi_addr = *cfg->pfx,
1078 			.i6vi_dev = idev,
1079 			.extack = extack,
1080 		};
1081 
1082 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1083 		err = notifier_to_errno(err);
1084 		if (err < 0)
1085 			goto out;
1086 	}
1087 
1088 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1089 	if (!ifa) {
1090 		err = -ENOBUFS;
1091 		goto out;
1092 	}
1093 
1094 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1095 	if (IS_ERR(f6i)) {
1096 		err = PTR_ERR(f6i);
1097 		f6i = NULL;
1098 		goto out;
1099 	}
1100 
1101 	if (net->ipv6.devconf_all->disable_policy ||
1102 	    idev->cnf.disable_policy)
1103 		f6i->dst_nopolicy = true;
1104 
1105 	neigh_parms_data_state_setall(idev->nd_parms);
1106 
1107 	ifa->addr = *cfg->pfx;
1108 	if (cfg->peer_pfx)
1109 		ifa->peer_addr = *cfg->peer_pfx;
1110 
1111 	spin_lock_init(&ifa->lock);
1112 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1113 	INIT_HLIST_NODE(&ifa->addr_lst);
1114 	ifa->scope = cfg->scope;
1115 	ifa->prefix_len = cfg->plen;
1116 	ifa->rt_priority = cfg->rt_priority;
1117 	ifa->flags = cfg->ifa_flags;
1118 	ifa->ifa_proto = cfg->ifa_proto;
1119 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1120 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1121 		ifa->flags |= IFA_F_TENTATIVE;
1122 	ifa->valid_lft = cfg->valid_lft;
1123 	ifa->prefered_lft = cfg->preferred_lft;
1124 	ifa->cstamp = ifa->tstamp = jiffies;
1125 	ifa->tokenized = false;
1126 
1127 	ifa->rt = f6i;
1128 
1129 	ifa->idev = idev;
1130 	in6_dev_hold(idev);
1131 
1132 	/* For caller */
1133 	refcount_set(&ifa->refcnt, 1);
1134 
1135 	rcu_read_lock_bh();
1136 
1137 	err = ipv6_add_addr_hash(idev->dev, ifa);
1138 	if (err < 0) {
1139 		rcu_read_unlock_bh();
1140 		goto out;
1141 	}
1142 
1143 	write_lock(&idev->lock);
1144 
1145 	/* Add to inet6_dev unicast addr list. */
1146 	ipv6_link_dev_addr(idev, ifa);
1147 
1148 	if (ifa->flags&IFA_F_TEMPORARY) {
1149 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1150 		in6_ifa_hold(ifa);
1151 	}
1152 
1153 	in6_ifa_hold(ifa);
1154 	write_unlock(&idev->lock);
1155 
1156 	rcu_read_unlock_bh();
1157 
1158 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1159 out:
1160 	if (unlikely(err < 0)) {
1161 		fib6_info_release(f6i);
1162 
1163 		if (ifa) {
1164 			if (ifa->idev)
1165 				in6_dev_put(ifa->idev);
1166 			kfree(ifa);
1167 		}
1168 		ifa = ERR_PTR(err);
1169 	}
1170 
1171 	return ifa;
1172 }
1173 
1174 enum cleanup_prefix_rt_t {
1175 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1176 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1177 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1178 };
1179 
1180 /*
1181  * Check, whether the prefix for ifp would still need a prefix route
1182  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1183  * constants.
1184  *
1185  * 1) we don't purge prefix if address was not permanent.
1186  *    prefix is managed by its own lifetime.
1187  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1188  * 3) if there are no addresses, delete prefix.
1189  * 4) if there are still other permanent address(es),
1190  *    corresponding prefix is still permanent.
1191  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1192  *    don't purge the prefix, assume user space is managing it.
1193  * 6) otherwise, update prefix lifetime to the
1194  *    longest valid lifetime among the corresponding
1195  *    addresses on the device.
1196  *    Note: subsequent RA will update lifetime.
1197  **/
1198 static enum cleanup_prefix_rt_t
1199 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1200 {
1201 	struct inet6_ifaddr *ifa;
1202 	struct inet6_dev *idev = ifp->idev;
1203 	unsigned long lifetime;
1204 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1205 
1206 	*expires = jiffies;
1207 
1208 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1209 		if (ifa == ifp)
1210 			continue;
1211 		if (ifa->prefix_len != ifp->prefix_len ||
1212 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1213 				       ifp->prefix_len))
1214 			continue;
1215 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1216 			return CLEANUP_PREFIX_RT_NOP;
1217 
1218 		action = CLEANUP_PREFIX_RT_EXPIRE;
1219 
1220 		spin_lock(&ifa->lock);
1221 
1222 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1223 		/*
1224 		 * Note: Because this address is
1225 		 * not permanent, lifetime <
1226 		 * LONG_MAX / HZ here.
1227 		 */
1228 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1229 			*expires = ifa->tstamp + lifetime * HZ;
1230 		spin_unlock(&ifa->lock);
1231 	}
1232 
1233 	return action;
1234 }
1235 
1236 static void
1237 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1238 		     bool del_rt, bool del_peer)
1239 {
1240 	struct fib6_info *f6i;
1241 
1242 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1243 					ifp->prefix_len,
1244 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1245 	if (f6i) {
1246 		if (del_rt)
1247 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1248 		else {
1249 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1250 				fib6_set_expires(f6i, expires);
1251 			fib6_info_release(f6i);
1252 		}
1253 	}
1254 }
1255 
1256 
1257 /* This function wants to get referenced ifp and releases it before return */
1258 
1259 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1260 {
1261 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1262 	struct net *net = dev_net(ifp->idev->dev);
1263 	unsigned long expires;
1264 	int state;
1265 
1266 	ASSERT_RTNL();
1267 
1268 	spin_lock_bh(&ifp->lock);
1269 	state = ifp->state;
1270 	ifp->state = INET6_IFADDR_STATE_DEAD;
1271 	spin_unlock_bh(&ifp->lock);
1272 
1273 	if (state == INET6_IFADDR_STATE_DEAD)
1274 		goto out;
1275 
1276 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1277 	hlist_del_init_rcu(&ifp->addr_lst);
1278 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1279 
1280 	write_lock_bh(&ifp->idev->lock);
1281 
1282 	if (ifp->flags&IFA_F_TEMPORARY) {
1283 		list_del(&ifp->tmp_list);
1284 		if (ifp->ifpub) {
1285 			in6_ifa_put(ifp->ifpub);
1286 			ifp->ifpub = NULL;
1287 		}
1288 		__in6_ifa_put(ifp);
1289 	}
1290 
1291 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1292 		action = check_cleanup_prefix_route(ifp, &expires);
1293 
1294 	list_del_rcu(&ifp->if_list);
1295 	__in6_ifa_put(ifp);
1296 
1297 	write_unlock_bh(&ifp->idev->lock);
1298 
1299 	addrconf_del_dad_work(ifp);
1300 
1301 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1302 
1303 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1304 
1305 	if (action != CLEANUP_PREFIX_RT_NOP) {
1306 		cleanup_prefix_route(ifp, expires,
1307 			action == CLEANUP_PREFIX_RT_DEL, false);
1308 	}
1309 
1310 	/* clean up prefsrc entries */
1311 	rt6_remove_prefsrc(ifp);
1312 out:
1313 	in6_ifa_put(ifp);
1314 }
1315 
1316 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1317 {
1318 	struct inet6_dev *idev = ifp->idev;
1319 	unsigned long tmp_tstamp, age;
1320 	unsigned long regen_advance;
1321 	unsigned long now = jiffies;
1322 	s32 cnf_temp_preferred_lft;
1323 	struct inet6_ifaddr *ift;
1324 	struct ifa6_config cfg;
1325 	long max_desync_factor;
1326 	struct in6_addr addr;
1327 	int ret = 0;
1328 
1329 	write_lock_bh(&idev->lock);
1330 
1331 retry:
1332 	in6_dev_hold(idev);
1333 	if (idev->cnf.use_tempaddr <= 0) {
1334 		write_unlock_bh(&idev->lock);
1335 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1336 		in6_dev_put(idev);
1337 		ret = -1;
1338 		goto out;
1339 	}
1340 	spin_lock_bh(&ifp->lock);
1341 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1342 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1343 		spin_unlock_bh(&ifp->lock);
1344 		write_unlock_bh(&idev->lock);
1345 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1346 			__func__);
1347 		in6_dev_put(idev);
1348 		ret = -1;
1349 		goto out;
1350 	}
1351 	in6_ifa_hold(ifp);
1352 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1353 	ipv6_gen_rnd_iid(&addr);
1354 
1355 	age = (now - ifp->tstamp) / HZ;
1356 
1357 	regen_advance = idev->cnf.regen_max_retry *
1358 			idev->cnf.dad_transmits *
1359 			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1360 
1361 	/* recalculate max_desync_factor each time and update
1362 	 * idev->desync_factor if it's larger
1363 	 */
1364 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1365 	max_desync_factor = min_t(__u32,
1366 				  idev->cnf.max_desync_factor,
1367 				  cnf_temp_preferred_lft - regen_advance);
1368 
1369 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1370 		if (max_desync_factor > 0) {
1371 			get_random_bytes(&idev->desync_factor,
1372 					 sizeof(idev->desync_factor));
1373 			idev->desync_factor %= max_desync_factor;
1374 		} else {
1375 			idev->desync_factor = 0;
1376 		}
1377 	}
1378 
1379 	memset(&cfg, 0, sizeof(cfg));
1380 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1381 			      idev->cnf.temp_valid_lft + age);
1382 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1383 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1384 
1385 	cfg.plen = ifp->prefix_len;
1386 	tmp_tstamp = ifp->tstamp;
1387 	spin_unlock_bh(&ifp->lock);
1388 
1389 	write_unlock_bh(&idev->lock);
1390 
1391 	/* A temporary address is created only if this calculated Preferred
1392 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1393 	 * an implementation must not create a temporary address with a zero
1394 	 * Preferred Lifetime.
1395 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1396 	 * temporary addresses being generated.
1397 	 */
1398 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1399 	if (cfg.preferred_lft <= regen_advance + age) {
1400 		in6_ifa_put(ifp);
1401 		in6_dev_put(idev);
1402 		ret = -1;
1403 		goto out;
1404 	}
1405 
1406 	cfg.ifa_flags = IFA_F_TEMPORARY;
1407 	/* set in addrconf_prefix_rcv() */
1408 	if (ifp->flags & IFA_F_OPTIMISTIC)
1409 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1410 
1411 	cfg.pfx = &addr;
1412 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1413 
1414 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1415 	if (IS_ERR(ift)) {
1416 		in6_ifa_put(ifp);
1417 		in6_dev_put(idev);
1418 		pr_info("%s: retry temporary address regeneration\n", __func__);
1419 		write_lock_bh(&idev->lock);
1420 		goto retry;
1421 	}
1422 
1423 	spin_lock_bh(&ift->lock);
1424 	ift->ifpub = ifp;
1425 	ift->cstamp = now;
1426 	ift->tstamp = tmp_tstamp;
1427 	spin_unlock_bh(&ift->lock);
1428 
1429 	addrconf_dad_start(ift);
1430 	in6_ifa_put(ift);
1431 	in6_dev_put(idev);
1432 out:
1433 	return ret;
1434 }
1435 
1436 /*
1437  *	Choose an appropriate source address (RFC3484)
1438  */
1439 enum {
1440 	IPV6_SADDR_RULE_INIT = 0,
1441 	IPV6_SADDR_RULE_LOCAL,
1442 	IPV6_SADDR_RULE_SCOPE,
1443 	IPV6_SADDR_RULE_PREFERRED,
1444 #ifdef CONFIG_IPV6_MIP6
1445 	IPV6_SADDR_RULE_HOA,
1446 #endif
1447 	IPV6_SADDR_RULE_OIF,
1448 	IPV6_SADDR_RULE_LABEL,
1449 	IPV6_SADDR_RULE_PRIVACY,
1450 	IPV6_SADDR_RULE_ORCHID,
1451 	IPV6_SADDR_RULE_PREFIX,
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1454 #endif
1455 	IPV6_SADDR_RULE_MAX
1456 };
1457 
1458 struct ipv6_saddr_score {
1459 	int			rule;
1460 	int			addr_type;
1461 	struct inet6_ifaddr	*ifa;
1462 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1463 	int			scopedist;
1464 	int			matchlen;
1465 };
1466 
1467 struct ipv6_saddr_dst {
1468 	const struct in6_addr *addr;
1469 	int ifindex;
1470 	int scope;
1471 	int label;
1472 	unsigned int prefs;
1473 };
1474 
1475 static inline int ipv6_saddr_preferred(int type)
1476 {
1477 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1478 		return 1;
1479 	return 0;
1480 }
1481 
1482 static bool ipv6_use_optimistic_addr(struct net *net,
1483 				     struct inet6_dev *idev)
1484 {
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486 	if (!idev)
1487 		return false;
1488 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1489 		return false;
1490 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1491 		return false;
1492 
1493 	return true;
1494 #else
1495 	return false;
1496 #endif
1497 }
1498 
1499 static bool ipv6_allow_optimistic_dad(struct net *net,
1500 				      struct inet6_dev *idev)
1501 {
1502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503 	if (!idev)
1504 		return false;
1505 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1506 		return false;
1507 
1508 	return true;
1509 #else
1510 	return false;
1511 #endif
1512 }
1513 
1514 static int ipv6_get_saddr_eval(struct net *net,
1515 			       struct ipv6_saddr_score *score,
1516 			       struct ipv6_saddr_dst *dst,
1517 			       int i)
1518 {
1519 	int ret;
1520 
1521 	if (i <= score->rule) {
1522 		switch (i) {
1523 		case IPV6_SADDR_RULE_SCOPE:
1524 			ret = score->scopedist;
1525 			break;
1526 		case IPV6_SADDR_RULE_PREFIX:
1527 			ret = score->matchlen;
1528 			break;
1529 		default:
1530 			ret = !!test_bit(i, score->scorebits);
1531 		}
1532 		goto out;
1533 	}
1534 
1535 	switch (i) {
1536 	case IPV6_SADDR_RULE_INIT:
1537 		/* Rule 0: remember if hiscore is not ready yet */
1538 		ret = !!score->ifa;
1539 		break;
1540 	case IPV6_SADDR_RULE_LOCAL:
1541 		/* Rule 1: Prefer same address */
1542 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1543 		break;
1544 	case IPV6_SADDR_RULE_SCOPE:
1545 		/* Rule 2: Prefer appropriate scope
1546 		 *
1547 		 *      ret
1548 		 *       ^
1549 		 *    -1 |  d 15
1550 		 *    ---+--+-+---> scope
1551 		 *       |
1552 		 *       |             d is scope of the destination.
1553 		 *  B-d  |  \
1554 		 *       |   \      <- smaller scope is better if
1555 		 *  B-15 |    \        if scope is enough for destination.
1556 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1557 		 * d-C-1 | /
1558 		 *       |/         <- greater is better
1559 		 *   -C  /             if scope is not enough for destination.
1560 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1561 		 *
1562 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1563 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1564 		 * Assume B = 0 and we get C > 29.
1565 		 */
1566 		ret = __ipv6_addr_src_scope(score->addr_type);
1567 		if (ret >= dst->scope)
1568 			ret = -ret;
1569 		else
1570 			ret -= 128;	/* 30 is enough */
1571 		score->scopedist = ret;
1572 		break;
1573 	case IPV6_SADDR_RULE_PREFERRED:
1574 	    {
1575 		/* Rule 3: Avoid deprecated and optimistic addresses */
1576 		u8 avoid = IFA_F_DEPRECATED;
1577 
1578 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1579 			avoid |= IFA_F_OPTIMISTIC;
1580 		ret = ipv6_saddr_preferred(score->addr_type) ||
1581 		      !(score->ifa->flags & avoid);
1582 		break;
1583 	    }
1584 #ifdef CONFIG_IPV6_MIP6
1585 	case IPV6_SADDR_RULE_HOA:
1586 	    {
1587 		/* Rule 4: Prefer home address */
1588 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1589 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1590 		break;
1591 	    }
1592 #endif
1593 	case IPV6_SADDR_RULE_OIF:
1594 		/* Rule 5: Prefer outgoing interface */
1595 		ret = (!dst->ifindex ||
1596 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1597 		break;
1598 	case IPV6_SADDR_RULE_LABEL:
1599 		/* Rule 6: Prefer matching label */
1600 		ret = ipv6_addr_label(net,
1601 				      &score->ifa->addr, score->addr_type,
1602 				      score->ifa->idev->dev->ifindex) == dst->label;
1603 		break;
1604 	case IPV6_SADDR_RULE_PRIVACY:
1605 	    {
1606 		/* Rule 7: Prefer public address
1607 		 * Note: prefer temporary address if use_tempaddr >= 2
1608 		 */
1609 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1610 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1611 				score->ifa->idev->cnf.use_tempaddr >= 2;
1612 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613 		break;
1614 	    }
1615 	case IPV6_SADDR_RULE_ORCHID:
1616 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1617 		 *	    non-ORCHID vs non-ORCHID
1618 		 */
1619 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1620 			ipv6_addr_orchid(dst->addr));
1621 		break;
1622 	case IPV6_SADDR_RULE_PREFIX:
1623 		/* Rule 8: Use longest matching prefix */
1624 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1625 		if (ret > score->ifa->prefix_len)
1626 			ret = score->ifa->prefix_len;
1627 		score->matchlen = ret;
1628 		break;
1629 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1630 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1631 		/* Optimistic addresses still have lower precedence than other
1632 		 * preferred addresses.
1633 		 */
1634 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635 		break;
1636 #endif
1637 	default:
1638 		ret = 0;
1639 	}
1640 
1641 	if (ret)
1642 		__set_bit(i, score->scorebits);
1643 	score->rule = i;
1644 out:
1645 	return ret;
1646 }
1647 
1648 static int __ipv6_dev_get_saddr(struct net *net,
1649 				struct ipv6_saddr_dst *dst,
1650 				struct inet6_dev *idev,
1651 				struct ipv6_saddr_score *scores,
1652 				int hiscore_idx)
1653 {
1654 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1655 
1656 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1657 		int i;
1658 
1659 		/*
1660 		 * - Tentative Address (RFC2462 section 5.4)
1661 		 *  - A tentative address is not considered
1662 		 *    "assigned to an interface" in the traditional
1663 		 *    sense, unless it is also flagged as optimistic.
1664 		 * - Candidate Source Address (section 4)
1665 		 *  - In any case, anycast addresses, multicast
1666 		 *    addresses, and the unspecified address MUST
1667 		 *    NOT be included in a candidate set.
1668 		 */
1669 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1670 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671 			continue;
1672 
1673 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1674 
1675 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1676 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1677 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1678 					    idev->dev->name);
1679 			continue;
1680 		}
1681 
1682 		score->rule = -1;
1683 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1684 
1685 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1686 			int minihiscore, miniscore;
1687 
1688 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1689 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1690 
1691 			if (minihiscore > miniscore) {
1692 				if (i == IPV6_SADDR_RULE_SCOPE &&
1693 				    score->scopedist > 0) {
1694 					/*
1695 					 * special case:
1696 					 * each remaining entry
1697 					 * has too small (not enough)
1698 					 * scope, because ifa entries
1699 					 * are sorted by their scope
1700 					 * values.
1701 					 */
1702 					goto out;
1703 				}
1704 				break;
1705 			} else if (minihiscore < miniscore) {
1706 				swap(hiscore, score);
1707 				hiscore_idx = 1 - hiscore_idx;
1708 
1709 				/* restore our iterator */
1710 				score->ifa = hiscore->ifa;
1711 
1712 				break;
1713 			}
1714 		}
1715 	}
1716 out:
1717 	return hiscore_idx;
1718 }
1719 
1720 static int ipv6_get_saddr_master(struct net *net,
1721 				 const struct net_device *dst_dev,
1722 				 const struct net_device *master,
1723 				 struct ipv6_saddr_dst *dst,
1724 				 struct ipv6_saddr_score *scores,
1725 				 int hiscore_idx)
1726 {
1727 	struct inet6_dev *idev;
1728 
1729 	idev = __in6_dev_get(dst_dev);
1730 	if (idev)
1731 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1732 						   scores, hiscore_idx);
1733 
1734 	idev = __in6_dev_get(master);
1735 	if (idev)
1736 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1737 						   scores, hiscore_idx);
1738 
1739 	return hiscore_idx;
1740 }
1741 
1742 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1743 		       const struct in6_addr *daddr, unsigned int prefs,
1744 		       struct in6_addr *saddr)
1745 {
1746 	struct ipv6_saddr_score scores[2], *hiscore;
1747 	struct ipv6_saddr_dst dst;
1748 	struct inet6_dev *idev;
1749 	struct net_device *dev;
1750 	int dst_type;
1751 	bool use_oif_addr = false;
1752 	int hiscore_idx = 0;
1753 	int ret = 0;
1754 
1755 	dst_type = __ipv6_addr_type(daddr);
1756 	dst.addr = daddr;
1757 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1758 	dst.scope = __ipv6_addr_src_scope(dst_type);
1759 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760 	dst.prefs = prefs;
1761 
1762 	scores[hiscore_idx].rule = -1;
1763 	scores[hiscore_idx].ifa = NULL;
1764 
1765 	rcu_read_lock();
1766 
1767 	/* Candidate Source Address (section 4)
1768 	 *  - multicast and link-local destination address,
1769 	 *    the set of candidate source address MUST only
1770 	 *    include addresses assigned to interfaces
1771 	 *    belonging to the same link as the outgoing
1772 	 *    interface.
1773 	 * (- For site-local destination addresses, the
1774 	 *    set of candidate source addresses MUST only
1775 	 *    include addresses assigned to interfaces
1776 	 *    belonging to the same site as the outgoing
1777 	 *    interface.)
1778 	 *  - "It is RECOMMENDED that the candidate source addresses
1779 	 *    be the set of unicast addresses assigned to the
1780 	 *    interface that will be used to send to the destination
1781 	 *    (the 'outgoing' interface)." (RFC 6724)
1782 	 */
1783 	if (dst_dev) {
1784 		idev = __in6_dev_get(dst_dev);
1785 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1786 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1787 		    (idev && idev->cnf.use_oif_addrs_only)) {
1788 			use_oif_addr = true;
1789 		}
1790 	}
1791 
1792 	if (use_oif_addr) {
1793 		if (idev)
1794 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1795 	} else {
1796 		const struct net_device *master;
1797 		int master_idx = 0;
1798 
1799 		/* if dst_dev exists and is enslaved to an L3 device, then
1800 		 * prefer addresses from dst_dev and then the master over
1801 		 * any other enslaved devices in the L3 domain.
1802 		 */
1803 		master = l3mdev_master_dev_rcu(dst_dev);
1804 		if (master) {
1805 			master_idx = master->ifindex;
1806 
1807 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1808 							    master, &dst,
1809 							    scores, hiscore_idx);
1810 
1811 			if (scores[hiscore_idx].ifa)
1812 				goto out;
1813 		}
1814 
1815 		for_each_netdev_rcu(net, dev) {
1816 			/* only consider addresses on devices in the
1817 			 * same L3 domain
1818 			 */
1819 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1820 				continue;
1821 			idev = __in6_dev_get(dev);
1822 			if (!idev)
1823 				continue;
1824 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1825 		}
1826 	}
1827 
1828 out:
1829 	hiscore = &scores[hiscore_idx];
1830 	if (!hiscore->ifa)
1831 		ret = -EADDRNOTAVAIL;
1832 	else
1833 		*saddr = hiscore->ifa->addr;
1834 
1835 	rcu_read_unlock();
1836 	return ret;
1837 }
1838 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1839 
1840 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841 			      u32 banned_flags)
1842 {
1843 	struct inet6_ifaddr *ifp;
1844 	int err = -EADDRNOTAVAIL;
1845 
1846 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1847 		if (ifp->scope > IFA_LINK)
1848 			break;
1849 		if (ifp->scope == IFA_LINK &&
1850 		    !(ifp->flags & banned_flags)) {
1851 			*addr = ifp->addr;
1852 			err = 0;
1853 			break;
1854 		}
1855 	}
1856 	return err;
1857 }
1858 
1859 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860 		    u32 banned_flags)
1861 {
1862 	struct inet6_dev *idev;
1863 	int err = -EADDRNOTAVAIL;
1864 
1865 	rcu_read_lock();
1866 	idev = __in6_dev_get(dev);
1867 	if (idev) {
1868 		read_lock_bh(&idev->lock);
1869 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1870 		read_unlock_bh(&idev->lock);
1871 	}
1872 	rcu_read_unlock();
1873 	return err;
1874 }
1875 
1876 static int ipv6_count_addresses(const struct inet6_dev *idev)
1877 {
1878 	const struct inet6_ifaddr *ifp;
1879 	int cnt = 0;
1880 
1881 	rcu_read_lock();
1882 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1883 		cnt++;
1884 	rcu_read_unlock();
1885 	return cnt;
1886 }
1887 
1888 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1889 		  const struct net_device *dev, int strict)
1890 {
1891 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1892 				       strict, IFA_F_TENTATIVE);
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr);
1895 
1896 /* device argument is used to find the L3 domain of interest. If
1897  * skip_dev_check is set, then the ifp device is not checked against
1898  * the passed in dev argument. So the 2 cases for addresses checks are:
1899  *   1. does the address exist in the L3 domain that dev is part of
1900  *      (skip_dev_check = true), or
1901  *
1902  *   2. does the address exist on the specific device
1903  *      (skip_dev_check = false)
1904  */
1905 static struct net_device *
1906 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1907 			  const struct net_device *dev, bool skip_dev_check,
1908 			  int strict, u32 banned_flags)
1909 {
1910 	unsigned int hash = inet6_addr_hash(net, addr);
1911 	struct net_device *l3mdev, *ndev;
1912 	struct inet6_ifaddr *ifp;
1913 	u32 ifp_flags;
1914 
1915 	rcu_read_lock();
1916 
1917 	l3mdev = l3mdev_master_dev_rcu(dev);
1918 	if (skip_dev_check)
1919 		dev = NULL;
1920 
1921 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1922 		ndev = ifp->idev->dev;
1923 
1924 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1925 			continue;
1926 
1927 		/* Decouple optimistic from tentative for evaluation here.
1928 		 * Ban optimistic addresses explicitly, when required.
1929 		 */
1930 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1931 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1932 			    : ifp->flags;
1933 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1934 		    !(ifp_flags&banned_flags) &&
1935 		    (!dev || ndev == dev ||
1936 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1937 			rcu_read_unlock();
1938 			return ndev;
1939 		}
1940 	}
1941 
1942 	rcu_read_unlock();
1943 	return NULL;
1944 }
1945 
1946 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1947 			    const struct net_device *dev, bool skip_dev_check,
1948 			    int strict, u32 banned_flags)
1949 {
1950 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1951 					 strict, banned_flags) ? 1 : 0;
1952 }
1953 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1954 
1955 
1956 /* Compares an address/prefix_len with addresses on device @dev.
1957  * If one is found it returns true.
1958  */
1959 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1960 	const unsigned int prefix_len, struct net_device *dev)
1961 {
1962 	const struct inet6_ifaddr *ifa;
1963 	const struct inet6_dev *idev;
1964 	bool ret = false;
1965 
1966 	rcu_read_lock();
1967 	idev = __in6_dev_get(dev);
1968 	if (idev) {
1969 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1970 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1971 			if (ret)
1972 				break;
1973 		}
1974 	}
1975 	rcu_read_unlock();
1976 
1977 	return ret;
1978 }
1979 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1980 
1981 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1982 {
1983 	const struct inet6_ifaddr *ifa;
1984 	const struct inet6_dev *idev;
1985 	int	onlink;
1986 
1987 	onlink = 0;
1988 	rcu_read_lock();
1989 	idev = __in6_dev_get(dev);
1990 	if (idev) {
1991 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1992 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1993 						   ifa->prefix_len);
1994 			if (onlink)
1995 				break;
1996 		}
1997 	}
1998 	rcu_read_unlock();
1999 	return onlink;
2000 }
2001 EXPORT_SYMBOL(ipv6_chk_prefix);
2002 
2003 /**
2004  * ipv6_dev_find - find the first device with a given source address.
2005  * @net: the net namespace
2006  * @addr: the source address
2007  * @dev: used to find the L3 domain of interest
2008  *
2009  * The caller should be protected by RCU, or RTNL.
2010  */
2011 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2012 				 struct net_device *dev)
2013 {
2014 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2015 					 IFA_F_TENTATIVE);
2016 }
2017 EXPORT_SYMBOL(ipv6_dev_find);
2018 
2019 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2020 				     struct net_device *dev, int strict)
2021 {
2022 	unsigned int hash = inet6_addr_hash(net, addr);
2023 	struct inet6_ifaddr *ifp, *result = NULL;
2024 
2025 	rcu_read_lock();
2026 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2027 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2028 			if (!dev || ifp->idev->dev == dev ||
2029 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2030 				result = ifp;
2031 				in6_ifa_hold(ifp);
2032 				break;
2033 			}
2034 		}
2035 	}
2036 	rcu_read_unlock();
2037 
2038 	return result;
2039 }
2040 
2041 /* Gets referenced address, destroys ifaddr */
2042 
2043 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2044 {
2045 	if (dad_failed)
2046 		ifp->flags |= IFA_F_DADFAILED;
2047 
2048 	if (ifp->flags&IFA_F_TEMPORARY) {
2049 		struct inet6_ifaddr *ifpub;
2050 		spin_lock_bh(&ifp->lock);
2051 		ifpub = ifp->ifpub;
2052 		if (ifpub) {
2053 			in6_ifa_hold(ifpub);
2054 			spin_unlock_bh(&ifp->lock);
2055 			ipv6_create_tempaddr(ifpub, true);
2056 			in6_ifa_put(ifpub);
2057 		} else {
2058 			spin_unlock_bh(&ifp->lock);
2059 		}
2060 		ipv6_del_addr(ifp);
2061 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2062 		spin_lock_bh(&ifp->lock);
2063 		addrconf_del_dad_work(ifp);
2064 		ifp->flags |= IFA_F_TENTATIVE;
2065 		if (dad_failed)
2066 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2067 		spin_unlock_bh(&ifp->lock);
2068 		if (dad_failed)
2069 			ipv6_ifa_notify(0, ifp);
2070 		in6_ifa_put(ifp);
2071 	} else {
2072 		ipv6_del_addr(ifp);
2073 	}
2074 }
2075 
2076 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2077 {
2078 	int err = -ENOENT;
2079 
2080 	spin_lock_bh(&ifp->lock);
2081 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2082 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2083 		err = 0;
2084 	}
2085 	spin_unlock_bh(&ifp->lock);
2086 
2087 	return err;
2088 }
2089 
2090 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2091 {
2092 	struct inet6_dev *idev = ifp->idev;
2093 	struct net *net = dev_net(idev->dev);
2094 
2095 	if (addrconf_dad_end(ifp)) {
2096 		in6_ifa_put(ifp);
2097 		return;
2098 	}
2099 
2100 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2101 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2102 
2103 	spin_lock_bh(&ifp->lock);
2104 
2105 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2106 		struct in6_addr new_addr;
2107 		struct inet6_ifaddr *ifp2;
2108 		int retries = ifp->stable_privacy_retry + 1;
2109 		struct ifa6_config cfg = {
2110 			.pfx = &new_addr,
2111 			.plen = ifp->prefix_len,
2112 			.ifa_flags = ifp->flags,
2113 			.valid_lft = ifp->valid_lft,
2114 			.preferred_lft = ifp->prefered_lft,
2115 			.scope = ifp->scope,
2116 		};
2117 
2118 		if (retries > net->ipv6.sysctl.idgen_retries) {
2119 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2120 					     ifp->idev->dev->name);
2121 			goto errdad;
2122 		}
2123 
2124 		new_addr = ifp->addr;
2125 		if (ipv6_generate_stable_address(&new_addr, retries,
2126 						 idev))
2127 			goto errdad;
2128 
2129 		spin_unlock_bh(&ifp->lock);
2130 
2131 		if (idev->cnf.max_addresses &&
2132 		    ipv6_count_addresses(idev) >=
2133 		    idev->cnf.max_addresses)
2134 			goto lock_errdad;
2135 
2136 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2137 				     ifp->idev->dev->name);
2138 
2139 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2140 		if (IS_ERR(ifp2))
2141 			goto lock_errdad;
2142 
2143 		spin_lock_bh(&ifp2->lock);
2144 		ifp2->stable_privacy_retry = retries;
2145 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2146 		spin_unlock_bh(&ifp2->lock);
2147 
2148 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2149 		in6_ifa_put(ifp2);
2150 lock_errdad:
2151 		spin_lock_bh(&ifp->lock);
2152 	}
2153 
2154 errdad:
2155 	/* transition from _POSTDAD to _ERRDAD */
2156 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2157 	spin_unlock_bh(&ifp->lock);
2158 
2159 	addrconf_mod_dad_work(ifp, 0);
2160 	in6_ifa_put(ifp);
2161 }
2162 
2163 /* Join to solicited addr multicast group.
2164  * caller must hold RTNL */
2165 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2166 {
2167 	struct in6_addr maddr;
2168 
2169 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2170 		return;
2171 
2172 	addrconf_addr_solict_mult(addr, &maddr);
2173 	ipv6_dev_mc_inc(dev, &maddr);
2174 }
2175 
2176 /* caller must hold RTNL */
2177 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2178 {
2179 	struct in6_addr maddr;
2180 
2181 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2182 		return;
2183 
2184 	addrconf_addr_solict_mult(addr, &maddr);
2185 	__ipv6_dev_mc_dec(idev, &maddr);
2186 }
2187 
2188 /* caller must hold RTNL */
2189 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2190 {
2191 	struct in6_addr addr;
2192 
2193 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2194 		return;
2195 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2196 	if (ipv6_addr_any(&addr))
2197 		return;
2198 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2199 }
2200 
2201 /* caller must hold RTNL */
2202 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2203 {
2204 	struct in6_addr addr;
2205 
2206 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2207 		return;
2208 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2209 	if (ipv6_addr_any(&addr))
2210 		return;
2211 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2212 }
2213 
2214 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2215 {
2216 	switch (dev->addr_len) {
2217 	case ETH_ALEN:
2218 		memcpy(eui, dev->dev_addr, 3);
2219 		eui[3] = 0xFF;
2220 		eui[4] = 0xFE;
2221 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2222 		break;
2223 	case EUI64_ADDR_LEN:
2224 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2225 		eui[0] ^= 2;
2226 		break;
2227 	default:
2228 		return -1;
2229 	}
2230 
2231 	return 0;
2232 }
2233 
2234 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2235 {
2236 	const union fwnet_hwaddr *ha;
2237 
2238 	if (dev->addr_len != FWNET_ALEN)
2239 		return -1;
2240 
2241 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2242 
2243 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2244 	eui[0] ^= 2;
2245 	return 0;
2246 }
2247 
2248 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2249 {
2250 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2251 	if (dev->addr_len != ARCNET_ALEN)
2252 		return -1;
2253 	memset(eui, 0, 7);
2254 	eui[7] = *(u8 *)dev->dev_addr;
2255 	return 0;
2256 }
2257 
2258 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2259 {
2260 	if (dev->addr_len != INFINIBAND_ALEN)
2261 		return -1;
2262 	memcpy(eui, dev->dev_addr + 12, 8);
2263 	eui[0] |= 2;
2264 	return 0;
2265 }
2266 
2267 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2268 {
2269 	if (addr == 0)
2270 		return -1;
2271 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2272 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2273 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2274 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2275 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2276 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2277 	eui[1] = 0;
2278 	eui[2] = 0x5E;
2279 	eui[3] = 0xFE;
2280 	memcpy(eui + 4, &addr, 4);
2281 	return 0;
2282 }
2283 
2284 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2285 {
2286 	if (dev->priv_flags & IFF_ISATAP)
2287 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2288 	return -1;
2289 }
2290 
2291 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2292 {
2293 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2294 }
2295 
2296 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2297 {
2298 	memcpy(eui, dev->perm_addr, 3);
2299 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2300 	eui[3] = 0xFF;
2301 	eui[4] = 0xFE;
2302 	eui[0] ^= 2;
2303 	return 0;
2304 }
2305 
2306 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2307 {
2308 	switch (dev->type) {
2309 	case ARPHRD_ETHER:
2310 	case ARPHRD_FDDI:
2311 		return addrconf_ifid_eui48(eui, dev);
2312 	case ARPHRD_ARCNET:
2313 		return addrconf_ifid_arcnet(eui, dev);
2314 	case ARPHRD_INFINIBAND:
2315 		return addrconf_ifid_infiniband(eui, dev);
2316 	case ARPHRD_SIT:
2317 		return addrconf_ifid_sit(eui, dev);
2318 	case ARPHRD_IPGRE:
2319 	case ARPHRD_TUNNEL:
2320 		return addrconf_ifid_gre(eui, dev);
2321 	case ARPHRD_6LOWPAN:
2322 		return addrconf_ifid_6lowpan(eui, dev);
2323 	case ARPHRD_IEEE1394:
2324 		return addrconf_ifid_ieee1394(eui, dev);
2325 	case ARPHRD_TUNNEL6:
2326 	case ARPHRD_IP6GRE:
2327 	case ARPHRD_RAWIP:
2328 		return addrconf_ifid_ip6tnl(eui, dev);
2329 	}
2330 	return -1;
2331 }
2332 
2333 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2334 {
2335 	int err = -1;
2336 	struct inet6_ifaddr *ifp;
2337 
2338 	read_lock_bh(&idev->lock);
2339 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2340 		if (ifp->scope > IFA_LINK)
2341 			break;
2342 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2343 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2344 			err = 0;
2345 			break;
2346 		}
2347 	}
2348 	read_unlock_bh(&idev->lock);
2349 	return err;
2350 }
2351 
2352 /* Generation of a randomized Interface Identifier
2353  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2354  */
2355 
2356 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2357 {
2358 regen:
2359 	get_random_bytes(&addr->s6_addr[8], 8);
2360 
2361 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2362 	 * check if generated address is not inappropriate:
2363 	 *
2364 	 * - Reserved IPv6 Interface Identifiers
2365 	 * - XXX: already assigned to an address on the device
2366 	 */
2367 
2368 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2369 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2370 		goto regen;
2371 
2372 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2373 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2374 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2375 	 */
2376 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2377 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2378 		goto regen;
2379 
2380 	/* Reserved subnet anycast addresses */
2381 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2382 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2383 		goto regen;
2384 }
2385 
2386 /*
2387  *	Add prefix route.
2388  */
2389 
2390 static void
2391 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2392 		      struct net_device *dev, unsigned long expires,
2393 		      u32 flags, gfp_t gfp_flags)
2394 {
2395 	struct fib6_config cfg = {
2396 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2397 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2398 		.fc_ifindex = dev->ifindex,
2399 		.fc_expires = expires,
2400 		.fc_dst_len = plen,
2401 		.fc_flags = RTF_UP | flags,
2402 		.fc_nlinfo.nl_net = dev_net(dev),
2403 		.fc_protocol = RTPROT_KERNEL,
2404 		.fc_type = RTN_UNICAST,
2405 	};
2406 
2407 	cfg.fc_dst = *pfx;
2408 
2409 	/* Prevent useless cloning on PtP SIT.
2410 	   This thing is done here expecting that the whole
2411 	   class of non-broadcast devices need not cloning.
2412 	 */
2413 #if IS_ENABLED(CONFIG_IPV6_SIT)
2414 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2415 		cfg.fc_flags |= RTF_NONEXTHOP;
2416 #endif
2417 
2418 	ip6_route_add(&cfg, gfp_flags, NULL);
2419 }
2420 
2421 
2422 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2423 						  int plen,
2424 						  const struct net_device *dev,
2425 						  u32 flags, u32 noflags,
2426 						  bool no_gw)
2427 {
2428 	struct fib6_node *fn;
2429 	struct fib6_info *rt = NULL;
2430 	struct fib6_table *table;
2431 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2432 
2433 	table = fib6_get_table(dev_net(dev), tb_id);
2434 	if (!table)
2435 		return NULL;
2436 
2437 	rcu_read_lock();
2438 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2439 	if (!fn)
2440 		goto out;
2441 
2442 	for_each_fib6_node_rt_rcu(fn) {
2443 		/* prefix routes only use builtin fib6_nh */
2444 		if (rt->nh)
2445 			continue;
2446 
2447 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2448 			continue;
2449 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2450 			continue;
2451 		if ((rt->fib6_flags & flags) != flags)
2452 			continue;
2453 		if ((rt->fib6_flags & noflags) != 0)
2454 			continue;
2455 		if (!fib6_info_hold_safe(rt))
2456 			continue;
2457 		break;
2458 	}
2459 out:
2460 	rcu_read_unlock();
2461 	return rt;
2462 }
2463 
2464 
2465 /* Create "default" multicast route to the interface */
2466 
2467 static void addrconf_add_mroute(struct net_device *dev)
2468 {
2469 	struct fib6_config cfg = {
2470 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2471 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2472 		.fc_ifindex = dev->ifindex,
2473 		.fc_dst_len = 8,
2474 		.fc_flags = RTF_UP,
2475 		.fc_type = RTN_MULTICAST,
2476 		.fc_nlinfo.nl_net = dev_net(dev),
2477 		.fc_protocol = RTPROT_KERNEL,
2478 	};
2479 
2480 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2481 
2482 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2483 }
2484 
2485 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2486 {
2487 	struct inet6_dev *idev;
2488 
2489 	ASSERT_RTNL();
2490 
2491 	idev = ipv6_find_idev(dev);
2492 	if (IS_ERR(idev))
2493 		return idev;
2494 
2495 	if (idev->cnf.disable_ipv6)
2496 		return ERR_PTR(-EACCES);
2497 
2498 	/* Add default multicast route */
2499 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2500 		addrconf_add_mroute(dev);
2501 
2502 	return idev;
2503 }
2504 
2505 static void manage_tempaddrs(struct inet6_dev *idev,
2506 			     struct inet6_ifaddr *ifp,
2507 			     __u32 valid_lft, __u32 prefered_lft,
2508 			     bool create, unsigned long now)
2509 {
2510 	u32 flags;
2511 	struct inet6_ifaddr *ift;
2512 
2513 	read_lock_bh(&idev->lock);
2514 	/* update all temporary addresses in the list */
2515 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2516 		int age, max_valid, max_prefered;
2517 
2518 		if (ifp != ift->ifpub)
2519 			continue;
2520 
2521 		/* RFC 4941 section 3.3:
2522 		 * If a received option will extend the lifetime of a public
2523 		 * address, the lifetimes of temporary addresses should
2524 		 * be extended, subject to the overall constraint that no
2525 		 * temporary addresses should ever remain "valid" or "preferred"
2526 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2527 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2528 		 */
2529 		age = (now - ift->cstamp) / HZ;
2530 		max_valid = idev->cnf.temp_valid_lft - age;
2531 		if (max_valid < 0)
2532 			max_valid = 0;
2533 
2534 		max_prefered = idev->cnf.temp_prefered_lft -
2535 			       idev->desync_factor - age;
2536 		if (max_prefered < 0)
2537 			max_prefered = 0;
2538 
2539 		if (valid_lft > max_valid)
2540 			valid_lft = max_valid;
2541 
2542 		if (prefered_lft > max_prefered)
2543 			prefered_lft = max_prefered;
2544 
2545 		spin_lock(&ift->lock);
2546 		flags = ift->flags;
2547 		ift->valid_lft = valid_lft;
2548 		ift->prefered_lft = prefered_lft;
2549 		ift->tstamp = now;
2550 		if (prefered_lft > 0)
2551 			ift->flags &= ~IFA_F_DEPRECATED;
2552 
2553 		spin_unlock(&ift->lock);
2554 		if (!(flags&IFA_F_TENTATIVE))
2555 			ipv6_ifa_notify(0, ift);
2556 	}
2557 
2558 	if ((create || list_empty(&idev->tempaddr_list)) &&
2559 	    idev->cnf.use_tempaddr > 0) {
2560 		/* When a new public address is created as described
2561 		 * in [ADDRCONF], also create a new temporary address.
2562 		 * Also create a temporary address if it's enabled but
2563 		 * no temporary address currently exists.
2564 		 */
2565 		read_unlock_bh(&idev->lock);
2566 		ipv6_create_tempaddr(ifp, false);
2567 	} else {
2568 		read_unlock_bh(&idev->lock);
2569 	}
2570 }
2571 
2572 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2573 {
2574 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2575 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2576 }
2577 
2578 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2579 				 const struct prefix_info *pinfo,
2580 				 struct inet6_dev *in6_dev,
2581 				 const struct in6_addr *addr, int addr_type,
2582 				 u32 addr_flags, bool sllao, bool tokenized,
2583 				 __u32 valid_lft, u32 prefered_lft)
2584 {
2585 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2586 	int create = 0, update_lft = 0;
2587 
2588 	if (!ifp && valid_lft) {
2589 		int max_addresses = in6_dev->cnf.max_addresses;
2590 		struct ifa6_config cfg = {
2591 			.pfx = addr,
2592 			.plen = pinfo->prefix_len,
2593 			.ifa_flags = addr_flags,
2594 			.valid_lft = valid_lft,
2595 			.preferred_lft = prefered_lft,
2596 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2597 			.ifa_proto = IFAPROT_KERNEL_RA
2598 		};
2599 
2600 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2601 		if ((net->ipv6.devconf_all->optimistic_dad ||
2602 		     in6_dev->cnf.optimistic_dad) &&
2603 		    !net->ipv6.devconf_all->forwarding && sllao)
2604 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2605 #endif
2606 
2607 		/* Do not allow to create too much of autoconfigured
2608 		 * addresses; this would be too easy way to crash kernel.
2609 		 */
2610 		if (!max_addresses ||
2611 		    ipv6_count_addresses(in6_dev) < max_addresses)
2612 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2613 
2614 		if (IS_ERR_OR_NULL(ifp))
2615 			return -1;
2616 
2617 		create = 1;
2618 		spin_lock_bh(&ifp->lock);
2619 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2620 		ifp->cstamp = jiffies;
2621 		ifp->tokenized = tokenized;
2622 		spin_unlock_bh(&ifp->lock);
2623 		addrconf_dad_start(ifp);
2624 	}
2625 
2626 	if (ifp) {
2627 		u32 flags;
2628 		unsigned long now;
2629 		u32 stored_lft;
2630 
2631 		/* update lifetime (RFC2462 5.5.3 e) */
2632 		spin_lock_bh(&ifp->lock);
2633 		now = jiffies;
2634 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2635 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2636 		else
2637 			stored_lft = 0;
2638 		if (!create && stored_lft) {
2639 			const u32 minimum_lft = min_t(u32,
2640 				stored_lft, MIN_VALID_LIFETIME);
2641 			valid_lft = max(valid_lft, minimum_lft);
2642 
2643 			/* RFC4862 Section 5.5.3e:
2644 			 * "Note that the preferred lifetime of the
2645 			 *  corresponding address is always reset to
2646 			 *  the Preferred Lifetime in the received
2647 			 *  Prefix Information option, regardless of
2648 			 *  whether the valid lifetime is also reset or
2649 			 *  ignored."
2650 			 *
2651 			 * So we should always update prefered_lft here.
2652 			 */
2653 			update_lft = 1;
2654 		}
2655 
2656 		if (update_lft) {
2657 			ifp->valid_lft = valid_lft;
2658 			ifp->prefered_lft = prefered_lft;
2659 			ifp->tstamp = now;
2660 			flags = ifp->flags;
2661 			ifp->flags &= ~IFA_F_DEPRECATED;
2662 			spin_unlock_bh(&ifp->lock);
2663 
2664 			if (!(flags&IFA_F_TENTATIVE))
2665 				ipv6_ifa_notify(0, ifp);
2666 		} else
2667 			spin_unlock_bh(&ifp->lock);
2668 
2669 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2670 				 create, now);
2671 
2672 		in6_ifa_put(ifp);
2673 		addrconf_verify(net);
2674 	}
2675 
2676 	return 0;
2677 }
2678 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2679 
2680 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2681 {
2682 	struct prefix_info *pinfo;
2683 	__u32 valid_lft;
2684 	__u32 prefered_lft;
2685 	int addr_type, err;
2686 	u32 addr_flags = 0;
2687 	struct inet6_dev *in6_dev;
2688 	struct net *net = dev_net(dev);
2689 
2690 	pinfo = (struct prefix_info *) opt;
2691 
2692 	if (len < sizeof(struct prefix_info)) {
2693 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2694 		return;
2695 	}
2696 
2697 	/*
2698 	 *	Validation checks ([ADDRCONF], page 19)
2699 	 */
2700 
2701 	addr_type = ipv6_addr_type(&pinfo->prefix);
2702 
2703 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2704 		return;
2705 
2706 	valid_lft = ntohl(pinfo->valid);
2707 	prefered_lft = ntohl(pinfo->prefered);
2708 
2709 	if (prefered_lft > valid_lft) {
2710 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2711 		return;
2712 	}
2713 
2714 	in6_dev = in6_dev_get(dev);
2715 
2716 	if (!in6_dev) {
2717 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2718 				    dev->name);
2719 		return;
2720 	}
2721 
2722 	/*
2723 	 *	Two things going on here:
2724 	 *	1) Add routes for on-link prefixes
2725 	 *	2) Configure prefixes with the auto flag set
2726 	 */
2727 
2728 	if (pinfo->onlink) {
2729 		struct fib6_info *rt;
2730 		unsigned long rt_expires;
2731 
2732 		/* Avoid arithmetic overflow. Really, we could
2733 		 * save rt_expires in seconds, likely valid_lft,
2734 		 * but it would require division in fib gc, that it
2735 		 * not good.
2736 		 */
2737 		if (HZ > USER_HZ)
2738 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2739 		else
2740 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2741 
2742 		if (addrconf_finite_timeout(rt_expires))
2743 			rt_expires *= HZ;
2744 
2745 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2746 					       pinfo->prefix_len,
2747 					       dev,
2748 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2749 					       RTF_DEFAULT, true);
2750 
2751 		if (rt) {
2752 			/* Autoconf prefix route */
2753 			if (valid_lft == 0) {
2754 				ip6_del_rt(net, rt, false);
2755 				rt = NULL;
2756 			} else if (addrconf_finite_timeout(rt_expires)) {
2757 				/* not infinity */
2758 				fib6_set_expires(rt, jiffies + rt_expires);
2759 			} else {
2760 				fib6_clean_expires(rt);
2761 			}
2762 		} else if (valid_lft) {
2763 			clock_t expires = 0;
2764 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2765 			if (addrconf_finite_timeout(rt_expires)) {
2766 				/* not infinity */
2767 				flags |= RTF_EXPIRES;
2768 				expires = jiffies_to_clock_t(rt_expires);
2769 			}
2770 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2771 					      0, dev, expires, flags,
2772 					      GFP_ATOMIC);
2773 		}
2774 		fib6_info_release(rt);
2775 	}
2776 
2777 	/* Try to figure out our local address for this prefix */
2778 
2779 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2780 		struct in6_addr addr;
2781 		bool tokenized = false, dev_addr_generated = false;
2782 
2783 		if (pinfo->prefix_len == 64) {
2784 			memcpy(&addr, &pinfo->prefix, 8);
2785 
2786 			if (!ipv6_addr_any(&in6_dev->token)) {
2787 				read_lock_bh(&in6_dev->lock);
2788 				memcpy(addr.s6_addr + 8,
2789 				       in6_dev->token.s6_addr + 8, 8);
2790 				read_unlock_bh(&in6_dev->lock);
2791 				tokenized = true;
2792 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2793 				   !ipv6_generate_stable_address(&addr, 0,
2794 								 in6_dev)) {
2795 				addr_flags |= IFA_F_STABLE_PRIVACY;
2796 				goto ok;
2797 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2798 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2799 				goto put;
2800 			} else {
2801 				dev_addr_generated = true;
2802 			}
2803 			goto ok;
2804 		}
2805 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2806 				    pinfo->prefix_len);
2807 		goto put;
2808 
2809 ok:
2810 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2811 						   &addr, addr_type,
2812 						   addr_flags, sllao,
2813 						   tokenized, valid_lft,
2814 						   prefered_lft);
2815 		if (err)
2816 			goto put;
2817 
2818 		/* Ignore error case here because previous prefix add addr was
2819 		 * successful which will be notified.
2820 		 */
2821 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2822 					      addr_type, addr_flags, sllao,
2823 					      tokenized, valid_lft,
2824 					      prefered_lft,
2825 					      dev_addr_generated);
2826 	}
2827 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2828 put:
2829 	in6_dev_put(in6_dev);
2830 }
2831 
2832 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2833 		struct in6_ifreq *ireq)
2834 {
2835 	struct ip_tunnel_parm p = { };
2836 	int err;
2837 
2838 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2839 		return -EADDRNOTAVAIL;
2840 
2841 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2842 	p.iph.version = 4;
2843 	p.iph.ihl = 5;
2844 	p.iph.protocol = IPPROTO_IPV6;
2845 	p.iph.ttl = 64;
2846 
2847 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2848 		return -EOPNOTSUPP;
2849 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2850 	if (err)
2851 		return err;
2852 
2853 	dev = __dev_get_by_name(net, p.name);
2854 	if (!dev)
2855 		return -ENOBUFS;
2856 	return dev_open(dev, NULL);
2857 }
2858 
2859 /*
2860  *	Set destination address.
2861  *	Special case for SIT interfaces where we create a new "virtual"
2862  *	device.
2863  */
2864 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2865 {
2866 	struct net_device *dev;
2867 	struct in6_ifreq ireq;
2868 	int err = -ENODEV;
2869 
2870 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2871 		return -ENODEV;
2872 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2873 		return -EFAULT;
2874 
2875 	rtnl_lock();
2876 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2877 	if (dev && dev->type == ARPHRD_SIT)
2878 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2879 	rtnl_unlock();
2880 	return err;
2881 }
2882 
2883 static int ipv6_mc_config(struct sock *sk, bool join,
2884 			  const struct in6_addr *addr, int ifindex)
2885 {
2886 	int ret;
2887 
2888 	ASSERT_RTNL();
2889 
2890 	lock_sock(sk);
2891 	if (join)
2892 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2893 	else
2894 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2895 	release_sock(sk);
2896 
2897 	return ret;
2898 }
2899 
2900 /*
2901  *	Manual configuration of address on an interface
2902  */
2903 static int inet6_addr_add(struct net *net, int ifindex,
2904 			  struct ifa6_config *cfg,
2905 			  struct netlink_ext_ack *extack)
2906 {
2907 	struct inet6_ifaddr *ifp;
2908 	struct inet6_dev *idev;
2909 	struct net_device *dev;
2910 	unsigned long timeout;
2911 	clock_t expires;
2912 	u32 flags;
2913 
2914 	ASSERT_RTNL();
2915 
2916 	if (cfg->plen > 128)
2917 		return -EINVAL;
2918 
2919 	/* check the lifetime */
2920 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2921 		return -EINVAL;
2922 
2923 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2924 		return -EINVAL;
2925 
2926 	dev = __dev_get_by_index(net, ifindex);
2927 	if (!dev)
2928 		return -ENODEV;
2929 
2930 	idev = addrconf_add_dev(dev);
2931 	if (IS_ERR(idev))
2932 		return PTR_ERR(idev);
2933 
2934 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2935 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2936 					 true, cfg->pfx, ifindex);
2937 
2938 		if (ret < 0)
2939 			return ret;
2940 	}
2941 
2942 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2943 
2944 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2945 	if (addrconf_finite_timeout(timeout)) {
2946 		expires = jiffies_to_clock_t(timeout * HZ);
2947 		cfg->valid_lft = timeout;
2948 		flags = RTF_EXPIRES;
2949 	} else {
2950 		expires = 0;
2951 		flags = 0;
2952 		cfg->ifa_flags |= IFA_F_PERMANENT;
2953 	}
2954 
2955 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2956 	if (addrconf_finite_timeout(timeout)) {
2957 		if (timeout == 0)
2958 			cfg->ifa_flags |= IFA_F_DEPRECATED;
2959 		cfg->preferred_lft = timeout;
2960 	}
2961 
2962 	ifp = ipv6_add_addr(idev, cfg, true, extack);
2963 	if (!IS_ERR(ifp)) {
2964 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2965 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2966 					      ifp->rt_priority, dev, expires,
2967 					      flags, GFP_KERNEL);
2968 		}
2969 
2970 		/* Send a netlink notification if DAD is enabled and
2971 		 * optimistic flag is not set
2972 		 */
2973 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2974 			ipv6_ifa_notify(0, ifp);
2975 		/*
2976 		 * Note that section 3.1 of RFC 4429 indicates
2977 		 * that the Optimistic flag should not be set for
2978 		 * manually configured addresses
2979 		 */
2980 		addrconf_dad_start(ifp);
2981 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2982 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
2983 					 cfg->preferred_lft, true, jiffies);
2984 		in6_ifa_put(ifp);
2985 		addrconf_verify_rtnl(net);
2986 		return 0;
2987 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2988 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2989 			       cfg->pfx, ifindex);
2990 	}
2991 
2992 	return PTR_ERR(ifp);
2993 }
2994 
2995 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2996 			  const struct in6_addr *pfx, unsigned int plen)
2997 {
2998 	struct inet6_ifaddr *ifp;
2999 	struct inet6_dev *idev;
3000 	struct net_device *dev;
3001 
3002 	if (plen > 128)
3003 		return -EINVAL;
3004 
3005 	dev = __dev_get_by_index(net, ifindex);
3006 	if (!dev)
3007 		return -ENODEV;
3008 
3009 	idev = __in6_dev_get(dev);
3010 	if (!idev)
3011 		return -ENXIO;
3012 
3013 	read_lock_bh(&idev->lock);
3014 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3015 		if (ifp->prefix_len == plen &&
3016 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3017 			in6_ifa_hold(ifp);
3018 			read_unlock_bh(&idev->lock);
3019 
3020 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3021 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
3022 				manage_tempaddrs(idev, ifp, 0, 0, false,
3023 						 jiffies);
3024 			ipv6_del_addr(ifp);
3025 			addrconf_verify_rtnl(net);
3026 			if (ipv6_addr_is_multicast(pfx)) {
3027 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3028 					       false, pfx, dev->ifindex);
3029 			}
3030 			return 0;
3031 		}
3032 	}
3033 	read_unlock_bh(&idev->lock);
3034 	return -EADDRNOTAVAIL;
3035 }
3036 
3037 
3038 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3039 {
3040 	struct ifa6_config cfg = {
3041 		.ifa_flags = IFA_F_PERMANENT,
3042 		.preferred_lft = INFINITY_LIFE_TIME,
3043 		.valid_lft = INFINITY_LIFE_TIME,
3044 	};
3045 	struct in6_ifreq ireq;
3046 	int err;
3047 
3048 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3049 		return -EPERM;
3050 
3051 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3052 		return -EFAULT;
3053 
3054 	cfg.pfx = &ireq.ifr6_addr;
3055 	cfg.plen = ireq.ifr6_prefixlen;
3056 
3057 	rtnl_lock();
3058 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3059 	rtnl_unlock();
3060 	return err;
3061 }
3062 
3063 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3064 {
3065 	struct in6_ifreq ireq;
3066 	int err;
3067 
3068 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3069 		return -EPERM;
3070 
3071 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3072 		return -EFAULT;
3073 
3074 	rtnl_lock();
3075 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3076 			     ireq.ifr6_prefixlen);
3077 	rtnl_unlock();
3078 	return err;
3079 }
3080 
3081 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3082 		     int plen, int scope, u8 proto)
3083 {
3084 	struct inet6_ifaddr *ifp;
3085 	struct ifa6_config cfg = {
3086 		.pfx = addr,
3087 		.plen = plen,
3088 		.ifa_flags = IFA_F_PERMANENT,
3089 		.valid_lft = INFINITY_LIFE_TIME,
3090 		.preferred_lft = INFINITY_LIFE_TIME,
3091 		.scope = scope,
3092 		.ifa_proto = proto
3093 	};
3094 
3095 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3096 	if (!IS_ERR(ifp)) {
3097 		spin_lock_bh(&ifp->lock);
3098 		ifp->flags &= ~IFA_F_TENTATIVE;
3099 		spin_unlock_bh(&ifp->lock);
3100 		rt_genid_bump_ipv6(dev_net(idev->dev));
3101 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3102 		in6_ifa_put(ifp);
3103 	}
3104 }
3105 
3106 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3107 static void add_v4_addrs(struct inet6_dev *idev)
3108 {
3109 	struct in6_addr addr;
3110 	struct net_device *dev;
3111 	struct net *net = dev_net(idev->dev);
3112 	int scope, plen, offset = 0;
3113 	u32 pflags = 0;
3114 
3115 	ASSERT_RTNL();
3116 
3117 	memset(&addr, 0, sizeof(struct in6_addr));
3118 	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3119 	if (idev->dev->addr_len == sizeof(struct in6_addr))
3120 		offset = sizeof(struct in6_addr) - 4;
3121 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3122 
3123 	if (idev->dev->flags&IFF_POINTOPOINT) {
3124 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3125 			return;
3126 
3127 		addr.s6_addr32[0] = htonl(0xfe800000);
3128 		scope = IFA_LINK;
3129 		plen = 64;
3130 	} else {
3131 		scope = IPV6_ADDR_COMPATv4;
3132 		plen = 96;
3133 		pflags |= RTF_NONEXTHOP;
3134 	}
3135 
3136 	if (addr.s6_addr32[3]) {
3137 		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3138 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3139 				      GFP_KERNEL);
3140 		return;
3141 	}
3142 
3143 	for_each_netdev(net, dev) {
3144 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3145 		if (in_dev && (dev->flags & IFF_UP)) {
3146 			struct in_ifaddr *ifa;
3147 			int flag = scope;
3148 
3149 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3150 				addr.s6_addr32[3] = ifa->ifa_local;
3151 
3152 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3153 					continue;
3154 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3155 					if (idev->dev->flags&IFF_POINTOPOINT)
3156 						continue;
3157 					flag |= IFA_HOST;
3158 				}
3159 
3160 				add_addr(idev, &addr, plen, flag,
3161 					 IFAPROT_UNSPEC);
3162 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3163 						      0, pflags, GFP_KERNEL);
3164 			}
3165 		}
3166 	}
3167 }
3168 #endif
3169 
3170 static void init_loopback(struct net_device *dev)
3171 {
3172 	struct inet6_dev  *idev;
3173 
3174 	/* ::1 */
3175 
3176 	ASSERT_RTNL();
3177 
3178 	idev = ipv6_find_idev(dev);
3179 	if (IS_ERR(idev)) {
3180 		pr_debug("%s: add_dev failed\n", __func__);
3181 		return;
3182 	}
3183 
3184 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3185 }
3186 
3187 void addrconf_add_linklocal(struct inet6_dev *idev,
3188 			    const struct in6_addr *addr, u32 flags)
3189 {
3190 	struct ifa6_config cfg = {
3191 		.pfx = addr,
3192 		.plen = 64,
3193 		.ifa_flags = flags | IFA_F_PERMANENT,
3194 		.valid_lft = INFINITY_LIFE_TIME,
3195 		.preferred_lft = INFINITY_LIFE_TIME,
3196 		.scope = IFA_LINK,
3197 		.ifa_proto = IFAPROT_KERNEL_LL
3198 	};
3199 	struct inet6_ifaddr *ifp;
3200 
3201 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3202 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3203 	     idev->cnf.optimistic_dad) &&
3204 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3205 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3206 #endif
3207 
3208 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3209 	if (!IS_ERR(ifp)) {
3210 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3211 				      0, 0, GFP_ATOMIC);
3212 		addrconf_dad_start(ifp);
3213 		in6_ifa_put(ifp);
3214 	}
3215 }
3216 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3217 
3218 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3219 {
3220 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3221 		return true;
3222 
3223 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3224 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3225 		return true;
3226 
3227 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3228 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3229 		return true;
3230 
3231 	return false;
3232 }
3233 
3234 static int ipv6_generate_stable_address(struct in6_addr *address,
3235 					u8 dad_count,
3236 					const struct inet6_dev *idev)
3237 {
3238 	static DEFINE_SPINLOCK(lock);
3239 	static __u32 digest[SHA1_DIGEST_WORDS];
3240 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3241 
3242 	static union {
3243 		char __data[SHA1_BLOCK_SIZE];
3244 		struct {
3245 			struct in6_addr secret;
3246 			__be32 prefix[2];
3247 			unsigned char hwaddr[MAX_ADDR_LEN];
3248 			u8 dad_count;
3249 		} __packed;
3250 	} data;
3251 
3252 	struct in6_addr secret;
3253 	struct in6_addr temp;
3254 	struct net *net = dev_net(idev->dev);
3255 
3256 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3257 
3258 	if (idev->cnf.stable_secret.initialized)
3259 		secret = idev->cnf.stable_secret.secret;
3260 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3261 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3262 	else
3263 		return -1;
3264 
3265 retry:
3266 	spin_lock_bh(&lock);
3267 
3268 	sha1_init(digest);
3269 	memset(&data, 0, sizeof(data));
3270 	memset(workspace, 0, sizeof(workspace));
3271 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3272 	data.prefix[0] = address->s6_addr32[0];
3273 	data.prefix[1] = address->s6_addr32[1];
3274 	data.secret = secret;
3275 	data.dad_count = dad_count;
3276 
3277 	sha1_transform(digest, data.__data, workspace);
3278 
3279 	temp = *address;
3280 	temp.s6_addr32[2] = (__force __be32)digest[0];
3281 	temp.s6_addr32[3] = (__force __be32)digest[1];
3282 
3283 	spin_unlock_bh(&lock);
3284 
3285 	if (ipv6_reserved_interfaceid(temp)) {
3286 		dad_count++;
3287 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3288 			return -1;
3289 		goto retry;
3290 	}
3291 
3292 	*address = temp;
3293 	return 0;
3294 }
3295 
3296 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3297 {
3298 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3299 
3300 	if (s->initialized)
3301 		return;
3302 	s = &idev->cnf.stable_secret;
3303 	get_random_bytes(&s->secret, sizeof(s->secret));
3304 	s->initialized = true;
3305 }
3306 
3307 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3308 {
3309 	struct in6_addr addr;
3310 
3311 	/* no link local addresses on L3 master devices */
3312 	if (netif_is_l3_master(idev->dev))
3313 		return;
3314 
3315 	/* no link local addresses on devices flagged as slaves */
3316 	if (idev->dev->flags & IFF_SLAVE)
3317 		return;
3318 
3319 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3320 
3321 	switch (idev->cnf.addr_gen_mode) {
3322 	case IN6_ADDR_GEN_MODE_RANDOM:
3323 		ipv6_gen_mode_random_init(idev);
3324 		fallthrough;
3325 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3326 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3327 			addrconf_add_linklocal(idev, &addr,
3328 					       IFA_F_STABLE_PRIVACY);
3329 		else if (prefix_route)
3330 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3331 					      0, 0, GFP_KERNEL);
3332 		break;
3333 	case IN6_ADDR_GEN_MODE_EUI64:
3334 		/* addrconf_add_linklocal also adds a prefix_route and we
3335 		 * only need to care about prefix routes if ipv6_generate_eui64
3336 		 * couldn't generate one.
3337 		 */
3338 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3339 			addrconf_add_linklocal(idev, &addr, 0);
3340 		else if (prefix_route)
3341 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3342 					      0, 0, GFP_KERNEL);
3343 		break;
3344 	case IN6_ADDR_GEN_MODE_NONE:
3345 	default:
3346 		/* will not add any link local address */
3347 		break;
3348 	}
3349 }
3350 
3351 static void addrconf_dev_config(struct net_device *dev)
3352 {
3353 	struct inet6_dev *idev;
3354 
3355 	ASSERT_RTNL();
3356 
3357 	if ((dev->type != ARPHRD_ETHER) &&
3358 	    (dev->type != ARPHRD_FDDI) &&
3359 	    (dev->type != ARPHRD_ARCNET) &&
3360 	    (dev->type != ARPHRD_INFINIBAND) &&
3361 	    (dev->type != ARPHRD_IEEE1394) &&
3362 	    (dev->type != ARPHRD_TUNNEL6) &&
3363 	    (dev->type != ARPHRD_6LOWPAN) &&
3364 	    (dev->type != ARPHRD_TUNNEL) &&
3365 	    (dev->type != ARPHRD_NONE) &&
3366 	    (dev->type != ARPHRD_RAWIP)) {
3367 		/* Alas, we support only Ethernet autoconfiguration. */
3368 		idev = __in6_dev_get(dev);
3369 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3370 		    dev->flags & IFF_MULTICAST)
3371 			ipv6_mc_up(idev);
3372 		return;
3373 	}
3374 
3375 	idev = addrconf_add_dev(dev);
3376 	if (IS_ERR(idev))
3377 		return;
3378 
3379 	/* this device type has no EUI support */
3380 	if (dev->type == ARPHRD_NONE &&
3381 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3382 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3383 
3384 	addrconf_addr_gen(idev, false);
3385 }
3386 
3387 #if IS_ENABLED(CONFIG_IPV6_SIT)
3388 static void addrconf_sit_config(struct net_device *dev)
3389 {
3390 	struct inet6_dev *idev;
3391 
3392 	ASSERT_RTNL();
3393 
3394 	/*
3395 	 * Configure the tunnel with one of our IPv4
3396 	 * addresses... we should configure all of
3397 	 * our v4 addrs in the tunnel
3398 	 */
3399 
3400 	idev = ipv6_find_idev(dev);
3401 	if (IS_ERR(idev)) {
3402 		pr_debug("%s: add_dev failed\n", __func__);
3403 		return;
3404 	}
3405 
3406 	if (dev->priv_flags & IFF_ISATAP) {
3407 		addrconf_addr_gen(idev, false);
3408 		return;
3409 	}
3410 
3411 	add_v4_addrs(idev);
3412 
3413 	if (dev->flags&IFF_POINTOPOINT)
3414 		addrconf_add_mroute(dev);
3415 }
3416 #endif
3417 
3418 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3419 static void addrconf_gre_config(struct net_device *dev)
3420 {
3421 	struct inet6_dev *idev;
3422 
3423 	ASSERT_RTNL();
3424 
3425 	idev = ipv6_find_idev(dev);
3426 	if (IS_ERR(idev)) {
3427 		pr_debug("%s: add_dev failed\n", __func__);
3428 		return;
3429 	}
3430 
3431 	if (dev->type == ARPHRD_ETHER) {
3432 		addrconf_addr_gen(idev, true);
3433 		return;
3434 	}
3435 
3436 	add_v4_addrs(idev);
3437 
3438 	if (dev->flags & IFF_POINTOPOINT)
3439 		addrconf_add_mroute(dev);
3440 }
3441 #endif
3442 
3443 static int fixup_permanent_addr(struct net *net,
3444 				struct inet6_dev *idev,
3445 				struct inet6_ifaddr *ifp)
3446 {
3447 	/* !fib6_node means the host route was removed from the
3448 	 * FIB, for example, if 'lo' device is taken down. In that
3449 	 * case regenerate the host route.
3450 	 */
3451 	if (!ifp->rt || !ifp->rt->fib6_node) {
3452 		struct fib6_info *f6i, *prev;
3453 
3454 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3455 					 GFP_ATOMIC);
3456 		if (IS_ERR(f6i))
3457 			return PTR_ERR(f6i);
3458 
3459 		/* ifp->rt can be accessed outside of rtnl */
3460 		spin_lock(&ifp->lock);
3461 		prev = ifp->rt;
3462 		ifp->rt = f6i;
3463 		spin_unlock(&ifp->lock);
3464 
3465 		fib6_info_release(prev);
3466 	}
3467 
3468 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3469 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3470 				      ifp->rt_priority, idev->dev, 0, 0,
3471 				      GFP_ATOMIC);
3472 	}
3473 
3474 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3475 		addrconf_dad_start(ifp);
3476 
3477 	return 0;
3478 }
3479 
3480 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3481 {
3482 	struct inet6_ifaddr *ifp, *tmp;
3483 	struct inet6_dev *idev;
3484 
3485 	idev = __in6_dev_get(dev);
3486 	if (!idev)
3487 		return;
3488 
3489 	write_lock_bh(&idev->lock);
3490 
3491 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3492 		if ((ifp->flags & IFA_F_PERMANENT) &&
3493 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3494 			write_unlock_bh(&idev->lock);
3495 			in6_ifa_hold(ifp);
3496 			ipv6_del_addr(ifp);
3497 			write_lock_bh(&idev->lock);
3498 
3499 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3500 					     idev->dev->name, &ifp->addr);
3501 		}
3502 	}
3503 
3504 	write_unlock_bh(&idev->lock);
3505 }
3506 
3507 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3508 			   void *ptr)
3509 {
3510 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3511 	struct netdev_notifier_change_info *change_info;
3512 	struct netdev_notifier_changeupper_info *info;
3513 	struct inet6_dev *idev = __in6_dev_get(dev);
3514 	struct net *net = dev_net(dev);
3515 	int run_pending = 0;
3516 	int err;
3517 
3518 	switch (event) {
3519 	case NETDEV_REGISTER:
3520 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3521 			idev = ipv6_add_dev(dev);
3522 			if (IS_ERR(idev))
3523 				return notifier_from_errno(PTR_ERR(idev));
3524 		}
3525 		break;
3526 
3527 	case NETDEV_CHANGEMTU:
3528 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3529 		if (dev->mtu < IPV6_MIN_MTU) {
3530 			addrconf_ifdown(dev, dev != net->loopback_dev);
3531 			break;
3532 		}
3533 
3534 		if (idev) {
3535 			rt6_mtu_change(dev, dev->mtu);
3536 			idev->cnf.mtu6 = dev->mtu;
3537 			break;
3538 		}
3539 
3540 		/* allocate new idev */
3541 		idev = ipv6_add_dev(dev);
3542 		if (IS_ERR(idev))
3543 			break;
3544 
3545 		/* device is still not ready */
3546 		if (!(idev->if_flags & IF_READY))
3547 			break;
3548 
3549 		run_pending = 1;
3550 		fallthrough;
3551 	case NETDEV_UP:
3552 	case NETDEV_CHANGE:
3553 		if (dev->flags & IFF_SLAVE)
3554 			break;
3555 
3556 		if (idev && idev->cnf.disable_ipv6)
3557 			break;
3558 
3559 		if (event == NETDEV_UP) {
3560 			/* restore routes for permanent addresses */
3561 			addrconf_permanent_addr(net, dev);
3562 
3563 			if (!addrconf_link_ready(dev)) {
3564 				/* device is not ready yet. */
3565 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3566 					 dev->name);
3567 				break;
3568 			}
3569 
3570 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3571 				idev = ipv6_add_dev(dev);
3572 
3573 			if (!IS_ERR_OR_NULL(idev)) {
3574 				idev->if_flags |= IF_READY;
3575 				run_pending = 1;
3576 			}
3577 		} else if (event == NETDEV_CHANGE) {
3578 			if (!addrconf_link_ready(dev)) {
3579 				/* device is still not ready. */
3580 				rt6_sync_down_dev(dev, event);
3581 				break;
3582 			}
3583 
3584 			if (!IS_ERR_OR_NULL(idev)) {
3585 				if (idev->if_flags & IF_READY) {
3586 					/* device is already configured -
3587 					 * but resend MLD reports, we might
3588 					 * have roamed and need to update
3589 					 * multicast snooping switches
3590 					 */
3591 					ipv6_mc_up(idev);
3592 					change_info = ptr;
3593 					if (change_info->flags_changed & IFF_NOARP)
3594 						addrconf_dad_run(idev, true);
3595 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3596 					break;
3597 				}
3598 				idev->if_flags |= IF_READY;
3599 			}
3600 
3601 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3602 				dev->name);
3603 
3604 			run_pending = 1;
3605 		}
3606 
3607 		switch (dev->type) {
3608 #if IS_ENABLED(CONFIG_IPV6_SIT)
3609 		case ARPHRD_SIT:
3610 			addrconf_sit_config(dev);
3611 			break;
3612 #endif
3613 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3614 		case ARPHRD_IP6GRE:
3615 		case ARPHRD_IPGRE:
3616 			addrconf_gre_config(dev);
3617 			break;
3618 #endif
3619 		case ARPHRD_LOOPBACK:
3620 			init_loopback(dev);
3621 			break;
3622 
3623 		default:
3624 			addrconf_dev_config(dev);
3625 			break;
3626 		}
3627 
3628 		if (!IS_ERR_OR_NULL(idev)) {
3629 			if (run_pending)
3630 				addrconf_dad_run(idev, false);
3631 
3632 			/* Device has an address by now */
3633 			rt6_sync_up(dev, RTNH_F_DEAD);
3634 
3635 			/*
3636 			 * If the MTU changed during the interface down,
3637 			 * when the interface up, the changed MTU must be
3638 			 * reflected in the idev as well as routers.
3639 			 */
3640 			if (idev->cnf.mtu6 != dev->mtu &&
3641 			    dev->mtu >= IPV6_MIN_MTU) {
3642 				rt6_mtu_change(dev, dev->mtu);
3643 				idev->cnf.mtu6 = dev->mtu;
3644 			}
3645 			idev->tstamp = jiffies;
3646 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3647 
3648 			/*
3649 			 * If the changed mtu during down is lower than
3650 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3651 			 */
3652 			if (dev->mtu < IPV6_MIN_MTU)
3653 				addrconf_ifdown(dev, dev != net->loopback_dev);
3654 		}
3655 		break;
3656 
3657 	case NETDEV_DOWN:
3658 	case NETDEV_UNREGISTER:
3659 		/*
3660 		 *	Remove all addresses from this interface.
3661 		 */
3662 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3663 		break;
3664 
3665 	case NETDEV_CHANGENAME:
3666 		if (idev) {
3667 			snmp6_unregister_dev(idev);
3668 			addrconf_sysctl_unregister(idev);
3669 			err = addrconf_sysctl_register(idev);
3670 			if (err)
3671 				return notifier_from_errno(err);
3672 			err = snmp6_register_dev(idev);
3673 			if (err) {
3674 				addrconf_sysctl_unregister(idev);
3675 				return notifier_from_errno(err);
3676 			}
3677 		}
3678 		break;
3679 
3680 	case NETDEV_PRE_TYPE_CHANGE:
3681 	case NETDEV_POST_TYPE_CHANGE:
3682 		if (idev)
3683 			addrconf_type_change(dev, event);
3684 		break;
3685 
3686 	case NETDEV_CHANGEUPPER:
3687 		info = ptr;
3688 
3689 		/* flush all routes if dev is linked to or unlinked from
3690 		 * an L3 master device (e.g., VRF)
3691 		 */
3692 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3693 			addrconf_ifdown(dev, false);
3694 	}
3695 
3696 	return NOTIFY_OK;
3697 }
3698 
3699 /*
3700  *	addrconf module should be notified of a device going up
3701  */
3702 static struct notifier_block ipv6_dev_notf = {
3703 	.notifier_call = addrconf_notify,
3704 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3705 };
3706 
3707 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3708 {
3709 	struct inet6_dev *idev;
3710 	ASSERT_RTNL();
3711 
3712 	idev = __in6_dev_get(dev);
3713 
3714 	if (event == NETDEV_POST_TYPE_CHANGE)
3715 		ipv6_mc_remap(idev);
3716 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3717 		ipv6_mc_unmap(idev);
3718 }
3719 
3720 static bool addr_is_local(const struct in6_addr *addr)
3721 {
3722 	return ipv6_addr_type(addr) &
3723 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3724 }
3725 
3726 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3727 {
3728 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3729 	struct net *net = dev_net(dev);
3730 	struct inet6_dev *idev;
3731 	struct inet6_ifaddr *ifa, *tmp;
3732 	bool keep_addr = false;
3733 	bool was_ready;
3734 	int state, i;
3735 
3736 	ASSERT_RTNL();
3737 
3738 	rt6_disable_ip(dev, event);
3739 
3740 	idev = __in6_dev_get(dev);
3741 	if (!idev)
3742 		return -ENODEV;
3743 
3744 	/*
3745 	 * Step 1: remove reference to ipv6 device from parent device.
3746 	 *	   Do not dev_put!
3747 	 */
3748 	if (unregister) {
3749 		idev->dead = 1;
3750 
3751 		/* protected by rtnl_lock */
3752 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3753 
3754 		/* Step 1.5: remove snmp6 entry */
3755 		snmp6_unregister_dev(idev);
3756 
3757 	}
3758 
3759 	/* combine the user config with event to determine if permanent
3760 	 * addresses are to be removed from address hash table
3761 	 */
3762 	if (!unregister && !idev->cnf.disable_ipv6) {
3763 		/* aggregate the system setting and interface setting */
3764 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3765 
3766 		if (!_keep_addr)
3767 			_keep_addr = idev->cnf.keep_addr_on_down;
3768 
3769 		keep_addr = (_keep_addr > 0);
3770 	}
3771 
3772 	/* Step 2: clear hash table */
3773 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3774 		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3775 
3776 		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3777 restart:
3778 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3779 			if (ifa->idev == idev) {
3780 				addrconf_del_dad_work(ifa);
3781 				/* combined flag + permanent flag decide if
3782 				 * address is retained on a down event
3783 				 */
3784 				if (!keep_addr ||
3785 				    !(ifa->flags & IFA_F_PERMANENT) ||
3786 				    addr_is_local(&ifa->addr)) {
3787 					hlist_del_init_rcu(&ifa->addr_lst);
3788 					goto restart;
3789 				}
3790 			}
3791 		}
3792 		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3793 	}
3794 
3795 	write_lock_bh(&idev->lock);
3796 
3797 	addrconf_del_rs_timer(idev);
3798 
3799 	/* Step 2: clear flags for stateless addrconf, repeated down
3800 	 *         detection
3801 	 */
3802 	was_ready = idev->if_flags & IF_READY;
3803 	if (!unregister)
3804 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3805 
3806 	/* Step 3: clear tempaddr list */
3807 	while (!list_empty(&idev->tempaddr_list)) {
3808 		ifa = list_first_entry(&idev->tempaddr_list,
3809 				       struct inet6_ifaddr, tmp_list);
3810 		list_del(&ifa->tmp_list);
3811 		write_unlock_bh(&idev->lock);
3812 		spin_lock_bh(&ifa->lock);
3813 
3814 		if (ifa->ifpub) {
3815 			in6_ifa_put(ifa->ifpub);
3816 			ifa->ifpub = NULL;
3817 		}
3818 		spin_unlock_bh(&ifa->lock);
3819 		in6_ifa_put(ifa);
3820 		write_lock_bh(&idev->lock);
3821 	}
3822 
3823 	list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3824 		struct fib6_info *rt = NULL;
3825 		bool keep;
3826 
3827 		addrconf_del_dad_work(ifa);
3828 
3829 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3830 			!addr_is_local(&ifa->addr);
3831 
3832 		write_unlock_bh(&idev->lock);
3833 		spin_lock_bh(&ifa->lock);
3834 
3835 		if (keep) {
3836 			/* set state to skip the notifier below */
3837 			state = INET6_IFADDR_STATE_DEAD;
3838 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3839 			if (!(ifa->flags & IFA_F_NODAD))
3840 				ifa->flags |= IFA_F_TENTATIVE;
3841 
3842 			rt = ifa->rt;
3843 			ifa->rt = NULL;
3844 		} else {
3845 			state = ifa->state;
3846 			ifa->state = INET6_IFADDR_STATE_DEAD;
3847 		}
3848 
3849 		spin_unlock_bh(&ifa->lock);
3850 
3851 		if (rt)
3852 			ip6_del_rt(net, rt, false);
3853 
3854 		if (state != INET6_IFADDR_STATE_DEAD) {
3855 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3856 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3857 		} else {
3858 			if (idev->cnf.forwarding)
3859 				addrconf_leave_anycast(ifa);
3860 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3861 		}
3862 
3863 		write_lock_bh(&idev->lock);
3864 		if (!keep) {
3865 			list_del_rcu(&ifa->if_list);
3866 			in6_ifa_put(ifa);
3867 		}
3868 	}
3869 
3870 	write_unlock_bh(&idev->lock);
3871 
3872 	/* Step 5: Discard anycast and multicast list */
3873 	if (unregister) {
3874 		ipv6_ac_destroy_dev(idev);
3875 		ipv6_mc_destroy_dev(idev);
3876 	} else if (was_ready) {
3877 		ipv6_mc_down(idev);
3878 	}
3879 
3880 	idev->tstamp = jiffies;
3881 	idev->ra_mtu = 0;
3882 
3883 	/* Last: Shot the device (if unregistered) */
3884 	if (unregister) {
3885 		addrconf_sysctl_unregister(idev);
3886 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3887 		neigh_ifdown(&nd_tbl, dev);
3888 		in6_dev_put(idev);
3889 	}
3890 	return 0;
3891 }
3892 
3893 static void addrconf_rs_timer(struct timer_list *t)
3894 {
3895 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3896 	struct net_device *dev = idev->dev;
3897 	struct in6_addr lladdr;
3898 
3899 	write_lock(&idev->lock);
3900 	if (idev->dead || !(idev->if_flags & IF_READY))
3901 		goto out;
3902 
3903 	if (!ipv6_accept_ra(idev))
3904 		goto out;
3905 
3906 	/* Announcement received after solicitation was sent */
3907 	if (idev->if_flags & IF_RA_RCVD)
3908 		goto out;
3909 
3910 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3911 		write_unlock(&idev->lock);
3912 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3913 			ndisc_send_rs(dev, &lladdr,
3914 				      &in6addr_linklocal_allrouters);
3915 		else
3916 			goto put;
3917 
3918 		write_lock(&idev->lock);
3919 		idev->rs_interval = rfc3315_s14_backoff_update(
3920 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3921 		/* The wait after the last probe can be shorter */
3922 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3923 					     idev->cnf.rtr_solicits) ?
3924 				      idev->cnf.rtr_solicit_delay :
3925 				      idev->rs_interval);
3926 	} else {
3927 		/*
3928 		 * Note: we do not support deprecated "all on-link"
3929 		 * assumption any longer.
3930 		 */
3931 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3932 	}
3933 
3934 out:
3935 	write_unlock(&idev->lock);
3936 put:
3937 	in6_dev_put(idev);
3938 }
3939 
3940 /*
3941  *	Duplicate Address Detection
3942  */
3943 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3944 {
3945 	unsigned long rand_num;
3946 	struct inet6_dev *idev = ifp->idev;
3947 	u64 nonce;
3948 
3949 	if (ifp->flags & IFA_F_OPTIMISTIC)
3950 		rand_num = 0;
3951 	else
3952 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3953 
3954 	nonce = 0;
3955 	if (idev->cnf.enhanced_dad ||
3956 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3957 		do
3958 			get_random_bytes(&nonce, 6);
3959 		while (nonce == 0);
3960 	}
3961 	ifp->dad_nonce = nonce;
3962 	ifp->dad_probes = idev->cnf.dad_transmits;
3963 	addrconf_mod_dad_work(ifp, rand_num);
3964 }
3965 
3966 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3967 {
3968 	struct inet6_dev *idev = ifp->idev;
3969 	struct net_device *dev = idev->dev;
3970 	bool bump_id, notify = false;
3971 	struct net *net;
3972 
3973 	addrconf_join_solict(dev, &ifp->addr);
3974 
3975 	read_lock_bh(&idev->lock);
3976 	spin_lock(&ifp->lock);
3977 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3978 		goto out;
3979 
3980 	net = dev_net(dev);
3981 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3982 	    (net->ipv6.devconf_all->accept_dad < 1 &&
3983 	     idev->cnf.accept_dad < 1) ||
3984 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3985 	    ifp->flags & IFA_F_NODAD) {
3986 		bool send_na = false;
3987 
3988 		if (ifp->flags & IFA_F_TENTATIVE &&
3989 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3990 			send_na = true;
3991 		bump_id = ifp->flags & IFA_F_TENTATIVE;
3992 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3993 		spin_unlock(&ifp->lock);
3994 		read_unlock_bh(&idev->lock);
3995 
3996 		addrconf_dad_completed(ifp, bump_id, send_na);
3997 		return;
3998 	}
3999 
4000 	if (!(idev->if_flags & IF_READY)) {
4001 		spin_unlock(&ifp->lock);
4002 		read_unlock_bh(&idev->lock);
4003 		/*
4004 		 * If the device is not ready:
4005 		 * - keep it tentative if it is a permanent address.
4006 		 * - otherwise, kill it.
4007 		 */
4008 		in6_ifa_hold(ifp);
4009 		addrconf_dad_stop(ifp, 0);
4010 		return;
4011 	}
4012 
4013 	/*
4014 	 * Optimistic nodes can start receiving
4015 	 * Frames right away
4016 	 */
4017 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4018 		ip6_ins_rt(net, ifp->rt);
4019 		if (ipv6_use_optimistic_addr(net, idev)) {
4020 			/* Because optimistic nodes can use this address,
4021 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4022 			 */
4023 			notify = true;
4024 		}
4025 	}
4026 
4027 	addrconf_dad_kick(ifp);
4028 out:
4029 	spin_unlock(&ifp->lock);
4030 	read_unlock_bh(&idev->lock);
4031 	if (notify)
4032 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4033 }
4034 
4035 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4036 {
4037 	bool begin_dad = false;
4038 
4039 	spin_lock_bh(&ifp->lock);
4040 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4041 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4042 		begin_dad = true;
4043 	}
4044 	spin_unlock_bh(&ifp->lock);
4045 
4046 	if (begin_dad)
4047 		addrconf_mod_dad_work(ifp, 0);
4048 }
4049 
4050 static void addrconf_dad_work(struct work_struct *w)
4051 {
4052 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4053 						struct inet6_ifaddr,
4054 						dad_work);
4055 	struct inet6_dev *idev = ifp->idev;
4056 	bool bump_id, disable_ipv6 = false;
4057 	struct in6_addr mcaddr;
4058 
4059 	enum {
4060 		DAD_PROCESS,
4061 		DAD_BEGIN,
4062 		DAD_ABORT,
4063 	} action = DAD_PROCESS;
4064 
4065 	rtnl_lock();
4066 
4067 	spin_lock_bh(&ifp->lock);
4068 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4069 		action = DAD_BEGIN;
4070 		ifp->state = INET6_IFADDR_STATE_DAD;
4071 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4072 		action = DAD_ABORT;
4073 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4074 
4075 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4076 		     idev->cnf.accept_dad > 1) &&
4077 		    !idev->cnf.disable_ipv6 &&
4078 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4079 			struct in6_addr addr;
4080 
4081 			addr.s6_addr32[0] = htonl(0xfe800000);
4082 			addr.s6_addr32[1] = 0;
4083 
4084 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4085 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4086 				/* DAD failed for link-local based on MAC */
4087 				idev->cnf.disable_ipv6 = 1;
4088 
4089 				pr_info("%s: IPv6 being disabled!\n",
4090 					ifp->idev->dev->name);
4091 				disable_ipv6 = true;
4092 			}
4093 		}
4094 	}
4095 	spin_unlock_bh(&ifp->lock);
4096 
4097 	if (action == DAD_BEGIN) {
4098 		addrconf_dad_begin(ifp);
4099 		goto out;
4100 	} else if (action == DAD_ABORT) {
4101 		in6_ifa_hold(ifp);
4102 		addrconf_dad_stop(ifp, 1);
4103 		if (disable_ipv6)
4104 			addrconf_ifdown(idev->dev, false);
4105 		goto out;
4106 	}
4107 
4108 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4109 		goto out;
4110 
4111 	write_lock_bh(&idev->lock);
4112 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4113 		write_unlock_bh(&idev->lock);
4114 		goto out;
4115 	}
4116 
4117 	spin_lock(&ifp->lock);
4118 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4119 		spin_unlock(&ifp->lock);
4120 		write_unlock_bh(&idev->lock);
4121 		goto out;
4122 	}
4123 
4124 	if (ifp->dad_probes == 0) {
4125 		bool send_na = false;
4126 
4127 		/*
4128 		 * DAD was successful
4129 		 */
4130 
4131 		if (ifp->flags & IFA_F_TENTATIVE &&
4132 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4133 			send_na = true;
4134 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4135 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4136 		spin_unlock(&ifp->lock);
4137 		write_unlock_bh(&idev->lock);
4138 
4139 		addrconf_dad_completed(ifp, bump_id, send_na);
4140 
4141 		goto out;
4142 	}
4143 
4144 	ifp->dad_probes--;
4145 	addrconf_mod_dad_work(ifp,
4146 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4147 				  HZ/100));
4148 	spin_unlock(&ifp->lock);
4149 	write_unlock_bh(&idev->lock);
4150 
4151 	/* send a neighbour solicitation for our addr */
4152 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4153 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4154 		      ifp->dad_nonce);
4155 out:
4156 	in6_ifa_put(ifp);
4157 	rtnl_unlock();
4158 }
4159 
4160 /* ifp->idev must be at least read locked */
4161 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4162 {
4163 	struct inet6_ifaddr *ifpiter;
4164 	struct inet6_dev *idev = ifp->idev;
4165 
4166 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4167 		if (ifpiter->scope > IFA_LINK)
4168 			break;
4169 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4170 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4171 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4172 		    IFA_F_PERMANENT)
4173 			return false;
4174 	}
4175 	return true;
4176 }
4177 
4178 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4179 				   bool send_na)
4180 {
4181 	struct net_device *dev = ifp->idev->dev;
4182 	struct in6_addr lladdr;
4183 	bool send_rs, send_mld;
4184 
4185 	addrconf_del_dad_work(ifp);
4186 
4187 	/*
4188 	 *	Configure the address for reception. Now it is valid.
4189 	 */
4190 
4191 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4192 
4193 	/* If added prefix is link local and we are prepared to process
4194 	   router advertisements, start sending router solicitations.
4195 	 */
4196 
4197 	read_lock_bh(&ifp->idev->lock);
4198 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4199 	send_rs = send_mld &&
4200 		  ipv6_accept_ra(ifp->idev) &&
4201 		  ifp->idev->cnf.rtr_solicits != 0 &&
4202 		  (dev->flags&IFF_LOOPBACK) == 0;
4203 	read_unlock_bh(&ifp->idev->lock);
4204 
4205 	/* While dad is in progress mld report's source address is in6_addrany.
4206 	 * Resend with proper ll now.
4207 	 */
4208 	if (send_mld)
4209 		ipv6_mc_dad_complete(ifp->idev);
4210 
4211 	/* send unsolicited NA if enabled */
4212 	if (send_na &&
4213 	    (ifp->idev->cnf.ndisc_notify ||
4214 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4215 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4216 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4217 			      /*solicited=*/ false, /*override=*/ true,
4218 			      /*inc_opt=*/ true);
4219 	}
4220 
4221 	if (send_rs) {
4222 		/*
4223 		 *	If a host as already performed a random delay
4224 		 *	[...] as part of DAD [...] there is no need
4225 		 *	to delay again before sending the first RS
4226 		 */
4227 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4228 			return;
4229 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4230 
4231 		write_lock_bh(&ifp->idev->lock);
4232 		spin_lock(&ifp->lock);
4233 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4234 			ifp->idev->cnf.rtr_solicit_interval);
4235 		ifp->idev->rs_probes = 1;
4236 		ifp->idev->if_flags |= IF_RS_SENT;
4237 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4238 		spin_unlock(&ifp->lock);
4239 		write_unlock_bh(&ifp->idev->lock);
4240 	}
4241 
4242 	if (bump_id)
4243 		rt_genid_bump_ipv6(dev_net(dev));
4244 
4245 	/* Make sure that a new temporary address will be created
4246 	 * before this temporary address becomes deprecated.
4247 	 */
4248 	if (ifp->flags & IFA_F_TEMPORARY)
4249 		addrconf_verify_rtnl(dev_net(dev));
4250 }
4251 
4252 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4253 {
4254 	struct inet6_ifaddr *ifp;
4255 
4256 	read_lock_bh(&idev->lock);
4257 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4258 		spin_lock(&ifp->lock);
4259 		if ((ifp->flags & IFA_F_TENTATIVE &&
4260 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4261 			if (restart)
4262 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4263 			addrconf_dad_kick(ifp);
4264 		}
4265 		spin_unlock(&ifp->lock);
4266 	}
4267 	read_unlock_bh(&idev->lock);
4268 }
4269 
4270 #ifdef CONFIG_PROC_FS
4271 struct if6_iter_state {
4272 	struct seq_net_private p;
4273 	int bucket;
4274 	int offset;
4275 };
4276 
4277 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4278 {
4279 	struct if6_iter_state *state = seq->private;
4280 	struct net *net = seq_file_net(seq);
4281 	struct inet6_ifaddr *ifa = NULL;
4282 	int p = 0;
4283 
4284 	/* initial bucket if pos is 0 */
4285 	if (pos == 0) {
4286 		state->bucket = 0;
4287 		state->offset = 0;
4288 	}
4289 
4290 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4291 		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4292 					 addr_lst) {
4293 			/* sync with offset */
4294 			if (p < state->offset) {
4295 				p++;
4296 				continue;
4297 			}
4298 			return ifa;
4299 		}
4300 
4301 		/* prepare for next bucket */
4302 		state->offset = 0;
4303 		p = 0;
4304 	}
4305 	return NULL;
4306 }
4307 
4308 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4309 					 struct inet6_ifaddr *ifa)
4310 {
4311 	struct if6_iter_state *state = seq->private;
4312 	struct net *net = seq_file_net(seq);
4313 
4314 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4315 		state->offset++;
4316 		return ifa;
4317 	}
4318 
4319 	state->offset = 0;
4320 	while (++state->bucket < IN6_ADDR_HSIZE) {
4321 		hlist_for_each_entry_rcu(ifa,
4322 				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4323 			return ifa;
4324 		}
4325 	}
4326 
4327 	return NULL;
4328 }
4329 
4330 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4331 	__acquires(rcu)
4332 {
4333 	rcu_read_lock();
4334 	return if6_get_first(seq, *pos);
4335 }
4336 
4337 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4338 {
4339 	struct inet6_ifaddr *ifa;
4340 
4341 	ifa = if6_get_next(seq, v);
4342 	++*pos;
4343 	return ifa;
4344 }
4345 
4346 static void if6_seq_stop(struct seq_file *seq, void *v)
4347 	__releases(rcu)
4348 {
4349 	rcu_read_unlock();
4350 }
4351 
4352 static int if6_seq_show(struct seq_file *seq, void *v)
4353 {
4354 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4355 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4356 		   &ifp->addr,
4357 		   ifp->idev->dev->ifindex,
4358 		   ifp->prefix_len,
4359 		   ifp->scope,
4360 		   (u8) ifp->flags,
4361 		   ifp->idev->dev->name);
4362 	return 0;
4363 }
4364 
4365 static const struct seq_operations if6_seq_ops = {
4366 	.start	= if6_seq_start,
4367 	.next	= if6_seq_next,
4368 	.show	= if6_seq_show,
4369 	.stop	= if6_seq_stop,
4370 };
4371 
4372 static int __net_init if6_proc_net_init(struct net *net)
4373 {
4374 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4375 			sizeof(struct if6_iter_state)))
4376 		return -ENOMEM;
4377 	return 0;
4378 }
4379 
4380 static void __net_exit if6_proc_net_exit(struct net *net)
4381 {
4382 	remove_proc_entry("if_inet6", net->proc_net);
4383 }
4384 
4385 static struct pernet_operations if6_proc_net_ops = {
4386 	.init = if6_proc_net_init,
4387 	.exit = if6_proc_net_exit,
4388 };
4389 
4390 int __init if6_proc_init(void)
4391 {
4392 	return register_pernet_subsys(&if6_proc_net_ops);
4393 }
4394 
4395 void if6_proc_exit(void)
4396 {
4397 	unregister_pernet_subsys(&if6_proc_net_ops);
4398 }
4399 #endif	/* CONFIG_PROC_FS */
4400 
4401 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4402 /* Check if address is a home address configured on any interface. */
4403 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4404 {
4405 	unsigned int hash = inet6_addr_hash(net, addr);
4406 	struct inet6_ifaddr *ifp = NULL;
4407 	int ret = 0;
4408 
4409 	rcu_read_lock();
4410 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4411 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4412 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4413 			ret = 1;
4414 			break;
4415 		}
4416 	}
4417 	rcu_read_unlock();
4418 	return ret;
4419 }
4420 #endif
4421 
4422 /* RFC6554 has some algorithm to avoid loops in segment routing by
4423  * checking if the segments contains any of a local interface address.
4424  *
4425  * Quote:
4426  *
4427  * To detect loops in the SRH, a router MUST determine if the SRH
4428  * includes multiple addresses assigned to any interface on that router.
4429  * If such addresses appear more than once and are separated by at least
4430  * one address not assigned to that router.
4431  */
4432 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4433 			  unsigned char nsegs)
4434 {
4435 	const struct in6_addr *addr;
4436 	int i, ret = 0, found = 0;
4437 	struct inet6_ifaddr *ifp;
4438 	bool separated = false;
4439 	unsigned int hash;
4440 	bool hash_found;
4441 
4442 	rcu_read_lock();
4443 	for (i = 0; i < nsegs; i++) {
4444 		addr = &segs[i];
4445 		hash = inet6_addr_hash(net, addr);
4446 
4447 		hash_found = false;
4448 		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4449 
4450 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4451 				hash_found = true;
4452 				break;
4453 			}
4454 		}
4455 
4456 		if (hash_found) {
4457 			if (found > 1 && separated) {
4458 				ret = 1;
4459 				break;
4460 			}
4461 
4462 			separated = false;
4463 			found++;
4464 		} else {
4465 			separated = true;
4466 		}
4467 	}
4468 	rcu_read_unlock();
4469 
4470 	return ret;
4471 }
4472 
4473 /*
4474  *	Periodic address status verification
4475  */
4476 
4477 static void addrconf_verify_rtnl(struct net *net)
4478 {
4479 	unsigned long now, next, next_sec, next_sched;
4480 	struct inet6_ifaddr *ifp;
4481 	int i;
4482 
4483 	ASSERT_RTNL();
4484 
4485 	rcu_read_lock_bh();
4486 	now = jiffies;
4487 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4488 
4489 	cancel_delayed_work(&net->ipv6.addr_chk_work);
4490 
4491 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4492 restart:
4493 		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4494 			unsigned long age;
4495 
4496 			/* When setting preferred_lft to a value not zero or
4497 			 * infinity, while valid_lft is infinity
4498 			 * IFA_F_PERMANENT has a non-infinity life time.
4499 			 */
4500 			if ((ifp->flags & IFA_F_PERMANENT) &&
4501 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4502 				continue;
4503 
4504 			spin_lock(&ifp->lock);
4505 			/* We try to batch several events at once. */
4506 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4507 
4508 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4509 			    age >= ifp->valid_lft) {
4510 				spin_unlock(&ifp->lock);
4511 				in6_ifa_hold(ifp);
4512 				rcu_read_unlock_bh();
4513 				ipv6_del_addr(ifp);
4514 				rcu_read_lock_bh();
4515 				goto restart;
4516 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4517 				spin_unlock(&ifp->lock);
4518 				continue;
4519 			} else if (age >= ifp->prefered_lft) {
4520 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4521 				int deprecate = 0;
4522 
4523 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4524 					deprecate = 1;
4525 					ifp->flags |= IFA_F_DEPRECATED;
4526 				}
4527 
4528 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4529 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4530 					next = ifp->tstamp + ifp->valid_lft * HZ;
4531 
4532 				spin_unlock(&ifp->lock);
4533 
4534 				if (deprecate) {
4535 					in6_ifa_hold(ifp);
4536 
4537 					ipv6_ifa_notify(0, ifp);
4538 					in6_ifa_put(ifp);
4539 					goto restart;
4540 				}
4541 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4542 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4543 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4544 					ifp->idev->cnf.dad_transmits *
4545 					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4546 
4547 				if (age >= ifp->prefered_lft - regen_advance) {
4548 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4549 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4550 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4551 					if (!ifp->regen_count && ifpub) {
4552 						ifp->regen_count++;
4553 						in6_ifa_hold(ifp);
4554 						in6_ifa_hold(ifpub);
4555 						spin_unlock(&ifp->lock);
4556 
4557 						spin_lock(&ifpub->lock);
4558 						ifpub->regen_count = 0;
4559 						spin_unlock(&ifpub->lock);
4560 						rcu_read_unlock_bh();
4561 						ipv6_create_tempaddr(ifpub, true);
4562 						in6_ifa_put(ifpub);
4563 						in6_ifa_put(ifp);
4564 						rcu_read_lock_bh();
4565 						goto restart;
4566 					}
4567 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4568 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4569 				spin_unlock(&ifp->lock);
4570 			} else {
4571 				/* ifp->prefered_lft <= ifp->valid_lft */
4572 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4573 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4574 				spin_unlock(&ifp->lock);
4575 			}
4576 		}
4577 	}
4578 
4579 	next_sec = round_jiffies_up(next);
4580 	next_sched = next;
4581 
4582 	/* If rounded timeout is accurate enough, accept it. */
4583 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4584 		next_sched = next_sec;
4585 
4586 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4587 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4588 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4589 
4590 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4591 		 now, next, next_sec, next_sched);
4592 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4593 	rcu_read_unlock_bh();
4594 }
4595 
4596 static void addrconf_verify_work(struct work_struct *w)
4597 {
4598 	struct net *net = container_of(to_delayed_work(w), struct net,
4599 				       ipv6.addr_chk_work);
4600 
4601 	rtnl_lock();
4602 	addrconf_verify_rtnl(net);
4603 	rtnl_unlock();
4604 }
4605 
4606 static void addrconf_verify(struct net *net)
4607 {
4608 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4609 }
4610 
4611 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4612 				     struct in6_addr **peer_pfx)
4613 {
4614 	struct in6_addr *pfx = NULL;
4615 
4616 	*peer_pfx = NULL;
4617 
4618 	if (addr)
4619 		pfx = nla_data(addr);
4620 
4621 	if (local) {
4622 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4623 			*peer_pfx = pfx;
4624 		pfx = nla_data(local);
4625 	}
4626 
4627 	return pfx;
4628 }
4629 
4630 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4631 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4632 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4633 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4634 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4635 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4636 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4637 	[IFA_PROTO]		= { .type = NLA_U8 },
4638 };
4639 
4640 static int
4641 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4642 		  struct netlink_ext_ack *extack)
4643 {
4644 	struct net *net = sock_net(skb->sk);
4645 	struct ifaddrmsg *ifm;
4646 	struct nlattr *tb[IFA_MAX+1];
4647 	struct in6_addr *pfx, *peer_pfx;
4648 	u32 ifa_flags;
4649 	int err;
4650 
4651 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4652 				     ifa_ipv6_policy, extack);
4653 	if (err < 0)
4654 		return err;
4655 
4656 	ifm = nlmsg_data(nlh);
4657 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4658 	if (!pfx)
4659 		return -EINVAL;
4660 
4661 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4662 
4663 	/* We ignore other flags so far. */
4664 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4665 
4666 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4667 			      ifm->ifa_prefixlen);
4668 }
4669 
4670 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4671 			       unsigned long expires, u32 flags,
4672 			       bool modify_peer)
4673 {
4674 	struct fib6_info *f6i;
4675 	u32 prio;
4676 
4677 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4678 					ifp->prefix_len,
4679 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4680 	if (!f6i)
4681 		return -ENOENT;
4682 
4683 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4684 	if (f6i->fib6_metric != prio) {
4685 		/* delete old one */
4686 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4687 
4688 		/* add new one */
4689 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4690 				      ifp->prefix_len,
4691 				      ifp->rt_priority, ifp->idev->dev,
4692 				      expires, flags, GFP_KERNEL);
4693 	} else {
4694 		if (!expires)
4695 			fib6_clean_expires(f6i);
4696 		else
4697 			fib6_set_expires(f6i, expires);
4698 
4699 		fib6_info_release(f6i);
4700 	}
4701 
4702 	return 0;
4703 }
4704 
4705 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4706 			     struct ifa6_config *cfg)
4707 {
4708 	u32 flags;
4709 	clock_t expires;
4710 	unsigned long timeout;
4711 	bool was_managetempaddr;
4712 	bool had_prefixroute;
4713 	bool new_peer = false;
4714 
4715 	ASSERT_RTNL();
4716 
4717 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4718 		return -EINVAL;
4719 
4720 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4721 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4722 		return -EINVAL;
4723 
4724 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4725 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4726 
4727 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4728 	if (addrconf_finite_timeout(timeout)) {
4729 		expires = jiffies_to_clock_t(timeout * HZ);
4730 		cfg->valid_lft = timeout;
4731 		flags = RTF_EXPIRES;
4732 	} else {
4733 		expires = 0;
4734 		flags = 0;
4735 		cfg->ifa_flags |= IFA_F_PERMANENT;
4736 	}
4737 
4738 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4739 	if (addrconf_finite_timeout(timeout)) {
4740 		if (timeout == 0)
4741 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4742 		cfg->preferred_lft = timeout;
4743 	}
4744 
4745 	if (cfg->peer_pfx &&
4746 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4747 		if (!ipv6_addr_any(&ifp->peer_addr))
4748 			cleanup_prefix_route(ifp, expires, true, true);
4749 		new_peer = true;
4750 	}
4751 
4752 	spin_lock_bh(&ifp->lock);
4753 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4754 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4755 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4756 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4757 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4758 			IFA_F_NOPREFIXROUTE);
4759 	ifp->flags |= cfg->ifa_flags;
4760 	ifp->tstamp = jiffies;
4761 	ifp->valid_lft = cfg->valid_lft;
4762 	ifp->prefered_lft = cfg->preferred_lft;
4763 	ifp->ifa_proto = cfg->ifa_proto;
4764 
4765 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4766 		ifp->rt_priority = cfg->rt_priority;
4767 
4768 	if (new_peer)
4769 		ifp->peer_addr = *cfg->peer_pfx;
4770 
4771 	spin_unlock_bh(&ifp->lock);
4772 	if (!(ifp->flags&IFA_F_TENTATIVE))
4773 		ipv6_ifa_notify(0, ifp);
4774 
4775 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4776 		int rc = -ENOENT;
4777 
4778 		if (had_prefixroute)
4779 			rc = modify_prefix_route(ifp, expires, flags, false);
4780 
4781 		/* prefix route could have been deleted; if so restore it */
4782 		if (rc == -ENOENT) {
4783 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4784 					      ifp->rt_priority, ifp->idev->dev,
4785 					      expires, flags, GFP_KERNEL);
4786 		}
4787 
4788 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4789 			rc = modify_prefix_route(ifp, expires, flags, true);
4790 
4791 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4792 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4793 					      ifp->rt_priority, ifp->idev->dev,
4794 					      expires, flags, GFP_KERNEL);
4795 		}
4796 	} else if (had_prefixroute) {
4797 		enum cleanup_prefix_rt_t action;
4798 		unsigned long rt_expires;
4799 
4800 		write_lock_bh(&ifp->idev->lock);
4801 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4802 		write_unlock_bh(&ifp->idev->lock);
4803 
4804 		if (action != CLEANUP_PREFIX_RT_NOP) {
4805 			cleanup_prefix_route(ifp, rt_expires,
4806 				action == CLEANUP_PREFIX_RT_DEL, false);
4807 		}
4808 	}
4809 
4810 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4811 		if (was_managetempaddr &&
4812 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4813 			cfg->valid_lft = 0;
4814 			cfg->preferred_lft = 0;
4815 		}
4816 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4817 				 cfg->preferred_lft, !was_managetempaddr,
4818 				 jiffies);
4819 	}
4820 
4821 	addrconf_verify_rtnl(net);
4822 
4823 	return 0;
4824 }
4825 
4826 static int
4827 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4828 		  struct netlink_ext_ack *extack)
4829 {
4830 	struct net *net = sock_net(skb->sk);
4831 	struct ifaddrmsg *ifm;
4832 	struct nlattr *tb[IFA_MAX+1];
4833 	struct in6_addr *peer_pfx;
4834 	struct inet6_ifaddr *ifa;
4835 	struct net_device *dev;
4836 	struct inet6_dev *idev;
4837 	struct ifa6_config cfg;
4838 	int err;
4839 
4840 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4841 				     ifa_ipv6_policy, extack);
4842 	if (err < 0)
4843 		return err;
4844 
4845 	memset(&cfg, 0, sizeof(cfg));
4846 
4847 	ifm = nlmsg_data(nlh);
4848 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4849 	if (!cfg.pfx)
4850 		return -EINVAL;
4851 
4852 	cfg.peer_pfx = peer_pfx;
4853 	cfg.plen = ifm->ifa_prefixlen;
4854 	if (tb[IFA_RT_PRIORITY])
4855 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4856 
4857 	if (tb[IFA_PROTO])
4858 		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4859 
4860 	cfg.valid_lft = INFINITY_LIFE_TIME;
4861 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4862 
4863 	if (tb[IFA_CACHEINFO]) {
4864 		struct ifa_cacheinfo *ci;
4865 
4866 		ci = nla_data(tb[IFA_CACHEINFO]);
4867 		cfg.valid_lft = ci->ifa_valid;
4868 		cfg.preferred_lft = ci->ifa_prefered;
4869 	}
4870 
4871 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4872 	if (!dev)
4873 		return -ENODEV;
4874 
4875 	if (tb[IFA_FLAGS])
4876 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4877 	else
4878 		cfg.ifa_flags = ifm->ifa_flags;
4879 
4880 	/* We ignore other flags so far. */
4881 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4882 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4883 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4884 
4885 	idev = ipv6_find_idev(dev);
4886 	if (IS_ERR(idev))
4887 		return PTR_ERR(idev);
4888 
4889 	if (!ipv6_allow_optimistic_dad(net, idev))
4890 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4891 
4892 	if (cfg.ifa_flags & IFA_F_NODAD &&
4893 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4894 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4895 		return -EINVAL;
4896 	}
4897 
4898 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4899 	if (!ifa) {
4900 		/*
4901 		 * It would be best to check for !NLM_F_CREATE here but
4902 		 * userspace already relies on not having to provide this.
4903 		 */
4904 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4905 	}
4906 
4907 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4908 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4909 		err = -EEXIST;
4910 	else
4911 		err = inet6_addr_modify(net, ifa, &cfg);
4912 
4913 	in6_ifa_put(ifa);
4914 
4915 	return err;
4916 }
4917 
4918 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4919 			  u8 scope, int ifindex)
4920 {
4921 	struct ifaddrmsg *ifm;
4922 
4923 	ifm = nlmsg_data(nlh);
4924 	ifm->ifa_family = AF_INET6;
4925 	ifm->ifa_prefixlen = prefixlen;
4926 	ifm->ifa_flags = flags;
4927 	ifm->ifa_scope = scope;
4928 	ifm->ifa_index = ifindex;
4929 }
4930 
4931 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4932 			 unsigned long tstamp, u32 preferred, u32 valid)
4933 {
4934 	struct ifa_cacheinfo ci;
4935 
4936 	ci.cstamp = cstamp_delta(cstamp);
4937 	ci.tstamp = cstamp_delta(tstamp);
4938 	ci.ifa_prefered = preferred;
4939 	ci.ifa_valid = valid;
4940 
4941 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4942 }
4943 
4944 static inline int rt_scope(int ifa_scope)
4945 {
4946 	if (ifa_scope & IFA_HOST)
4947 		return RT_SCOPE_HOST;
4948 	else if (ifa_scope & IFA_LINK)
4949 		return RT_SCOPE_LINK;
4950 	else if (ifa_scope & IFA_SITE)
4951 		return RT_SCOPE_SITE;
4952 	else
4953 		return RT_SCOPE_UNIVERSE;
4954 }
4955 
4956 static inline int inet6_ifaddr_msgsize(void)
4957 {
4958 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4959 	       + nla_total_size(16) /* IFA_LOCAL */
4960 	       + nla_total_size(16) /* IFA_ADDRESS */
4961 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4962 	       + nla_total_size(4)  /* IFA_FLAGS */
4963 	       + nla_total_size(1)  /* IFA_PROTO */
4964 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4965 }
4966 
4967 enum addr_type_t {
4968 	UNICAST_ADDR,
4969 	MULTICAST_ADDR,
4970 	ANYCAST_ADDR,
4971 };
4972 
4973 struct inet6_fill_args {
4974 	u32 portid;
4975 	u32 seq;
4976 	int event;
4977 	unsigned int flags;
4978 	int netnsid;
4979 	int ifindex;
4980 	enum addr_type_t type;
4981 };
4982 
4983 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4984 			     struct inet6_fill_args *args)
4985 {
4986 	struct nlmsghdr  *nlh;
4987 	u32 preferred, valid;
4988 
4989 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4990 			sizeof(struct ifaddrmsg), args->flags);
4991 	if (!nlh)
4992 		return -EMSGSIZE;
4993 
4994 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4995 		      ifa->idev->dev->ifindex);
4996 
4997 	if (args->netnsid >= 0 &&
4998 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4999 		goto error;
5000 
5001 	spin_lock_bh(&ifa->lock);
5002 	if (!((ifa->flags&IFA_F_PERMANENT) &&
5003 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5004 		preferred = ifa->prefered_lft;
5005 		valid = ifa->valid_lft;
5006 		if (preferred != INFINITY_LIFE_TIME) {
5007 			long tval = (jiffies - ifa->tstamp)/HZ;
5008 			if (preferred > tval)
5009 				preferred -= tval;
5010 			else
5011 				preferred = 0;
5012 			if (valid != INFINITY_LIFE_TIME) {
5013 				if (valid > tval)
5014 					valid -= tval;
5015 				else
5016 					valid = 0;
5017 			}
5018 		}
5019 	} else {
5020 		preferred = INFINITY_LIFE_TIME;
5021 		valid = INFINITY_LIFE_TIME;
5022 	}
5023 	spin_unlock_bh(&ifa->lock);
5024 
5025 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5026 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5027 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5028 			goto error;
5029 	} else
5030 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5031 			goto error;
5032 
5033 	if (ifa->rt_priority &&
5034 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5035 		goto error;
5036 
5037 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5038 		goto error;
5039 
5040 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5041 		goto error;
5042 
5043 	if (ifa->ifa_proto &&
5044 	    nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5045 		goto error;
5046 
5047 	nlmsg_end(skb, nlh);
5048 	return 0;
5049 
5050 error:
5051 	nlmsg_cancel(skb, nlh);
5052 	return -EMSGSIZE;
5053 }
5054 
5055 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5056 			       struct inet6_fill_args *args)
5057 {
5058 	struct nlmsghdr  *nlh;
5059 	u8 scope = RT_SCOPE_UNIVERSE;
5060 	int ifindex = ifmca->idev->dev->ifindex;
5061 
5062 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5063 		scope = RT_SCOPE_SITE;
5064 
5065 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5066 			sizeof(struct ifaddrmsg), args->flags);
5067 	if (!nlh)
5068 		return -EMSGSIZE;
5069 
5070 	if (args->netnsid >= 0 &&
5071 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5072 		nlmsg_cancel(skb, nlh);
5073 		return -EMSGSIZE;
5074 	}
5075 
5076 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5077 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5078 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5079 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5080 		nlmsg_cancel(skb, nlh);
5081 		return -EMSGSIZE;
5082 	}
5083 
5084 	nlmsg_end(skb, nlh);
5085 	return 0;
5086 }
5087 
5088 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5089 			       struct inet6_fill_args *args)
5090 {
5091 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5092 	int ifindex = dev ? dev->ifindex : 1;
5093 	struct nlmsghdr  *nlh;
5094 	u8 scope = RT_SCOPE_UNIVERSE;
5095 
5096 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5097 		scope = RT_SCOPE_SITE;
5098 
5099 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5100 			sizeof(struct ifaddrmsg), args->flags);
5101 	if (!nlh)
5102 		return -EMSGSIZE;
5103 
5104 	if (args->netnsid >= 0 &&
5105 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5106 		nlmsg_cancel(skb, nlh);
5107 		return -EMSGSIZE;
5108 	}
5109 
5110 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5111 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5112 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5113 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5114 		nlmsg_cancel(skb, nlh);
5115 		return -EMSGSIZE;
5116 	}
5117 
5118 	nlmsg_end(skb, nlh);
5119 	return 0;
5120 }
5121 
5122 /* called with rcu_read_lock() */
5123 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5124 			  struct netlink_callback *cb, int s_ip_idx,
5125 			  struct inet6_fill_args *fillargs)
5126 {
5127 	struct ifmcaddr6 *ifmca;
5128 	struct ifacaddr6 *ifaca;
5129 	int ip_idx = 0;
5130 	int err = 1;
5131 
5132 	read_lock_bh(&idev->lock);
5133 	switch (fillargs->type) {
5134 	case UNICAST_ADDR: {
5135 		struct inet6_ifaddr *ifa;
5136 		fillargs->event = RTM_NEWADDR;
5137 
5138 		/* unicast address incl. temp addr */
5139 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
5140 			if (ip_idx < s_ip_idx)
5141 				goto next;
5142 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5143 			if (err < 0)
5144 				break;
5145 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5146 next:
5147 			ip_idx++;
5148 		}
5149 		break;
5150 	}
5151 	case MULTICAST_ADDR:
5152 		read_unlock_bh(&idev->lock);
5153 		fillargs->event = RTM_GETMULTICAST;
5154 
5155 		/* multicast address */
5156 		for (ifmca = rcu_dereference(idev->mc_list);
5157 		     ifmca;
5158 		     ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5159 			if (ip_idx < s_ip_idx)
5160 				continue;
5161 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5162 			if (err < 0)
5163 				break;
5164 		}
5165 		read_lock_bh(&idev->lock);
5166 		break;
5167 	case ANYCAST_ADDR:
5168 		fillargs->event = RTM_GETANYCAST;
5169 		/* anycast address */
5170 		for (ifaca = idev->ac_list; ifaca;
5171 		     ifaca = ifaca->aca_next, ip_idx++) {
5172 			if (ip_idx < s_ip_idx)
5173 				continue;
5174 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5175 			if (err < 0)
5176 				break;
5177 		}
5178 		break;
5179 	default:
5180 		break;
5181 	}
5182 	read_unlock_bh(&idev->lock);
5183 	cb->args[2] = ip_idx;
5184 	return err;
5185 }
5186 
5187 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5188 				       struct inet6_fill_args *fillargs,
5189 				       struct net **tgt_net, struct sock *sk,
5190 				       struct netlink_callback *cb)
5191 {
5192 	struct netlink_ext_ack *extack = cb->extack;
5193 	struct nlattr *tb[IFA_MAX+1];
5194 	struct ifaddrmsg *ifm;
5195 	int err, i;
5196 
5197 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5198 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5199 		return -EINVAL;
5200 	}
5201 
5202 	ifm = nlmsg_data(nlh);
5203 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5204 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5205 		return -EINVAL;
5206 	}
5207 
5208 	fillargs->ifindex = ifm->ifa_index;
5209 	if (fillargs->ifindex) {
5210 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5211 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5212 	}
5213 
5214 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5215 					    ifa_ipv6_policy, extack);
5216 	if (err < 0)
5217 		return err;
5218 
5219 	for (i = 0; i <= IFA_MAX; ++i) {
5220 		if (!tb[i])
5221 			continue;
5222 
5223 		if (i == IFA_TARGET_NETNSID) {
5224 			struct net *net;
5225 
5226 			fillargs->netnsid = nla_get_s32(tb[i]);
5227 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5228 			if (IS_ERR(net)) {
5229 				fillargs->netnsid = -1;
5230 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5231 				return PTR_ERR(net);
5232 			}
5233 			*tgt_net = net;
5234 		} else {
5235 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5236 			return -EINVAL;
5237 		}
5238 	}
5239 
5240 	return 0;
5241 }
5242 
5243 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5244 			   enum addr_type_t type)
5245 {
5246 	const struct nlmsghdr *nlh = cb->nlh;
5247 	struct inet6_fill_args fillargs = {
5248 		.portid = NETLINK_CB(cb->skb).portid,
5249 		.seq = cb->nlh->nlmsg_seq,
5250 		.flags = NLM_F_MULTI,
5251 		.netnsid = -1,
5252 		.type = type,
5253 	};
5254 	struct net *tgt_net = sock_net(skb->sk);
5255 	int idx, s_idx, s_ip_idx;
5256 	int h, s_h;
5257 	struct net_device *dev;
5258 	struct inet6_dev *idev;
5259 	struct hlist_head *head;
5260 	int err = 0;
5261 
5262 	s_h = cb->args[0];
5263 	s_idx = idx = cb->args[1];
5264 	s_ip_idx = cb->args[2];
5265 
5266 	if (cb->strict_check) {
5267 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5268 						  skb->sk, cb);
5269 		if (err < 0)
5270 			goto put_tgt_net;
5271 
5272 		err = 0;
5273 		if (fillargs.ifindex) {
5274 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5275 			if (!dev) {
5276 				err = -ENODEV;
5277 				goto put_tgt_net;
5278 			}
5279 			idev = __in6_dev_get(dev);
5280 			if (idev) {
5281 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5282 						     &fillargs);
5283 				if (err > 0)
5284 					err = 0;
5285 			}
5286 			goto put_tgt_net;
5287 		}
5288 	}
5289 
5290 	rcu_read_lock();
5291 	cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5292 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5293 		idx = 0;
5294 		head = &tgt_net->dev_index_head[h];
5295 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5296 			if (idx < s_idx)
5297 				goto cont;
5298 			if (h > s_h || idx > s_idx)
5299 				s_ip_idx = 0;
5300 			idev = __in6_dev_get(dev);
5301 			if (!idev)
5302 				goto cont;
5303 
5304 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5305 					   &fillargs) < 0)
5306 				goto done;
5307 cont:
5308 			idx++;
5309 		}
5310 	}
5311 done:
5312 	rcu_read_unlock();
5313 	cb->args[0] = h;
5314 	cb->args[1] = idx;
5315 put_tgt_net:
5316 	if (fillargs.netnsid >= 0)
5317 		put_net(tgt_net);
5318 
5319 	return skb->len ? : err;
5320 }
5321 
5322 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5323 {
5324 	enum addr_type_t type = UNICAST_ADDR;
5325 
5326 	return inet6_dump_addr(skb, cb, type);
5327 }
5328 
5329 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5330 {
5331 	enum addr_type_t type = MULTICAST_ADDR;
5332 
5333 	return inet6_dump_addr(skb, cb, type);
5334 }
5335 
5336 
5337 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5338 {
5339 	enum addr_type_t type = ANYCAST_ADDR;
5340 
5341 	return inet6_dump_addr(skb, cb, type);
5342 }
5343 
5344 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5345 				       const struct nlmsghdr *nlh,
5346 				       struct nlattr **tb,
5347 				       struct netlink_ext_ack *extack)
5348 {
5349 	struct ifaddrmsg *ifm;
5350 	int i, err;
5351 
5352 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5353 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5354 		return -EINVAL;
5355 	}
5356 
5357 	if (!netlink_strict_get_check(skb))
5358 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5359 					      ifa_ipv6_policy, extack);
5360 
5361 	ifm = nlmsg_data(nlh);
5362 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5363 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5364 		return -EINVAL;
5365 	}
5366 
5367 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5368 					    ifa_ipv6_policy, extack);
5369 	if (err)
5370 		return err;
5371 
5372 	for (i = 0; i <= IFA_MAX; i++) {
5373 		if (!tb[i])
5374 			continue;
5375 
5376 		switch (i) {
5377 		case IFA_TARGET_NETNSID:
5378 		case IFA_ADDRESS:
5379 		case IFA_LOCAL:
5380 			break;
5381 		default:
5382 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5383 			return -EINVAL;
5384 		}
5385 	}
5386 
5387 	return 0;
5388 }
5389 
5390 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5391 			     struct netlink_ext_ack *extack)
5392 {
5393 	struct net *tgt_net = sock_net(in_skb->sk);
5394 	struct inet6_fill_args fillargs = {
5395 		.portid = NETLINK_CB(in_skb).portid,
5396 		.seq = nlh->nlmsg_seq,
5397 		.event = RTM_NEWADDR,
5398 		.flags = 0,
5399 		.netnsid = -1,
5400 	};
5401 	struct ifaddrmsg *ifm;
5402 	struct nlattr *tb[IFA_MAX+1];
5403 	struct in6_addr *addr = NULL, *peer;
5404 	struct net_device *dev = NULL;
5405 	struct inet6_ifaddr *ifa;
5406 	struct sk_buff *skb;
5407 	int err;
5408 
5409 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5410 	if (err < 0)
5411 		return err;
5412 
5413 	if (tb[IFA_TARGET_NETNSID]) {
5414 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5415 
5416 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5417 						  fillargs.netnsid);
5418 		if (IS_ERR(tgt_net))
5419 			return PTR_ERR(tgt_net);
5420 	}
5421 
5422 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5423 	if (!addr)
5424 		return -EINVAL;
5425 
5426 	ifm = nlmsg_data(nlh);
5427 	if (ifm->ifa_index)
5428 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5429 
5430 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5431 	if (!ifa) {
5432 		err = -EADDRNOTAVAIL;
5433 		goto errout;
5434 	}
5435 
5436 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5437 	if (!skb) {
5438 		err = -ENOBUFS;
5439 		goto errout_ifa;
5440 	}
5441 
5442 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5443 	if (err < 0) {
5444 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5445 		WARN_ON(err == -EMSGSIZE);
5446 		kfree_skb(skb);
5447 		goto errout_ifa;
5448 	}
5449 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5450 errout_ifa:
5451 	in6_ifa_put(ifa);
5452 errout:
5453 	dev_put(dev);
5454 	if (fillargs.netnsid >= 0)
5455 		put_net(tgt_net);
5456 
5457 	return err;
5458 }
5459 
5460 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5461 {
5462 	struct sk_buff *skb;
5463 	struct net *net = dev_net(ifa->idev->dev);
5464 	struct inet6_fill_args fillargs = {
5465 		.portid = 0,
5466 		.seq = 0,
5467 		.event = event,
5468 		.flags = 0,
5469 		.netnsid = -1,
5470 	};
5471 	int err = -ENOBUFS;
5472 
5473 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5474 	if (!skb)
5475 		goto errout;
5476 
5477 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5478 	if (err < 0) {
5479 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5480 		WARN_ON(err == -EMSGSIZE);
5481 		kfree_skb(skb);
5482 		goto errout;
5483 	}
5484 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5485 	return;
5486 errout:
5487 	if (err < 0)
5488 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5489 }
5490 
5491 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5492 				__s32 *array, int bytes)
5493 {
5494 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5495 
5496 	memset(array, 0, bytes);
5497 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5498 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5499 	array[DEVCONF_MTU6] = cnf->mtu6;
5500 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5501 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5502 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5503 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5504 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5505 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5506 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5507 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5508 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5509 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5510 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5511 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5512 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5513 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5514 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5515 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5516 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5517 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5518 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5519 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5520 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5521 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5522 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5523 	array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5524 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5525 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5526 #ifdef CONFIG_IPV6_ROUTER_PREF
5527 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5528 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5529 		jiffies_to_msecs(cnf->rtr_probe_interval);
5530 #ifdef CONFIG_IPV6_ROUTE_INFO
5531 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5532 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5533 #endif
5534 #endif
5535 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5536 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5537 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5538 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5539 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5540 #endif
5541 #ifdef CONFIG_IPV6_MROUTE
5542 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5543 #endif
5544 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5545 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5546 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5547 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5548 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5549 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5550 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5551 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5552 	/* we omit DEVCONF_STABLE_SECRET for now */
5553 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5554 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5555 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5556 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5557 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5558 #ifdef CONFIG_IPV6_SEG6_HMAC
5559 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5560 #endif
5561 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5562 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5563 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5564 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5565 	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5566 	array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5567 	array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5568 	array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5569 	array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5570 }
5571 
5572 static inline size_t inet6_ifla6_size(void)
5573 {
5574 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5575 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5576 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5577 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5578 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5579 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5580 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5581 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5582 	     + 0;
5583 }
5584 
5585 static inline size_t inet6_if_nlmsg_size(void)
5586 {
5587 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5588 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5589 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5590 	       + nla_total_size(4) /* IFLA_MTU */
5591 	       + nla_total_size(4) /* IFLA_LINK */
5592 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5593 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5594 }
5595 
5596 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5597 					int bytes)
5598 {
5599 	int i;
5600 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5601 	BUG_ON(pad < 0);
5602 
5603 	/* Use put_unaligned() because stats may not be aligned for u64. */
5604 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5605 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5606 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5607 
5608 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5609 }
5610 
5611 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5612 					int bytes, size_t syncpoff)
5613 {
5614 	int i, c;
5615 	u64 buff[IPSTATS_MIB_MAX];
5616 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5617 
5618 	BUG_ON(pad < 0);
5619 
5620 	memset(buff, 0, sizeof(buff));
5621 	buff[0] = IPSTATS_MIB_MAX;
5622 
5623 	for_each_possible_cpu(c) {
5624 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5625 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5626 	}
5627 
5628 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5629 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5630 }
5631 
5632 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5633 			     int bytes)
5634 {
5635 	switch (attrtype) {
5636 	case IFLA_INET6_STATS:
5637 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5638 				     offsetof(struct ipstats_mib, syncp));
5639 		break;
5640 	case IFLA_INET6_ICMP6STATS:
5641 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5642 		break;
5643 	}
5644 }
5645 
5646 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5647 				  u32 ext_filter_mask)
5648 {
5649 	struct nlattr *nla;
5650 	struct ifla_cacheinfo ci;
5651 
5652 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5653 		goto nla_put_failure;
5654 	ci.max_reasm_len = IPV6_MAXPLEN;
5655 	ci.tstamp = cstamp_delta(idev->tstamp);
5656 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5657 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5658 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5659 		goto nla_put_failure;
5660 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5661 	if (!nla)
5662 		goto nla_put_failure;
5663 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5664 
5665 	/* XXX - MC not implemented */
5666 
5667 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5668 		return 0;
5669 
5670 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5671 	if (!nla)
5672 		goto nla_put_failure;
5673 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5674 
5675 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5676 	if (!nla)
5677 		goto nla_put_failure;
5678 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5679 
5680 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5681 	if (!nla)
5682 		goto nla_put_failure;
5683 	read_lock_bh(&idev->lock);
5684 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5685 	read_unlock_bh(&idev->lock);
5686 
5687 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5688 		goto nla_put_failure;
5689 
5690 	if (idev->ra_mtu &&
5691 	    nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5692 		goto nla_put_failure;
5693 
5694 	return 0;
5695 
5696 nla_put_failure:
5697 	return -EMSGSIZE;
5698 }
5699 
5700 static size_t inet6_get_link_af_size(const struct net_device *dev,
5701 				     u32 ext_filter_mask)
5702 {
5703 	if (!__in6_dev_get(dev))
5704 		return 0;
5705 
5706 	return inet6_ifla6_size();
5707 }
5708 
5709 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5710 			      u32 ext_filter_mask)
5711 {
5712 	struct inet6_dev *idev = __in6_dev_get(dev);
5713 
5714 	if (!idev)
5715 		return -ENODATA;
5716 
5717 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5718 		return -EMSGSIZE;
5719 
5720 	return 0;
5721 }
5722 
5723 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5724 			     struct netlink_ext_ack *extack)
5725 {
5726 	struct inet6_ifaddr *ifp;
5727 	struct net_device *dev = idev->dev;
5728 	bool clear_token, update_rs = false;
5729 	struct in6_addr ll_addr;
5730 
5731 	ASSERT_RTNL();
5732 
5733 	if (!token)
5734 		return -EINVAL;
5735 
5736 	if (dev->flags & IFF_LOOPBACK) {
5737 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5738 		return -EINVAL;
5739 	}
5740 
5741 	if (dev->flags & IFF_NOARP) {
5742 		NL_SET_ERR_MSG_MOD(extack,
5743 				   "Device does not do neighbour discovery");
5744 		return -EINVAL;
5745 	}
5746 
5747 	if (!ipv6_accept_ra(idev)) {
5748 		NL_SET_ERR_MSG_MOD(extack,
5749 				   "Router advertisement is disabled on device");
5750 		return -EINVAL;
5751 	}
5752 
5753 	if (idev->cnf.rtr_solicits == 0) {
5754 		NL_SET_ERR_MSG(extack,
5755 			       "Router solicitation is disabled on device");
5756 		return -EINVAL;
5757 	}
5758 
5759 	write_lock_bh(&idev->lock);
5760 
5761 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5762 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5763 
5764 	write_unlock_bh(&idev->lock);
5765 
5766 	clear_token = ipv6_addr_any(token);
5767 	if (clear_token)
5768 		goto update_lft;
5769 
5770 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5771 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5772 			     IFA_F_OPTIMISTIC)) {
5773 		/* If we're not ready, then normal ifup will take care
5774 		 * of this. Otherwise, we need to request our rs here.
5775 		 */
5776 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5777 		update_rs = true;
5778 	}
5779 
5780 update_lft:
5781 	write_lock_bh(&idev->lock);
5782 
5783 	if (update_rs) {
5784 		idev->if_flags |= IF_RS_SENT;
5785 		idev->rs_interval = rfc3315_s14_backoff_init(
5786 			idev->cnf.rtr_solicit_interval);
5787 		idev->rs_probes = 1;
5788 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5789 	}
5790 
5791 	/* Well, that's kinda nasty ... */
5792 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5793 		spin_lock(&ifp->lock);
5794 		if (ifp->tokenized) {
5795 			ifp->valid_lft = 0;
5796 			ifp->prefered_lft = 0;
5797 		}
5798 		spin_unlock(&ifp->lock);
5799 	}
5800 
5801 	write_unlock_bh(&idev->lock);
5802 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5803 	addrconf_verify_rtnl(dev_net(dev));
5804 	return 0;
5805 }
5806 
5807 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5808 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5809 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5810 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5811 					    .reject_message =
5812 						"IFLA_INET6_RA_MTU can not be set" },
5813 };
5814 
5815 static int check_addr_gen_mode(int mode)
5816 {
5817 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5818 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5819 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5820 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5821 		return -EINVAL;
5822 	return 1;
5823 }
5824 
5825 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5826 				int mode)
5827 {
5828 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5829 	    !idev->cnf.stable_secret.initialized &&
5830 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5831 		return -EINVAL;
5832 	return 1;
5833 }
5834 
5835 static int inet6_validate_link_af(const struct net_device *dev,
5836 				  const struct nlattr *nla,
5837 				  struct netlink_ext_ack *extack)
5838 {
5839 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5840 	struct inet6_dev *idev = NULL;
5841 	int err;
5842 
5843 	if (dev) {
5844 		idev = __in6_dev_get(dev);
5845 		if (!idev)
5846 			return -EAFNOSUPPORT;
5847 	}
5848 
5849 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5850 					  inet6_af_policy, extack);
5851 	if (err)
5852 		return err;
5853 
5854 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5855 		return -EINVAL;
5856 
5857 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5858 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5859 
5860 		if (check_addr_gen_mode(mode) < 0)
5861 			return -EINVAL;
5862 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5863 			return -EINVAL;
5864 	}
5865 
5866 	return 0;
5867 }
5868 
5869 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5870 			     struct netlink_ext_ack *extack)
5871 {
5872 	struct inet6_dev *idev = __in6_dev_get(dev);
5873 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5874 	int err;
5875 
5876 	if (!idev)
5877 		return -EAFNOSUPPORT;
5878 
5879 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5880 		return -EINVAL;
5881 
5882 	if (tb[IFLA_INET6_TOKEN]) {
5883 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5884 					extack);
5885 		if (err)
5886 			return err;
5887 	}
5888 
5889 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5890 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5891 
5892 		idev->cnf.addr_gen_mode = mode;
5893 	}
5894 
5895 	return 0;
5896 }
5897 
5898 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5899 			     u32 portid, u32 seq, int event, unsigned int flags)
5900 {
5901 	struct net_device *dev = idev->dev;
5902 	struct ifinfomsg *hdr;
5903 	struct nlmsghdr *nlh;
5904 	void *protoinfo;
5905 
5906 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5907 	if (!nlh)
5908 		return -EMSGSIZE;
5909 
5910 	hdr = nlmsg_data(nlh);
5911 	hdr->ifi_family = AF_INET6;
5912 	hdr->__ifi_pad = 0;
5913 	hdr->ifi_type = dev->type;
5914 	hdr->ifi_index = dev->ifindex;
5915 	hdr->ifi_flags = dev_get_flags(dev);
5916 	hdr->ifi_change = 0;
5917 
5918 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5919 	    (dev->addr_len &&
5920 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5921 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5922 	    (dev->ifindex != dev_get_iflink(dev) &&
5923 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5924 	    nla_put_u8(skb, IFLA_OPERSTATE,
5925 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5926 		goto nla_put_failure;
5927 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5928 	if (!protoinfo)
5929 		goto nla_put_failure;
5930 
5931 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5932 		goto nla_put_failure;
5933 
5934 	nla_nest_end(skb, protoinfo);
5935 	nlmsg_end(skb, nlh);
5936 	return 0;
5937 
5938 nla_put_failure:
5939 	nlmsg_cancel(skb, nlh);
5940 	return -EMSGSIZE;
5941 }
5942 
5943 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5944 				   struct netlink_ext_ack *extack)
5945 {
5946 	struct ifinfomsg *ifm;
5947 
5948 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5949 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5950 		return -EINVAL;
5951 	}
5952 
5953 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5954 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5955 		return -EINVAL;
5956 	}
5957 
5958 	ifm = nlmsg_data(nlh);
5959 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5960 	    ifm->ifi_change || ifm->ifi_index) {
5961 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5962 		return -EINVAL;
5963 	}
5964 
5965 	return 0;
5966 }
5967 
5968 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5969 {
5970 	struct net *net = sock_net(skb->sk);
5971 	int h, s_h;
5972 	int idx = 0, s_idx;
5973 	struct net_device *dev;
5974 	struct inet6_dev *idev;
5975 	struct hlist_head *head;
5976 
5977 	/* only requests using strict checking can pass data to
5978 	 * influence the dump
5979 	 */
5980 	if (cb->strict_check) {
5981 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5982 
5983 		if (err < 0)
5984 			return err;
5985 	}
5986 
5987 	s_h = cb->args[0];
5988 	s_idx = cb->args[1];
5989 
5990 	rcu_read_lock();
5991 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5992 		idx = 0;
5993 		head = &net->dev_index_head[h];
5994 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5995 			if (idx < s_idx)
5996 				goto cont;
5997 			idev = __in6_dev_get(dev);
5998 			if (!idev)
5999 				goto cont;
6000 			if (inet6_fill_ifinfo(skb, idev,
6001 					      NETLINK_CB(cb->skb).portid,
6002 					      cb->nlh->nlmsg_seq,
6003 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
6004 				goto out;
6005 cont:
6006 			idx++;
6007 		}
6008 	}
6009 out:
6010 	rcu_read_unlock();
6011 	cb->args[1] = idx;
6012 	cb->args[0] = h;
6013 
6014 	return skb->len;
6015 }
6016 
6017 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6018 {
6019 	struct sk_buff *skb;
6020 	struct net *net = dev_net(idev->dev);
6021 	int err = -ENOBUFS;
6022 
6023 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6024 	if (!skb)
6025 		goto errout;
6026 
6027 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6028 	if (err < 0) {
6029 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6030 		WARN_ON(err == -EMSGSIZE);
6031 		kfree_skb(skb);
6032 		goto errout;
6033 	}
6034 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6035 	return;
6036 errout:
6037 	if (err < 0)
6038 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6039 }
6040 
6041 static inline size_t inet6_prefix_nlmsg_size(void)
6042 {
6043 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6044 	       + nla_total_size(sizeof(struct in6_addr))
6045 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6046 }
6047 
6048 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6049 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6050 			     int event, unsigned int flags)
6051 {
6052 	struct prefixmsg *pmsg;
6053 	struct nlmsghdr *nlh;
6054 	struct prefix_cacheinfo	ci;
6055 
6056 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6057 	if (!nlh)
6058 		return -EMSGSIZE;
6059 
6060 	pmsg = nlmsg_data(nlh);
6061 	pmsg->prefix_family = AF_INET6;
6062 	pmsg->prefix_pad1 = 0;
6063 	pmsg->prefix_pad2 = 0;
6064 	pmsg->prefix_ifindex = idev->dev->ifindex;
6065 	pmsg->prefix_len = pinfo->prefix_len;
6066 	pmsg->prefix_type = pinfo->type;
6067 	pmsg->prefix_pad3 = 0;
6068 	pmsg->prefix_flags = 0;
6069 	if (pinfo->onlink)
6070 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6071 	if (pinfo->autoconf)
6072 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6073 
6074 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6075 		goto nla_put_failure;
6076 	ci.preferred_time = ntohl(pinfo->prefered);
6077 	ci.valid_time = ntohl(pinfo->valid);
6078 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6079 		goto nla_put_failure;
6080 	nlmsg_end(skb, nlh);
6081 	return 0;
6082 
6083 nla_put_failure:
6084 	nlmsg_cancel(skb, nlh);
6085 	return -EMSGSIZE;
6086 }
6087 
6088 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6089 			 struct prefix_info *pinfo)
6090 {
6091 	struct sk_buff *skb;
6092 	struct net *net = dev_net(idev->dev);
6093 	int err = -ENOBUFS;
6094 
6095 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6096 	if (!skb)
6097 		goto errout;
6098 
6099 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6100 	if (err < 0) {
6101 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6102 		WARN_ON(err == -EMSGSIZE);
6103 		kfree_skb(skb);
6104 		goto errout;
6105 	}
6106 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6107 	return;
6108 errout:
6109 	if (err < 0)
6110 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6111 }
6112 
6113 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6114 {
6115 	struct net *net = dev_net(ifp->idev->dev);
6116 
6117 	if (event)
6118 		ASSERT_RTNL();
6119 
6120 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6121 
6122 	switch (event) {
6123 	case RTM_NEWADDR:
6124 		/*
6125 		 * If the address was optimistic we inserted the route at the
6126 		 * start of our DAD process, so we don't need to do it again.
6127 		 * If the device was taken down in the middle of the DAD
6128 		 * cycle there is a race where we could get here without a
6129 		 * host route, so nothing to insert. That will be fixed when
6130 		 * the device is brought up.
6131 		 */
6132 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6133 			ip6_ins_rt(net, ifp->rt);
6134 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6135 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6136 				&ifp->addr, ifp->idev->dev->name);
6137 		}
6138 
6139 		if (ifp->idev->cnf.forwarding)
6140 			addrconf_join_anycast(ifp);
6141 		if (!ipv6_addr_any(&ifp->peer_addr))
6142 			addrconf_prefix_route(&ifp->peer_addr, 128,
6143 					      ifp->rt_priority, ifp->idev->dev,
6144 					      0, 0, GFP_ATOMIC);
6145 		break;
6146 	case RTM_DELADDR:
6147 		if (ifp->idev->cnf.forwarding)
6148 			addrconf_leave_anycast(ifp);
6149 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6150 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6151 			struct fib6_info *rt;
6152 
6153 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6154 						       ifp->idev->dev, 0, 0,
6155 						       false);
6156 			if (rt)
6157 				ip6_del_rt(net, rt, false);
6158 		}
6159 		if (ifp->rt) {
6160 			ip6_del_rt(net, ifp->rt, false);
6161 			ifp->rt = NULL;
6162 		}
6163 		rt_genid_bump_ipv6(net);
6164 		break;
6165 	}
6166 	atomic_inc(&net->ipv6.dev_addr_genid);
6167 }
6168 
6169 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6170 {
6171 	if (likely(ifp->idev->dead == 0))
6172 		__ipv6_ifa_notify(event, ifp);
6173 }
6174 
6175 #ifdef CONFIG_SYSCTL
6176 
6177 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6178 		void *buffer, size_t *lenp, loff_t *ppos)
6179 {
6180 	int *valp = ctl->data;
6181 	int val = *valp;
6182 	loff_t pos = *ppos;
6183 	struct ctl_table lctl;
6184 	int ret;
6185 
6186 	/*
6187 	 * ctl->data points to idev->cnf.forwarding, we should
6188 	 * not modify it until we get the rtnl lock.
6189 	 */
6190 	lctl = *ctl;
6191 	lctl.data = &val;
6192 
6193 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6194 
6195 	if (write)
6196 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6197 	if (ret)
6198 		*ppos = pos;
6199 	return ret;
6200 }
6201 
6202 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6203 		void *buffer, size_t *lenp, loff_t *ppos)
6204 {
6205 	struct inet6_dev *idev = ctl->extra1;
6206 	int min_mtu = IPV6_MIN_MTU;
6207 	struct ctl_table lctl;
6208 
6209 	lctl = *ctl;
6210 	lctl.extra1 = &min_mtu;
6211 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6212 
6213 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6214 }
6215 
6216 static void dev_disable_change(struct inet6_dev *idev)
6217 {
6218 	struct netdev_notifier_info info;
6219 
6220 	if (!idev || !idev->dev)
6221 		return;
6222 
6223 	netdev_notifier_info_init(&info, idev->dev);
6224 	if (idev->cnf.disable_ipv6)
6225 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6226 	else
6227 		addrconf_notify(NULL, NETDEV_UP, &info);
6228 }
6229 
6230 static void addrconf_disable_change(struct net *net, __s32 newf)
6231 {
6232 	struct net_device *dev;
6233 	struct inet6_dev *idev;
6234 
6235 	for_each_netdev(net, dev) {
6236 		idev = __in6_dev_get(dev);
6237 		if (idev) {
6238 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6239 			idev->cnf.disable_ipv6 = newf;
6240 			if (changed)
6241 				dev_disable_change(idev);
6242 		}
6243 	}
6244 }
6245 
6246 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6247 {
6248 	struct net *net;
6249 	int old;
6250 
6251 	if (!rtnl_trylock())
6252 		return restart_syscall();
6253 
6254 	net = (struct net *)table->extra2;
6255 	old = *p;
6256 	*p = newf;
6257 
6258 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6259 		rtnl_unlock();
6260 		return 0;
6261 	}
6262 
6263 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6264 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
6265 		addrconf_disable_change(net, newf);
6266 	} else if ((!newf) ^ (!old))
6267 		dev_disable_change((struct inet6_dev *)table->extra1);
6268 
6269 	rtnl_unlock();
6270 	return 0;
6271 }
6272 
6273 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6274 		void *buffer, size_t *lenp, loff_t *ppos)
6275 {
6276 	int *valp = ctl->data;
6277 	int val = *valp;
6278 	loff_t pos = *ppos;
6279 	struct ctl_table lctl;
6280 	int ret;
6281 
6282 	/*
6283 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6284 	 * not modify it until we get the rtnl lock.
6285 	 */
6286 	lctl = *ctl;
6287 	lctl.data = &val;
6288 
6289 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6290 
6291 	if (write)
6292 		ret = addrconf_disable_ipv6(ctl, valp, val);
6293 	if (ret)
6294 		*ppos = pos;
6295 	return ret;
6296 }
6297 
6298 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6299 		void *buffer, size_t *lenp, loff_t *ppos)
6300 {
6301 	int *valp = ctl->data;
6302 	int ret;
6303 	int old, new;
6304 
6305 	old = *valp;
6306 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6307 	new = *valp;
6308 
6309 	if (write && old != new) {
6310 		struct net *net = ctl->extra2;
6311 
6312 		if (!rtnl_trylock())
6313 			return restart_syscall();
6314 
6315 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6316 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6317 						     NETCONFA_PROXY_NEIGH,
6318 						     NETCONFA_IFINDEX_DEFAULT,
6319 						     net->ipv6.devconf_dflt);
6320 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6321 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6322 						     NETCONFA_PROXY_NEIGH,
6323 						     NETCONFA_IFINDEX_ALL,
6324 						     net->ipv6.devconf_all);
6325 		else {
6326 			struct inet6_dev *idev = ctl->extra1;
6327 
6328 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6329 						     NETCONFA_PROXY_NEIGH,
6330 						     idev->dev->ifindex,
6331 						     &idev->cnf);
6332 		}
6333 		rtnl_unlock();
6334 	}
6335 
6336 	return ret;
6337 }
6338 
6339 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6340 					 void *buffer, size_t *lenp,
6341 					 loff_t *ppos)
6342 {
6343 	int ret = 0;
6344 	u32 new_val;
6345 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6346 	struct net *net = (struct net *)ctl->extra2;
6347 	struct ctl_table tmp = {
6348 		.data = &new_val,
6349 		.maxlen = sizeof(new_val),
6350 		.mode = ctl->mode,
6351 	};
6352 
6353 	if (!rtnl_trylock())
6354 		return restart_syscall();
6355 
6356 	new_val = *((u32 *)ctl->data);
6357 
6358 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6359 	if (ret != 0)
6360 		goto out;
6361 
6362 	if (write) {
6363 		if (check_addr_gen_mode(new_val) < 0) {
6364 			ret = -EINVAL;
6365 			goto out;
6366 		}
6367 
6368 		if (idev) {
6369 			if (check_stable_privacy(idev, net, new_val) < 0) {
6370 				ret = -EINVAL;
6371 				goto out;
6372 			}
6373 
6374 			if (idev->cnf.addr_gen_mode != new_val) {
6375 				idev->cnf.addr_gen_mode = new_val;
6376 				addrconf_dev_config(idev->dev);
6377 			}
6378 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6379 			struct net_device *dev;
6380 
6381 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6382 			for_each_netdev(net, dev) {
6383 				idev = __in6_dev_get(dev);
6384 				if (idev &&
6385 				    idev->cnf.addr_gen_mode != new_val) {
6386 					idev->cnf.addr_gen_mode = new_val;
6387 					addrconf_dev_config(idev->dev);
6388 				}
6389 			}
6390 		}
6391 
6392 		*((u32 *)ctl->data) = new_val;
6393 	}
6394 
6395 out:
6396 	rtnl_unlock();
6397 
6398 	return ret;
6399 }
6400 
6401 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6402 					 void *buffer, size_t *lenp,
6403 					 loff_t *ppos)
6404 {
6405 	int err;
6406 	struct in6_addr addr;
6407 	char str[IPV6_MAX_STRLEN];
6408 	struct ctl_table lctl = *ctl;
6409 	struct net *net = ctl->extra2;
6410 	struct ipv6_stable_secret *secret = ctl->data;
6411 
6412 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6413 		return -EIO;
6414 
6415 	lctl.maxlen = IPV6_MAX_STRLEN;
6416 	lctl.data = str;
6417 
6418 	if (!rtnl_trylock())
6419 		return restart_syscall();
6420 
6421 	if (!write && !secret->initialized) {
6422 		err = -EIO;
6423 		goto out;
6424 	}
6425 
6426 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6427 	if (err >= sizeof(str)) {
6428 		err = -EIO;
6429 		goto out;
6430 	}
6431 
6432 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6433 	if (err || !write)
6434 		goto out;
6435 
6436 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6437 		err = -EIO;
6438 		goto out;
6439 	}
6440 
6441 	secret->initialized = true;
6442 	secret->secret = addr;
6443 
6444 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6445 		struct net_device *dev;
6446 
6447 		for_each_netdev(net, dev) {
6448 			struct inet6_dev *idev = __in6_dev_get(dev);
6449 
6450 			if (idev) {
6451 				idev->cnf.addr_gen_mode =
6452 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6453 			}
6454 		}
6455 	} else {
6456 		struct inet6_dev *idev = ctl->extra1;
6457 
6458 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6459 	}
6460 
6461 out:
6462 	rtnl_unlock();
6463 
6464 	return err;
6465 }
6466 
6467 static
6468 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6469 						int write, void *buffer,
6470 						size_t *lenp,
6471 						loff_t *ppos)
6472 {
6473 	int *valp = ctl->data;
6474 	int val = *valp;
6475 	loff_t pos = *ppos;
6476 	struct ctl_table lctl;
6477 	int ret;
6478 
6479 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6480 	 * we should not modify it until we get the rtnl lock.
6481 	 */
6482 	lctl = *ctl;
6483 	lctl.data = &val;
6484 
6485 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6486 
6487 	if (write)
6488 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6489 	if (ret)
6490 		*ppos = pos;
6491 	return ret;
6492 }
6493 
6494 static
6495 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6496 {
6497 	if (rt) {
6498 		if (action)
6499 			rt->dst.flags |= DST_NOPOLICY;
6500 		else
6501 			rt->dst.flags &= ~DST_NOPOLICY;
6502 	}
6503 }
6504 
6505 static
6506 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6507 {
6508 	struct inet6_ifaddr *ifa;
6509 
6510 	read_lock_bh(&idev->lock);
6511 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6512 		spin_lock(&ifa->lock);
6513 		if (ifa->rt) {
6514 			/* host routes only use builtin fib6_nh */
6515 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6516 			int cpu;
6517 
6518 			rcu_read_lock();
6519 			ifa->rt->dst_nopolicy = val ? true : false;
6520 			if (nh->rt6i_pcpu) {
6521 				for_each_possible_cpu(cpu) {
6522 					struct rt6_info **rtp;
6523 
6524 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6525 					addrconf_set_nopolicy(*rtp, val);
6526 				}
6527 			}
6528 			rcu_read_unlock();
6529 		}
6530 		spin_unlock(&ifa->lock);
6531 	}
6532 	read_unlock_bh(&idev->lock);
6533 }
6534 
6535 static
6536 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6537 {
6538 	struct inet6_dev *idev;
6539 	struct net *net;
6540 
6541 	if (!rtnl_trylock())
6542 		return restart_syscall();
6543 
6544 	*valp = val;
6545 
6546 	net = (struct net *)ctl->extra2;
6547 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6548 		rtnl_unlock();
6549 		return 0;
6550 	}
6551 
6552 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6553 		struct net_device *dev;
6554 
6555 		for_each_netdev(net, dev) {
6556 			idev = __in6_dev_get(dev);
6557 			if (idev)
6558 				addrconf_disable_policy_idev(idev, val);
6559 		}
6560 	} else {
6561 		idev = (struct inet6_dev *)ctl->extra1;
6562 		addrconf_disable_policy_idev(idev, val);
6563 	}
6564 
6565 	rtnl_unlock();
6566 	return 0;
6567 }
6568 
6569 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6570 				   void *buffer, size_t *lenp, loff_t *ppos)
6571 {
6572 	int *valp = ctl->data;
6573 	int val = *valp;
6574 	loff_t pos = *ppos;
6575 	struct ctl_table lctl;
6576 	int ret;
6577 
6578 	lctl = *ctl;
6579 	lctl.data = &val;
6580 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6581 
6582 	if (write && (*valp != val))
6583 		ret = addrconf_disable_policy(ctl, valp, val);
6584 
6585 	if (ret)
6586 		*ppos = pos;
6587 
6588 	return ret;
6589 }
6590 
6591 static int minus_one = -1;
6592 static const int two_five_five = 255;
6593 static u32 ioam6_if_id_max = U16_MAX;
6594 
6595 static const struct ctl_table addrconf_sysctl[] = {
6596 	{
6597 		.procname	= "forwarding",
6598 		.data		= &ipv6_devconf.forwarding,
6599 		.maxlen		= sizeof(int),
6600 		.mode		= 0644,
6601 		.proc_handler	= addrconf_sysctl_forward,
6602 	},
6603 	{
6604 		.procname	= "hop_limit",
6605 		.data		= &ipv6_devconf.hop_limit,
6606 		.maxlen		= sizeof(int),
6607 		.mode		= 0644,
6608 		.proc_handler	= proc_dointvec_minmax,
6609 		.extra1		= (void *)SYSCTL_ONE,
6610 		.extra2		= (void *)&two_five_five,
6611 	},
6612 	{
6613 		.procname	= "mtu",
6614 		.data		= &ipv6_devconf.mtu6,
6615 		.maxlen		= sizeof(int),
6616 		.mode		= 0644,
6617 		.proc_handler	= addrconf_sysctl_mtu,
6618 	},
6619 	{
6620 		.procname	= "accept_ra",
6621 		.data		= &ipv6_devconf.accept_ra,
6622 		.maxlen		= sizeof(int),
6623 		.mode		= 0644,
6624 		.proc_handler	= proc_dointvec,
6625 	},
6626 	{
6627 		.procname	= "accept_redirects",
6628 		.data		= &ipv6_devconf.accept_redirects,
6629 		.maxlen		= sizeof(int),
6630 		.mode		= 0644,
6631 		.proc_handler	= proc_dointvec,
6632 	},
6633 	{
6634 		.procname	= "autoconf",
6635 		.data		= &ipv6_devconf.autoconf,
6636 		.maxlen		= sizeof(int),
6637 		.mode		= 0644,
6638 		.proc_handler	= proc_dointvec,
6639 	},
6640 	{
6641 		.procname	= "dad_transmits",
6642 		.data		= &ipv6_devconf.dad_transmits,
6643 		.maxlen		= sizeof(int),
6644 		.mode		= 0644,
6645 		.proc_handler	= proc_dointvec,
6646 	},
6647 	{
6648 		.procname	= "router_solicitations",
6649 		.data		= &ipv6_devconf.rtr_solicits,
6650 		.maxlen		= sizeof(int),
6651 		.mode		= 0644,
6652 		.proc_handler	= proc_dointvec_minmax,
6653 		.extra1		= &minus_one,
6654 	},
6655 	{
6656 		.procname	= "router_solicitation_interval",
6657 		.data		= &ipv6_devconf.rtr_solicit_interval,
6658 		.maxlen		= sizeof(int),
6659 		.mode		= 0644,
6660 		.proc_handler	= proc_dointvec_jiffies,
6661 	},
6662 	{
6663 		.procname	= "router_solicitation_max_interval",
6664 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6665 		.maxlen		= sizeof(int),
6666 		.mode		= 0644,
6667 		.proc_handler	= proc_dointvec_jiffies,
6668 	},
6669 	{
6670 		.procname	= "router_solicitation_delay",
6671 		.data		= &ipv6_devconf.rtr_solicit_delay,
6672 		.maxlen		= sizeof(int),
6673 		.mode		= 0644,
6674 		.proc_handler	= proc_dointvec_jiffies,
6675 	},
6676 	{
6677 		.procname	= "force_mld_version",
6678 		.data		= &ipv6_devconf.force_mld_version,
6679 		.maxlen		= sizeof(int),
6680 		.mode		= 0644,
6681 		.proc_handler	= proc_dointvec,
6682 	},
6683 	{
6684 		.procname	= "mldv1_unsolicited_report_interval",
6685 		.data		=
6686 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6687 		.maxlen		= sizeof(int),
6688 		.mode		= 0644,
6689 		.proc_handler	= proc_dointvec_ms_jiffies,
6690 	},
6691 	{
6692 		.procname	= "mldv2_unsolicited_report_interval",
6693 		.data		=
6694 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6695 		.maxlen		= sizeof(int),
6696 		.mode		= 0644,
6697 		.proc_handler	= proc_dointvec_ms_jiffies,
6698 	},
6699 	{
6700 		.procname	= "use_tempaddr",
6701 		.data		= &ipv6_devconf.use_tempaddr,
6702 		.maxlen		= sizeof(int),
6703 		.mode		= 0644,
6704 		.proc_handler	= proc_dointvec,
6705 	},
6706 	{
6707 		.procname	= "temp_valid_lft",
6708 		.data		= &ipv6_devconf.temp_valid_lft,
6709 		.maxlen		= sizeof(int),
6710 		.mode		= 0644,
6711 		.proc_handler	= proc_dointvec,
6712 	},
6713 	{
6714 		.procname	= "temp_prefered_lft",
6715 		.data		= &ipv6_devconf.temp_prefered_lft,
6716 		.maxlen		= sizeof(int),
6717 		.mode		= 0644,
6718 		.proc_handler	= proc_dointvec,
6719 	},
6720 	{
6721 		.procname	= "regen_max_retry",
6722 		.data		= &ipv6_devconf.regen_max_retry,
6723 		.maxlen		= sizeof(int),
6724 		.mode		= 0644,
6725 		.proc_handler	= proc_dointvec,
6726 	},
6727 	{
6728 		.procname	= "max_desync_factor",
6729 		.data		= &ipv6_devconf.max_desync_factor,
6730 		.maxlen		= sizeof(int),
6731 		.mode		= 0644,
6732 		.proc_handler	= proc_dointvec,
6733 	},
6734 	{
6735 		.procname	= "max_addresses",
6736 		.data		= &ipv6_devconf.max_addresses,
6737 		.maxlen		= sizeof(int),
6738 		.mode		= 0644,
6739 		.proc_handler	= proc_dointvec,
6740 	},
6741 	{
6742 		.procname	= "accept_ra_defrtr",
6743 		.data		= &ipv6_devconf.accept_ra_defrtr,
6744 		.maxlen		= sizeof(int),
6745 		.mode		= 0644,
6746 		.proc_handler	= proc_dointvec,
6747 	},
6748 	{
6749 		.procname	= "ra_defrtr_metric",
6750 		.data		= &ipv6_devconf.ra_defrtr_metric,
6751 		.maxlen		= sizeof(u32),
6752 		.mode		= 0644,
6753 		.proc_handler	= proc_douintvec_minmax,
6754 		.extra1		= (void *)SYSCTL_ONE,
6755 	},
6756 	{
6757 		.procname	= "accept_ra_min_hop_limit",
6758 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6759 		.maxlen		= sizeof(int),
6760 		.mode		= 0644,
6761 		.proc_handler	= proc_dointvec,
6762 	},
6763 	{
6764 		.procname	= "accept_ra_pinfo",
6765 		.data		= &ipv6_devconf.accept_ra_pinfo,
6766 		.maxlen		= sizeof(int),
6767 		.mode		= 0644,
6768 		.proc_handler	= proc_dointvec,
6769 	},
6770 #ifdef CONFIG_IPV6_ROUTER_PREF
6771 	{
6772 		.procname	= "accept_ra_rtr_pref",
6773 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6774 		.maxlen		= sizeof(int),
6775 		.mode		= 0644,
6776 		.proc_handler	= proc_dointvec,
6777 	},
6778 	{
6779 		.procname	= "router_probe_interval",
6780 		.data		= &ipv6_devconf.rtr_probe_interval,
6781 		.maxlen		= sizeof(int),
6782 		.mode		= 0644,
6783 		.proc_handler	= proc_dointvec_jiffies,
6784 	},
6785 #ifdef CONFIG_IPV6_ROUTE_INFO
6786 	{
6787 		.procname	= "accept_ra_rt_info_min_plen",
6788 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6789 		.maxlen		= sizeof(int),
6790 		.mode		= 0644,
6791 		.proc_handler	= proc_dointvec,
6792 	},
6793 	{
6794 		.procname	= "accept_ra_rt_info_max_plen",
6795 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6796 		.maxlen		= sizeof(int),
6797 		.mode		= 0644,
6798 		.proc_handler	= proc_dointvec,
6799 	},
6800 #endif
6801 #endif
6802 	{
6803 		.procname	= "proxy_ndp",
6804 		.data		= &ipv6_devconf.proxy_ndp,
6805 		.maxlen		= sizeof(int),
6806 		.mode		= 0644,
6807 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6808 	},
6809 	{
6810 		.procname	= "accept_source_route",
6811 		.data		= &ipv6_devconf.accept_source_route,
6812 		.maxlen		= sizeof(int),
6813 		.mode		= 0644,
6814 		.proc_handler	= proc_dointvec,
6815 	},
6816 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6817 	{
6818 		.procname	= "optimistic_dad",
6819 		.data		= &ipv6_devconf.optimistic_dad,
6820 		.maxlen		= sizeof(int),
6821 		.mode		= 0644,
6822 		.proc_handler   = proc_dointvec,
6823 	},
6824 	{
6825 		.procname	= "use_optimistic",
6826 		.data		= &ipv6_devconf.use_optimistic,
6827 		.maxlen		= sizeof(int),
6828 		.mode		= 0644,
6829 		.proc_handler	= proc_dointvec,
6830 	},
6831 #endif
6832 #ifdef CONFIG_IPV6_MROUTE
6833 	{
6834 		.procname	= "mc_forwarding",
6835 		.data		= &ipv6_devconf.mc_forwarding,
6836 		.maxlen		= sizeof(int),
6837 		.mode		= 0444,
6838 		.proc_handler	= proc_dointvec,
6839 	},
6840 #endif
6841 	{
6842 		.procname	= "disable_ipv6",
6843 		.data		= &ipv6_devconf.disable_ipv6,
6844 		.maxlen		= sizeof(int),
6845 		.mode		= 0644,
6846 		.proc_handler	= addrconf_sysctl_disable,
6847 	},
6848 	{
6849 		.procname	= "accept_dad",
6850 		.data		= &ipv6_devconf.accept_dad,
6851 		.maxlen		= sizeof(int),
6852 		.mode		= 0644,
6853 		.proc_handler	= proc_dointvec,
6854 	},
6855 	{
6856 		.procname	= "force_tllao",
6857 		.data		= &ipv6_devconf.force_tllao,
6858 		.maxlen		= sizeof(int),
6859 		.mode		= 0644,
6860 		.proc_handler	= proc_dointvec
6861 	},
6862 	{
6863 		.procname	= "ndisc_notify",
6864 		.data		= &ipv6_devconf.ndisc_notify,
6865 		.maxlen		= sizeof(int),
6866 		.mode		= 0644,
6867 		.proc_handler	= proc_dointvec
6868 	},
6869 	{
6870 		.procname	= "suppress_frag_ndisc",
6871 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6872 		.maxlen		= sizeof(int),
6873 		.mode		= 0644,
6874 		.proc_handler	= proc_dointvec
6875 	},
6876 	{
6877 		.procname	= "accept_ra_from_local",
6878 		.data		= &ipv6_devconf.accept_ra_from_local,
6879 		.maxlen		= sizeof(int),
6880 		.mode		= 0644,
6881 		.proc_handler	= proc_dointvec,
6882 	},
6883 	{
6884 		.procname	= "accept_ra_mtu",
6885 		.data		= &ipv6_devconf.accept_ra_mtu,
6886 		.maxlen		= sizeof(int),
6887 		.mode		= 0644,
6888 		.proc_handler	= proc_dointvec,
6889 	},
6890 	{
6891 		.procname	= "stable_secret",
6892 		.data		= &ipv6_devconf.stable_secret,
6893 		.maxlen		= IPV6_MAX_STRLEN,
6894 		.mode		= 0600,
6895 		.proc_handler	= addrconf_sysctl_stable_secret,
6896 	},
6897 	{
6898 		.procname	= "use_oif_addrs_only",
6899 		.data		= &ipv6_devconf.use_oif_addrs_only,
6900 		.maxlen		= sizeof(int),
6901 		.mode		= 0644,
6902 		.proc_handler	= proc_dointvec,
6903 	},
6904 	{
6905 		.procname	= "ignore_routes_with_linkdown",
6906 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6907 		.maxlen		= sizeof(int),
6908 		.mode		= 0644,
6909 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6910 	},
6911 	{
6912 		.procname	= "drop_unicast_in_l2_multicast",
6913 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6914 		.maxlen		= sizeof(int),
6915 		.mode		= 0644,
6916 		.proc_handler	= proc_dointvec,
6917 	},
6918 	{
6919 		.procname	= "drop_unsolicited_na",
6920 		.data		= &ipv6_devconf.drop_unsolicited_na,
6921 		.maxlen		= sizeof(int),
6922 		.mode		= 0644,
6923 		.proc_handler	= proc_dointvec,
6924 	},
6925 	{
6926 		.procname	= "keep_addr_on_down",
6927 		.data		= &ipv6_devconf.keep_addr_on_down,
6928 		.maxlen		= sizeof(int),
6929 		.mode		= 0644,
6930 		.proc_handler	= proc_dointvec,
6931 
6932 	},
6933 	{
6934 		.procname	= "seg6_enabled",
6935 		.data		= &ipv6_devconf.seg6_enabled,
6936 		.maxlen		= sizeof(int),
6937 		.mode		= 0644,
6938 		.proc_handler	= proc_dointvec,
6939 	},
6940 #ifdef CONFIG_IPV6_SEG6_HMAC
6941 	{
6942 		.procname	= "seg6_require_hmac",
6943 		.data		= &ipv6_devconf.seg6_require_hmac,
6944 		.maxlen		= sizeof(int),
6945 		.mode		= 0644,
6946 		.proc_handler	= proc_dointvec,
6947 	},
6948 #endif
6949 	{
6950 		.procname       = "enhanced_dad",
6951 		.data           = &ipv6_devconf.enhanced_dad,
6952 		.maxlen         = sizeof(int),
6953 		.mode           = 0644,
6954 		.proc_handler   = proc_dointvec,
6955 	},
6956 	{
6957 		.procname	= "addr_gen_mode",
6958 		.data		= &ipv6_devconf.addr_gen_mode,
6959 		.maxlen		= sizeof(int),
6960 		.mode		= 0644,
6961 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6962 	},
6963 	{
6964 		.procname       = "disable_policy",
6965 		.data           = &ipv6_devconf.disable_policy,
6966 		.maxlen         = sizeof(int),
6967 		.mode           = 0644,
6968 		.proc_handler   = addrconf_sysctl_disable_policy,
6969 	},
6970 	{
6971 		.procname	= "ndisc_tclass",
6972 		.data		= &ipv6_devconf.ndisc_tclass,
6973 		.maxlen		= sizeof(int),
6974 		.mode		= 0644,
6975 		.proc_handler	= proc_dointvec_minmax,
6976 		.extra1		= (void *)SYSCTL_ZERO,
6977 		.extra2		= (void *)&two_five_five,
6978 	},
6979 	{
6980 		.procname	= "rpl_seg_enabled",
6981 		.data		= &ipv6_devconf.rpl_seg_enabled,
6982 		.maxlen		= sizeof(int),
6983 		.mode		= 0644,
6984 		.proc_handler	= proc_dointvec,
6985 	},
6986 	{
6987 		.procname	= "ioam6_enabled",
6988 		.data		= &ipv6_devconf.ioam6_enabled,
6989 		.maxlen		= sizeof(u8),
6990 		.mode		= 0644,
6991 		.proc_handler	= proc_dou8vec_minmax,
6992 		.extra1		= (void *)SYSCTL_ZERO,
6993 		.extra2		= (void *)SYSCTL_ONE,
6994 	},
6995 	{
6996 		.procname	= "ioam6_id",
6997 		.data		= &ipv6_devconf.ioam6_id,
6998 		.maxlen		= sizeof(u32),
6999 		.mode		= 0644,
7000 		.proc_handler	= proc_douintvec_minmax,
7001 		.extra1		= (void *)SYSCTL_ZERO,
7002 		.extra2		= (void *)&ioam6_if_id_max,
7003 	},
7004 	{
7005 		.procname	= "ioam6_id_wide",
7006 		.data		= &ipv6_devconf.ioam6_id_wide,
7007 		.maxlen		= sizeof(u32),
7008 		.mode		= 0644,
7009 		.proc_handler	= proc_douintvec,
7010 	},
7011 	{
7012 		.procname	= "ndisc_evict_nocarrier",
7013 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
7014 		.maxlen		= sizeof(u8),
7015 		.mode		= 0644,
7016 		.proc_handler	= proc_dou8vec_minmax,
7017 		.extra1		= (void *)SYSCTL_ZERO,
7018 		.extra2		= (void *)SYSCTL_ONE,
7019 	},
7020 	{
7021 		/* sentinel */
7022 	}
7023 };
7024 
7025 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7026 		struct inet6_dev *idev, struct ipv6_devconf *p)
7027 {
7028 	int i, ifindex;
7029 	struct ctl_table *table;
7030 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7031 
7032 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
7033 	if (!table)
7034 		goto out;
7035 
7036 	for (i = 0; table[i].data; i++) {
7037 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7038 		/* If one of these is already set, then it is not safe to
7039 		 * overwrite either of them: this makes proc_dointvec_minmax
7040 		 * usable.
7041 		 */
7042 		if (!table[i].extra1 && !table[i].extra2) {
7043 			table[i].extra1 = idev; /* embedded; no ref */
7044 			table[i].extra2 = net;
7045 		}
7046 	}
7047 
7048 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7049 
7050 	p->sysctl_header = register_net_sysctl(net, path, table);
7051 	if (!p->sysctl_header)
7052 		goto free;
7053 
7054 	if (!strcmp(dev_name, "all"))
7055 		ifindex = NETCONFA_IFINDEX_ALL;
7056 	else if (!strcmp(dev_name, "default"))
7057 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7058 	else
7059 		ifindex = idev->dev->ifindex;
7060 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7061 				     ifindex, p);
7062 	return 0;
7063 
7064 free:
7065 	kfree(table);
7066 out:
7067 	return -ENOBUFS;
7068 }
7069 
7070 static void __addrconf_sysctl_unregister(struct net *net,
7071 					 struct ipv6_devconf *p, int ifindex)
7072 {
7073 	struct ctl_table *table;
7074 
7075 	if (!p->sysctl_header)
7076 		return;
7077 
7078 	table = p->sysctl_header->ctl_table_arg;
7079 	unregister_net_sysctl_table(p->sysctl_header);
7080 	p->sysctl_header = NULL;
7081 	kfree(table);
7082 
7083 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7084 }
7085 
7086 static int addrconf_sysctl_register(struct inet6_dev *idev)
7087 {
7088 	int err;
7089 
7090 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7091 		return -EINVAL;
7092 
7093 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7094 				    &ndisc_ifinfo_sysctl_change);
7095 	if (err)
7096 		return err;
7097 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7098 					 idev, &idev->cnf);
7099 	if (err)
7100 		neigh_sysctl_unregister(idev->nd_parms);
7101 
7102 	return err;
7103 }
7104 
7105 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7106 {
7107 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7108 				     idev->dev->ifindex);
7109 	neigh_sysctl_unregister(idev->nd_parms);
7110 }
7111 
7112 
7113 #endif
7114 
7115 static int __net_init addrconf_init_net(struct net *net)
7116 {
7117 	int err = -ENOMEM;
7118 	struct ipv6_devconf *all, *dflt;
7119 
7120 	spin_lock_init(&net->ipv6.addrconf_hash_lock);
7121 	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7122 	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7123 					   sizeof(struct hlist_head),
7124 					   GFP_KERNEL);
7125 	if (!net->ipv6.inet6_addr_lst)
7126 		goto err_alloc_addr;
7127 
7128 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7129 	if (!all)
7130 		goto err_alloc_all;
7131 
7132 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7133 	if (!dflt)
7134 		goto err_alloc_dflt;
7135 
7136 	if (IS_ENABLED(CONFIG_SYSCTL) &&
7137 	    !net_eq(net, &init_net)) {
7138 		switch (sysctl_devconf_inherit_init_net) {
7139 		case 1:  /* copy from init_net */
7140 			memcpy(all, init_net.ipv6.devconf_all,
7141 			       sizeof(ipv6_devconf));
7142 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7143 			       sizeof(ipv6_devconf_dflt));
7144 			break;
7145 		case 3: /* copy from the current netns */
7146 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7147 			       sizeof(ipv6_devconf));
7148 			memcpy(dflt,
7149 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7150 			       sizeof(ipv6_devconf_dflt));
7151 			break;
7152 		case 0:
7153 		case 2:
7154 			/* use compiled values */
7155 			break;
7156 		}
7157 	}
7158 
7159 	/* these will be inherited by all namespaces */
7160 	dflt->autoconf = ipv6_defaults.autoconf;
7161 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7162 
7163 	dflt->stable_secret.initialized = false;
7164 	all->stable_secret.initialized = false;
7165 
7166 	net->ipv6.devconf_all = all;
7167 	net->ipv6.devconf_dflt = dflt;
7168 
7169 #ifdef CONFIG_SYSCTL
7170 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7171 	if (err < 0)
7172 		goto err_reg_all;
7173 
7174 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7175 	if (err < 0)
7176 		goto err_reg_dflt;
7177 #endif
7178 	return 0;
7179 
7180 #ifdef CONFIG_SYSCTL
7181 err_reg_dflt:
7182 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7183 err_reg_all:
7184 	kfree(dflt);
7185 #endif
7186 err_alloc_dflt:
7187 	kfree(all);
7188 err_alloc_all:
7189 	kfree(net->ipv6.inet6_addr_lst);
7190 err_alloc_addr:
7191 	return err;
7192 }
7193 
7194 static void __net_exit addrconf_exit_net(struct net *net)
7195 {
7196 	int i;
7197 
7198 #ifdef CONFIG_SYSCTL
7199 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7200 				     NETCONFA_IFINDEX_DEFAULT);
7201 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7202 				     NETCONFA_IFINDEX_ALL);
7203 #endif
7204 	kfree(net->ipv6.devconf_dflt);
7205 	net->ipv6.devconf_dflt = NULL;
7206 	kfree(net->ipv6.devconf_all);
7207 	net->ipv6.devconf_all = NULL;
7208 
7209 	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7210 	/*
7211 	 *	Check hash table, then free it.
7212 	 */
7213 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7214 		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7215 
7216 	kfree(net->ipv6.inet6_addr_lst);
7217 	net->ipv6.inet6_addr_lst = NULL;
7218 }
7219 
7220 static struct pernet_operations addrconf_ops = {
7221 	.init = addrconf_init_net,
7222 	.exit = addrconf_exit_net,
7223 };
7224 
7225 static struct rtnl_af_ops inet6_ops __read_mostly = {
7226 	.family		  = AF_INET6,
7227 	.fill_link_af	  = inet6_fill_link_af,
7228 	.get_link_af_size = inet6_get_link_af_size,
7229 	.validate_link_af = inet6_validate_link_af,
7230 	.set_link_af	  = inet6_set_link_af,
7231 };
7232 
7233 /*
7234  *	Init / cleanup code
7235  */
7236 
7237 int __init addrconf_init(void)
7238 {
7239 	struct inet6_dev *idev;
7240 	int err;
7241 
7242 	err = ipv6_addr_label_init();
7243 	if (err < 0) {
7244 		pr_crit("%s: cannot initialize default policy table: %d\n",
7245 			__func__, err);
7246 		goto out;
7247 	}
7248 
7249 	err = register_pernet_subsys(&addrconf_ops);
7250 	if (err < 0)
7251 		goto out_addrlabel;
7252 
7253 	addrconf_wq = create_workqueue("ipv6_addrconf");
7254 	if (!addrconf_wq) {
7255 		err = -ENOMEM;
7256 		goto out_nowq;
7257 	}
7258 
7259 	rtnl_lock();
7260 	idev = ipv6_add_dev(blackhole_netdev);
7261 	rtnl_unlock();
7262 	if (IS_ERR(idev)) {
7263 		err = PTR_ERR(idev);
7264 		goto errlo;
7265 	}
7266 
7267 	ip6_route_init_special_entries();
7268 
7269 	register_netdevice_notifier(&ipv6_dev_notf);
7270 
7271 	addrconf_verify(&init_net);
7272 
7273 	rtnl_af_register(&inet6_ops);
7274 
7275 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7276 				   NULL, inet6_dump_ifinfo, 0);
7277 	if (err < 0)
7278 		goto errout;
7279 
7280 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7281 				   inet6_rtm_newaddr, NULL, 0);
7282 	if (err < 0)
7283 		goto errout;
7284 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7285 				   inet6_rtm_deladdr, NULL, 0);
7286 	if (err < 0)
7287 		goto errout;
7288 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7289 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7290 				   RTNL_FLAG_DOIT_UNLOCKED);
7291 	if (err < 0)
7292 		goto errout;
7293 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7294 				   NULL, inet6_dump_ifmcaddr, 0);
7295 	if (err < 0)
7296 		goto errout;
7297 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7298 				   NULL, inet6_dump_ifacaddr, 0);
7299 	if (err < 0)
7300 		goto errout;
7301 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7302 				   inet6_netconf_get_devconf,
7303 				   inet6_netconf_dump_devconf,
7304 				   RTNL_FLAG_DOIT_UNLOCKED);
7305 	if (err < 0)
7306 		goto errout;
7307 	err = ipv6_addr_label_rtnl_register();
7308 	if (err < 0)
7309 		goto errout;
7310 
7311 	return 0;
7312 errout:
7313 	rtnl_unregister_all(PF_INET6);
7314 	rtnl_af_unregister(&inet6_ops);
7315 	unregister_netdevice_notifier(&ipv6_dev_notf);
7316 errlo:
7317 	destroy_workqueue(addrconf_wq);
7318 out_nowq:
7319 	unregister_pernet_subsys(&addrconf_ops);
7320 out_addrlabel:
7321 	ipv6_addr_label_cleanup();
7322 out:
7323 	return err;
7324 }
7325 
7326 void addrconf_cleanup(void)
7327 {
7328 	struct net_device *dev;
7329 
7330 	unregister_netdevice_notifier(&ipv6_dev_notf);
7331 	unregister_pernet_subsys(&addrconf_ops);
7332 	ipv6_addr_label_cleanup();
7333 
7334 	rtnl_af_unregister(&inet6_ops);
7335 
7336 	rtnl_lock();
7337 
7338 	/* clean dev list */
7339 	for_each_netdev(&init_net, dev) {
7340 		if (__in6_dev_get(dev) == NULL)
7341 			continue;
7342 		addrconf_ifdown(dev, true);
7343 	}
7344 	addrconf_ifdown(init_net.loopback_dev, true);
7345 
7346 	rtnl_unlock();
7347 
7348 	destroy_workqueue(addrconf_wq);
7349 }
7350