xref: /openbmc/linux/net/ipv6/addrconf.c (revision 71a15258f3c92eb1c4ae98bbfca9459f4723d5d3)
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 = get_random_u32_inclusive(900000, 1100000) * (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 = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116 	do_div(tmp, 1000000);
117 	if ((s32)tmp > mrt) {
118 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 		tmp = get_random_u32_inclusive(900000, 1100000) * (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_min_lft	= 0,
206 	.accept_ra_pinfo	= 1,
207 #ifdef CONFIG_IPV6_ROUTER_PREF
208 	.accept_ra_rtr_pref	= 1,
209 	.rtr_probe_interval	= 60 * HZ,
210 #ifdef CONFIG_IPV6_ROUTE_INFO
211 	.accept_ra_rt_info_min_plen = 0,
212 	.accept_ra_rt_info_max_plen = 0,
213 #endif
214 #endif
215 	.proxy_ndp		= 0,
216 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
217 	.disable_ipv6		= 0,
218 	.accept_dad		= 0,
219 	.suppress_frag_ndisc	= 1,
220 	.accept_ra_mtu		= 1,
221 	.stable_secret		= {
222 		.initialized = false,
223 	},
224 	.use_oif_addrs_only	= 0,
225 	.ignore_routes_with_linkdown = 0,
226 	.keep_addr_on_down	= 0,
227 	.seg6_enabled		= 0,
228 #ifdef CONFIG_IPV6_SEG6_HMAC
229 	.seg6_require_hmac	= 0,
230 #endif
231 	.enhanced_dad           = 1,
232 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
233 	.disable_policy		= 0,
234 	.rpl_seg_enabled	= 0,
235 	.ioam6_enabled		= 0,
236 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
237 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
238 	.ndisc_evict_nocarrier	= 1,
239 };
240 
241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
242 	.forwarding		= 0,
243 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
244 	.mtu6			= IPV6_MIN_MTU,
245 	.accept_ra		= 1,
246 	.accept_redirects	= 1,
247 	.autoconf		= 1,
248 	.force_mld_version	= 0,
249 	.mldv1_unsolicited_report_interval = 10 * HZ,
250 	.mldv2_unsolicited_report_interval = HZ,
251 	.dad_transmits		= 1,
252 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
253 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
254 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
255 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
256 	.use_tempaddr		= 0,
257 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
258 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
259 	.regen_max_retry	= REGEN_MAX_RETRY,
260 	.max_desync_factor	= MAX_DESYNC_FACTOR,
261 	.max_addresses		= IPV6_MAX_ADDRESSES,
262 	.accept_ra_defrtr	= 1,
263 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
264 	.accept_ra_from_local	= 0,
265 	.accept_ra_min_hop_limit= 1,
266 	.accept_ra_min_lft	= 0,
267 	.accept_ra_pinfo	= 1,
268 #ifdef CONFIG_IPV6_ROUTER_PREF
269 	.accept_ra_rtr_pref	= 1,
270 	.rtr_probe_interval	= 60 * HZ,
271 #ifdef CONFIG_IPV6_ROUTE_INFO
272 	.accept_ra_rt_info_min_plen = 0,
273 	.accept_ra_rt_info_max_plen = 0,
274 #endif
275 #endif
276 	.proxy_ndp		= 0,
277 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
278 	.disable_ipv6		= 0,
279 	.accept_dad		= 1,
280 	.suppress_frag_ndisc	= 1,
281 	.accept_ra_mtu		= 1,
282 	.stable_secret		= {
283 		.initialized = false,
284 	},
285 	.use_oif_addrs_only	= 0,
286 	.ignore_routes_with_linkdown = 0,
287 	.keep_addr_on_down	= 0,
288 	.seg6_enabled		= 0,
289 #ifdef CONFIG_IPV6_SEG6_HMAC
290 	.seg6_require_hmac	= 0,
291 #endif
292 	.enhanced_dad           = 1,
293 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
294 	.disable_policy		= 0,
295 	.rpl_seg_enabled	= 0,
296 	.ioam6_enabled		= 0,
297 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
298 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
299 	.ndisc_evict_nocarrier	= 1,
300 };
301 
302 /* Check if link is ready: is it up and is a valid qdisc available */
303 static inline bool addrconf_link_ready(const struct net_device *dev)
304 {
305 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
306 }
307 
308 static void addrconf_del_rs_timer(struct inet6_dev *idev)
309 {
310 	if (del_timer(&idev->rs_timer))
311 		__in6_dev_put(idev);
312 }
313 
314 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
315 {
316 	if (cancel_delayed_work(&ifp->dad_work))
317 		__in6_ifa_put(ifp);
318 }
319 
320 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
321 				  unsigned long when)
322 {
323 	if (!mod_timer(&idev->rs_timer, jiffies + when))
324 		in6_dev_hold(idev);
325 }
326 
327 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
328 				   unsigned long delay)
329 {
330 	in6_ifa_hold(ifp);
331 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
332 		in6_ifa_put(ifp);
333 }
334 
335 static int snmp6_alloc_dev(struct inet6_dev *idev)
336 {
337 	int i;
338 
339 	idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
340 	if (!idev->stats.ipv6)
341 		goto err_ip;
342 
343 	for_each_possible_cpu(i) {
344 		struct ipstats_mib *addrconf_stats;
345 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
346 		u64_stats_init(&addrconf_stats->syncp);
347 	}
348 
349 
350 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
351 					GFP_KERNEL);
352 	if (!idev->stats.icmpv6dev)
353 		goto err_icmp;
354 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
355 					   GFP_KERNEL_ACCOUNT);
356 	if (!idev->stats.icmpv6msgdev)
357 		goto err_icmpmsg;
358 
359 	return 0;
360 
361 err_icmpmsg:
362 	kfree(idev->stats.icmpv6dev);
363 err_icmp:
364 	free_percpu(idev->stats.ipv6);
365 err_ip:
366 	return -ENOMEM;
367 }
368 
369 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
370 {
371 	struct inet6_dev *ndev;
372 	int err = -ENOMEM;
373 
374 	ASSERT_RTNL();
375 
376 	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
377 		return ERR_PTR(-EINVAL);
378 
379 	ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
380 	if (!ndev)
381 		return ERR_PTR(err);
382 
383 	rwlock_init(&ndev->lock);
384 	ndev->dev = dev;
385 	INIT_LIST_HEAD(&ndev->addr_list);
386 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
387 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
388 
389 	if (ndev->cnf.stable_secret.initialized)
390 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
391 
392 	ndev->cnf.mtu6 = dev->mtu;
393 	ndev->ra_mtu = 0;
394 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
395 	if (!ndev->nd_parms) {
396 		kfree(ndev);
397 		return ERR_PTR(err);
398 	}
399 	if (ndev->cnf.forwarding)
400 		dev_disable_lro(dev);
401 	/* We refer to the device */
402 	netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
403 
404 	if (snmp6_alloc_dev(ndev) < 0) {
405 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
406 			   __func__);
407 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
408 		netdev_put(dev, &ndev->dev_tracker);
409 		kfree(ndev);
410 		return ERR_PTR(err);
411 	}
412 
413 	if (dev != blackhole_netdev) {
414 		if (snmp6_register_dev(ndev) < 0) {
415 			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
416 				   __func__, dev->name);
417 			goto err_release;
418 		}
419 	}
420 	/* One reference from device. */
421 	refcount_set(&ndev->refcnt, 1);
422 
423 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
424 		ndev->cnf.accept_dad = -1;
425 
426 #if IS_ENABLED(CONFIG_IPV6_SIT)
427 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
428 		pr_info("%s: Disabled Multicast RS\n", dev->name);
429 		ndev->cnf.rtr_solicits = 0;
430 	}
431 #endif
432 
433 	INIT_LIST_HEAD(&ndev->tempaddr_list);
434 	ndev->desync_factor = U32_MAX;
435 	if ((dev->flags&IFF_LOOPBACK) ||
436 	    dev->type == ARPHRD_TUNNEL ||
437 	    dev->type == ARPHRD_TUNNEL6 ||
438 	    dev->type == ARPHRD_SIT ||
439 	    dev->type == ARPHRD_NONE) {
440 		ndev->cnf.use_tempaddr = -1;
441 	}
442 
443 	ndev->token = in6addr_any;
444 
445 	if (netif_running(dev) && addrconf_link_ready(dev))
446 		ndev->if_flags |= IF_READY;
447 
448 	ipv6_mc_init_dev(ndev);
449 	ndev->tstamp = jiffies;
450 	if (dev != blackhole_netdev) {
451 		err = addrconf_sysctl_register(ndev);
452 		if (err) {
453 			ipv6_mc_destroy_dev(ndev);
454 			snmp6_unregister_dev(ndev);
455 			goto err_release;
456 		}
457 	}
458 	/* protected by rtnl_lock */
459 	rcu_assign_pointer(dev->ip6_ptr, ndev);
460 
461 	if (dev != blackhole_netdev) {
462 		/* Join interface-local all-node multicast group */
463 		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
464 
465 		/* Join all-node multicast group */
466 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
467 
468 		/* Join all-router multicast group if forwarding is set */
469 		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
470 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
471 	}
472 	return ndev;
473 
474 err_release:
475 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
476 	ndev->dead = 1;
477 	in6_dev_finish_destroy(ndev);
478 	return ERR_PTR(err);
479 }
480 
481 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
482 {
483 	struct inet6_dev *idev;
484 
485 	ASSERT_RTNL();
486 
487 	idev = __in6_dev_get(dev);
488 	if (!idev) {
489 		idev = ipv6_add_dev(dev);
490 		if (IS_ERR(idev))
491 			return idev;
492 	}
493 
494 	if (dev->flags&IFF_UP)
495 		ipv6_mc_up(idev);
496 	return idev;
497 }
498 
499 static int inet6_netconf_msgsize_devconf(int type)
500 {
501 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
502 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
503 	bool all = false;
504 
505 	if (type == NETCONFA_ALL)
506 		all = true;
507 
508 	if (all || type == NETCONFA_FORWARDING)
509 		size += nla_total_size(4);
510 #ifdef CONFIG_IPV6_MROUTE
511 	if (all || type == NETCONFA_MC_FORWARDING)
512 		size += nla_total_size(4);
513 #endif
514 	if (all || type == NETCONFA_PROXY_NEIGH)
515 		size += nla_total_size(4);
516 
517 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
518 		size += nla_total_size(4);
519 
520 	return size;
521 }
522 
523 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
524 				      struct ipv6_devconf *devconf, u32 portid,
525 				      u32 seq, int event, unsigned int flags,
526 				      int type)
527 {
528 	struct nlmsghdr  *nlh;
529 	struct netconfmsg *ncm;
530 	bool all = false;
531 
532 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
533 			flags);
534 	if (!nlh)
535 		return -EMSGSIZE;
536 
537 	if (type == NETCONFA_ALL)
538 		all = true;
539 
540 	ncm = nlmsg_data(nlh);
541 	ncm->ncm_family = AF_INET6;
542 
543 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
544 		goto nla_put_failure;
545 
546 	if (!devconf)
547 		goto out;
548 
549 	if ((all || type == NETCONFA_FORWARDING) &&
550 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
551 		goto nla_put_failure;
552 #ifdef CONFIG_IPV6_MROUTE
553 	if ((all || type == NETCONFA_MC_FORWARDING) &&
554 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
555 			atomic_read(&devconf->mc_forwarding)) < 0)
556 		goto nla_put_failure;
557 #endif
558 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
559 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
560 		goto nla_put_failure;
561 
562 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
563 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
564 			devconf->ignore_routes_with_linkdown) < 0)
565 		goto nla_put_failure;
566 
567 out:
568 	nlmsg_end(skb, nlh);
569 	return 0;
570 
571 nla_put_failure:
572 	nlmsg_cancel(skb, nlh);
573 	return -EMSGSIZE;
574 }
575 
576 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
577 				  int ifindex, struct ipv6_devconf *devconf)
578 {
579 	struct sk_buff *skb;
580 	int err = -ENOBUFS;
581 
582 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
583 	if (!skb)
584 		goto errout;
585 
586 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
587 					 event, 0, type);
588 	if (err < 0) {
589 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
590 		WARN_ON(err == -EMSGSIZE);
591 		kfree_skb(skb);
592 		goto errout;
593 	}
594 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
595 	return;
596 errout:
597 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
598 }
599 
600 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
601 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
602 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
603 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
604 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
605 };
606 
607 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
608 				       const struct nlmsghdr *nlh,
609 				       struct nlattr **tb,
610 				       struct netlink_ext_ack *extack)
611 {
612 	int i, err;
613 
614 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
615 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
616 		return -EINVAL;
617 	}
618 
619 	if (!netlink_strict_get_check(skb))
620 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
621 					      tb, NETCONFA_MAX,
622 					      devconf_ipv6_policy, extack);
623 
624 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
625 					    tb, NETCONFA_MAX,
626 					    devconf_ipv6_policy, extack);
627 	if (err)
628 		return err;
629 
630 	for (i = 0; i <= NETCONFA_MAX; i++) {
631 		if (!tb[i])
632 			continue;
633 
634 		switch (i) {
635 		case NETCONFA_IFINDEX:
636 			break;
637 		default:
638 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
639 			return -EINVAL;
640 		}
641 	}
642 
643 	return 0;
644 }
645 
646 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
647 				     struct nlmsghdr *nlh,
648 				     struct netlink_ext_ack *extack)
649 {
650 	struct net *net = sock_net(in_skb->sk);
651 	struct nlattr *tb[NETCONFA_MAX+1];
652 	struct inet6_dev *in6_dev = NULL;
653 	struct net_device *dev = NULL;
654 	struct sk_buff *skb;
655 	struct ipv6_devconf *devconf;
656 	int ifindex;
657 	int err;
658 
659 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
660 	if (err < 0)
661 		return err;
662 
663 	if (!tb[NETCONFA_IFINDEX])
664 		return -EINVAL;
665 
666 	err = -EINVAL;
667 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
668 	switch (ifindex) {
669 	case NETCONFA_IFINDEX_ALL:
670 		devconf = net->ipv6.devconf_all;
671 		break;
672 	case NETCONFA_IFINDEX_DEFAULT:
673 		devconf = net->ipv6.devconf_dflt;
674 		break;
675 	default:
676 		dev = dev_get_by_index(net, ifindex);
677 		if (!dev)
678 			return -EINVAL;
679 		in6_dev = in6_dev_get(dev);
680 		if (!in6_dev)
681 			goto errout;
682 		devconf = &in6_dev->cnf;
683 		break;
684 	}
685 
686 	err = -ENOBUFS;
687 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
688 	if (!skb)
689 		goto errout;
690 
691 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
692 					 NETLINK_CB(in_skb).portid,
693 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
694 					 NETCONFA_ALL);
695 	if (err < 0) {
696 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
697 		WARN_ON(err == -EMSGSIZE);
698 		kfree_skb(skb);
699 		goto errout;
700 	}
701 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
702 errout:
703 	if (in6_dev)
704 		in6_dev_put(in6_dev);
705 	dev_put(dev);
706 	return err;
707 }
708 
709 /* Combine dev_addr_genid and dev_base_seq to detect changes.
710  */
711 static u32 inet6_base_seq(const struct net *net)
712 {
713 	u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
714 		  net->dev_base_seq;
715 
716 	/* Must not return 0 (see nl_dump_check_consistent()).
717 	 * Chose a value far away from 0.
718 	 */
719 	if (!res)
720 		res = 0x80000000;
721 	return res;
722 }
723 
724 
725 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
726 				      struct netlink_callback *cb)
727 {
728 	const struct nlmsghdr *nlh = cb->nlh;
729 	struct net *net = sock_net(skb->sk);
730 	int h, s_h;
731 	int idx, s_idx;
732 	struct net_device *dev;
733 	struct inet6_dev *idev;
734 	struct hlist_head *head;
735 
736 	if (cb->strict_check) {
737 		struct netlink_ext_ack *extack = cb->extack;
738 		struct netconfmsg *ncm;
739 
740 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
741 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
742 			return -EINVAL;
743 		}
744 
745 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
746 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
747 			return -EINVAL;
748 		}
749 	}
750 
751 	s_h = cb->args[0];
752 	s_idx = idx = cb->args[1];
753 
754 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
755 		idx = 0;
756 		head = &net->dev_index_head[h];
757 		rcu_read_lock();
758 		cb->seq = inet6_base_seq(net);
759 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
760 			if (idx < s_idx)
761 				goto cont;
762 			idev = __in6_dev_get(dev);
763 			if (!idev)
764 				goto cont;
765 
766 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
767 						       &idev->cnf,
768 						       NETLINK_CB(cb->skb).portid,
769 						       nlh->nlmsg_seq,
770 						       RTM_NEWNETCONF,
771 						       NLM_F_MULTI,
772 						       NETCONFA_ALL) < 0) {
773 				rcu_read_unlock();
774 				goto done;
775 			}
776 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
777 cont:
778 			idx++;
779 		}
780 		rcu_read_unlock();
781 	}
782 	if (h == NETDEV_HASHENTRIES) {
783 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
784 					       net->ipv6.devconf_all,
785 					       NETLINK_CB(cb->skb).portid,
786 					       nlh->nlmsg_seq,
787 					       RTM_NEWNETCONF, NLM_F_MULTI,
788 					       NETCONFA_ALL) < 0)
789 			goto done;
790 		else
791 			h++;
792 	}
793 	if (h == NETDEV_HASHENTRIES + 1) {
794 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
795 					       net->ipv6.devconf_dflt,
796 					       NETLINK_CB(cb->skb).portid,
797 					       nlh->nlmsg_seq,
798 					       RTM_NEWNETCONF, NLM_F_MULTI,
799 					       NETCONFA_ALL) < 0)
800 			goto done;
801 		else
802 			h++;
803 	}
804 done:
805 	cb->args[0] = h;
806 	cb->args[1] = idx;
807 
808 	return skb->len;
809 }
810 
811 #ifdef CONFIG_SYSCTL
812 static void dev_forward_change(struct inet6_dev *idev)
813 {
814 	struct net_device *dev;
815 	struct inet6_ifaddr *ifa;
816 	LIST_HEAD(tmp_addr_list);
817 
818 	if (!idev)
819 		return;
820 	dev = idev->dev;
821 	if (idev->cnf.forwarding)
822 		dev_disable_lro(dev);
823 	if (dev->flags & IFF_MULTICAST) {
824 		if (idev->cnf.forwarding) {
825 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
826 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
827 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
828 		} else {
829 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
830 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
831 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
832 		}
833 	}
834 
835 	read_lock_bh(&idev->lock);
836 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
837 		if (ifa->flags&IFA_F_TENTATIVE)
838 			continue;
839 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
840 	}
841 	read_unlock_bh(&idev->lock);
842 
843 	while (!list_empty(&tmp_addr_list)) {
844 		ifa = list_first_entry(&tmp_addr_list,
845 				       struct inet6_ifaddr, if_list_aux);
846 		list_del(&ifa->if_list_aux);
847 		if (idev->cnf.forwarding)
848 			addrconf_join_anycast(ifa);
849 		else
850 			addrconf_leave_anycast(ifa);
851 	}
852 
853 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
854 				     NETCONFA_FORWARDING,
855 				     dev->ifindex, &idev->cnf);
856 }
857 
858 
859 static void addrconf_forward_change(struct net *net, __s32 newf)
860 {
861 	struct net_device *dev;
862 	struct inet6_dev *idev;
863 
864 	for_each_netdev(net, dev) {
865 		idev = __in6_dev_get(dev);
866 		if (idev) {
867 			int changed = (!idev->cnf.forwarding) ^ (!newf);
868 			idev->cnf.forwarding = newf;
869 			if (changed)
870 				dev_forward_change(idev);
871 		}
872 	}
873 }
874 
875 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
876 {
877 	struct net *net;
878 	int old;
879 
880 	if (!rtnl_trylock())
881 		return restart_syscall();
882 
883 	net = (struct net *)table->extra2;
884 	old = *p;
885 	*p = newf;
886 
887 	if (p == &net->ipv6.devconf_dflt->forwarding) {
888 		if ((!newf) ^ (!old))
889 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
890 						     NETCONFA_FORWARDING,
891 						     NETCONFA_IFINDEX_DEFAULT,
892 						     net->ipv6.devconf_dflt);
893 		rtnl_unlock();
894 		return 0;
895 	}
896 
897 	if (p == &net->ipv6.devconf_all->forwarding) {
898 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
899 
900 		net->ipv6.devconf_dflt->forwarding = newf;
901 		if ((!newf) ^ (!old_dflt))
902 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
903 						     NETCONFA_FORWARDING,
904 						     NETCONFA_IFINDEX_DEFAULT,
905 						     net->ipv6.devconf_dflt);
906 
907 		addrconf_forward_change(net, newf);
908 		if ((!newf) ^ (!old))
909 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
910 						     NETCONFA_FORWARDING,
911 						     NETCONFA_IFINDEX_ALL,
912 						     net->ipv6.devconf_all);
913 	} else if ((!newf) ^ (!old))
914 		dev_forward_change((struct inet6_dev *)table->extra1);
915 	rtnl_unlock();
916 
917 	if (newf)
918 		rt6_purge_dflt_routers(net);
919 	return 1;
920 }
921 
922 static void addrconf_linkdown_change(struct net *net, __s32 newf)
923 {
924 	struct net_device *dev;
925 	struct inet6_dev *idev;
926 
927 	for_each_netdev(net, dev) {
928 		idev = __in6_dev_get(dev);
929 		if (idev) {
930 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
931 
932 			idev->cnf.ignore_routes_with_linkdown = newf;
933 			if (changed)
934 				inet6_netconf_notify_devconf(dev_net(dev),
935 							     RTM_NEWNETCONF,
936 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
937 							     dev->ifindex,
938 							     &idev->cnf);
939 		}
940 	}
941 }
942 
943 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
944 {
945 	struct net *net;
946 	int old;
947 
948 	if (!rtnl_trylock())
949 		return restart_syscall();
950 
951 	net = (struct net *)table->extra2;
952 	old = *p;
953 	*p = newf;
954 
955 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
956 		if ((!newf) ^ (!old))
957 			inet6_netconf_notify_devconf(net,
958 						     RTM_NEWNETCONF,
959 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
960 						     NETCONFA_IFINDEX_DEFAULT,
961 						     net->ipv6.devconf_dflt);
962 		rtnl_unlock();
963 		return 0;
964 	}
965 
966 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
967 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
968 		addrconf_linkdown_change(net, newf);
969 		if ((!newf) ^ (!old))
970 			inet6_netconf_notify_devconf(net,
971 						     RTM_NEWNETCONF,
972 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
973 						     NETCONFA_IFINDEX_ALL,
974 						     net->ipv6.devconf_all);
975 	}
976 	rtnl_unlock();
977 
978 	return 1;
979 }
980 
981 #endif
982 
983 /* Nobody refers to this ifaddr, destroy it */
984 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
985 {
986 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
987 
988 #ifdef NET_REFCNT_DEBUG
989 	pr_debug("%s\n", __func__);
990 #endif
991 
992 	in6_dev_put(ifp->idev);
993 
994 	if (cancel_delayed_work(&ifp->dad_work))
995 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
996 			  ifp);
997 
998 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
999 		pr_warn("Freeing alive inet6 address %p\n", ifp);
1000 		return;
1001 	}
1002 
1003 	kfree_rcu(ifp, rcu);
1004 }
1005 
1006 static void
1007 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1008 {
1009 	struct list_head *p;
1010 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1011 
1012 	/*
1013 	 * Each device address list is sorted in order of scope -
1014 	 * global before linklocal.
1015 	 */
1016 	list_for_each(p, &idev->addr_list) {
1017 		struct inet6_ifaddr *ifa
1018 			= list_entry(p, struct inet6_ifaddr, if_list);
1019 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1020 			break;
1021 	}
1022 
1023 	list_add_tail_rcu(&ifp->if_list, p);
1024 }
1025 
1026 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1027 {
1028 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1029 
1030 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1031 }
1032 
1033 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1034 			       struct net_device *dev, unsigned int hash)
1035 {
1036 	struct inet6_ifaddr *ifp;
1037 
1038 	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1039 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1040 			if (!dev || ifp->idev->dev == dev)
1041 				return true;
1042 		}
1043 	}
1044 	return false;
1045 }
1046 
1047 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1048 {
1049 	struct net *net = dev_net(dev);
1050 	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1051 	int err = 0;
1052 
1053 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1054 
1055 	/* Ignore adding duplicate addresses on an interface */
1056 	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1057 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1058 		err = -EEXIST;
1059 	} else {
1060 		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1061 	}
1062 
1063 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1064 
1065 	return err;
1066 }
1067 
1068 /* On success it returns ifp with increased reference count */
1069 
1070 static struct inet6_ifaddr *
1071 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1072 	      bool can_block, struct netlink_ext_ack *extack)
1073 {
1074 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1075 	int addr_type = ipv6_addr_type(cfg->pfx);
1076 	struct net *net = dev_net(idev->dev);
1077 	struct inet6_ifaddr *ifa = NULL;
1078 	struct fib6_info *f6i = NULL;
1079 	int err = 0;
1080 
1081 	if (addr_type == IPV6_ADDR_ANY) {
1082 		NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1083 		return ERR_PTR(-EADDRNOTAVAIL);
1084 	} else if (addr_type & IPV6_ADDR_MULTICAST &&
1085 		   !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1086 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1087 		return ERR_PTR(-EADDRNOTAVAIL);
1088 	} else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1089 		   !netif_is_l3_master(idev->dev) &&
1090 		   addr_type & IPV6_ADDR_LOOPBACK) {
1091 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1092 		return ERR_PTR(-EADDRNOTAVAIL);
1093 	}
1094 
1095 	if (idev->dead) {
1096 		NL_SET_ERR_MSG_MOD(extack, "device is going away");
1097 		err = -ENODEV;
1098 		goto out;
1099 	}
1100 
1101 	if (idev->cnf.disable_ipv6) {
1102 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1103 		err = -EACCES;
1104 		goto out;
1105 	}
1106 
1107 	/* validator notifier needs to be blocking;
1108 	 * do not call in atomic context
1109 	 */
1110 	if (can_block) {
1111 		struct in6_validator_info i6vi = {
1112 			.i6vi_addr = *cfg->pfx,
1113 			.i6vi_dev = idev,
1114 			.extack = extack,
1115 		};
1116 
1117 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1118 		err = notifier_to_errno(err);
1119 		if (err < 0)
1120 			goto out;
1121 	}
1122 
1123 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1124 	if (!ifa) {
1125 		err = -ENOBUFS;
1126 		goto out;
1127 	}
1128 
1129 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1130 	if (IS_ERR(f6i)) {
1131 		err = PTR_ERR(f6i);
1132 		f6i = NULL;
1133 		goto out;
1134 	}
1135 
1136 	neigh_parms_data_state_setall(idev->nd_parms);
1137 
1138 	ifa->addr = *cfg->pfx;
1139 	if (cfg->peer_pfx)
1140 		ifa->peer_addr = *cfg->peer_pfx;
1141 
1142 	spin_lock_init(&ifa->lock);
1143 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1144 	INIT_HLIST_NODE(&ifa->addr_lst);
1145 	ifa->scope = cfg->scope;
1146 	ifa->prefix_len = cfg->plen;
1147 	ifa->rt_priority = cfg->rt_priority;
1148 	ifa->flags = cfg->ifa_flags;
1149 	ifa->ifa_proto = cfg->ifa_proto;
1150 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1151 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1152 		ifa->flags |= IFA_F_TENTATIVE;
1153 	ifa->valid_lft = cfg->valid_lft;
1154 	ifa->prefered_lft = cfg->preferred_lft;
1155 	ifa->cstamp = ifa->tstamp = jiffies;
1156 	ifa->tokenized = false;
1157 
1158 	ifa->rt = f6i;
1159 
1160 	ifa->idev = idev;
1161 	in6_dev_hold(idev);
1162 
1163 	/* For caller */
1164 	refcount_set(&ifa->refcnt, 1);
1165 
1166 	rcu_read_lock();
1167 
1168 	err = ipv6_add_addr_hash(idev->dev, ifa);
1169 	if (err < 0) {
1170 		rcu_read_unlock();
1171 		goto out;
1172 	}
1173 
1174 	write_lock_bh(&idev->lock);
1175 
1176 	/* Add to inet6_dev unicast addr list. */
1177 	ipv6_link_dev_addr(idev, ifa);
1178 
1179 	if (ifa->flags&IFA_F_TEMPORARY) {
1180 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1181 		in6_ifa_hold(ifa);
1182 	}
1183 
1184 	in6_ifa_hold(ifa);
1185 	write_unlock_bh(&idev->lock);
1186 
1187 	rcu_read_unlock();
1188 
1189 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1190 out:
1191 	if (unlikely(err < 0)) {
1192 		fib6_info_release(f6i);
1193 
1194 		if (ifa) {
1195 			if (ifa->idev)
1196 				in6_dev_put(ifa->idev);
1197 			kfree(ifa);
1198 		}
1199 		ifa = ERR_PTR(err);
1200 	}
1201 
1202 	return ifa;
1203 }
1204 
1205 enum cleanup_prefix_rt_t {
1206 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1207 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1208 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1209 };
1210 
1211 /*
1212  * Check, whether the prefix for ifp would still need a prefix route
1213  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1214  * constants.
1215  *
1216  * 1) we don't purge prefix if address was not permanent.
1217  *    prefix is managed by its own lifetime.
1218  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1219  * 3) if there are no addresses, delete prefix.
1220  * 4) if there are still other permanent address(es),
1221  *    corresponding prefix is still permanent.
1222  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1223  *    don't purge the prefix, assume user space is managing it.
1224  * 6) otherwise, update prefix lifetime to the
1225  *    longest valid lifetime among the corresponding
1226  *    addresses on the device.
1227  *    Note: subsequent RA will update lifetime.
1228  **/
1229 static enum cleanup_prefix_rt_t
1230 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1231 {
1232 	struct inet6_ifaddr *ifa;
1233 	struct inet6_dev *idev = ifp->idev;
1234 	unsigned long lifetime;
1235 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1236 
1237 	*expires = jiffies;
1238 
1239 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1240 		if (ifa == ifp)
1241 			continue;
1242 		if (ifa->prefix_len != ifp->prefix_len ||
1243 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1244 				       ifp->prefix_len))
1245 			continue;
1246 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1247 			return CLEANUP_PREFIX_RT_NOP;
1248 
1249 		action = CLEANUP_PREFIX_RT_EXPIRE;
1250 
1251 		spin_lock(&ifa->lock);
1252 
1253 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1254 		/*
1255 		 * Note: Because this address is
1256 		 * not permanent, lifetime <
1257 		 * LONG_MAX / HZ here.
1258 		 */
1259 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1260 			*expires = ifa->tstamp + lifetime * HZ;
1261 		spin_unlock(&ifa->lock);
1262 	}
1263 
1264 	return action;
1265 }
1266 
1267 static void
1268 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1269 		     bool del_rt, bool del_peer)
1270 {
1271 	struct fib6_info *f6i;
1272 
1273 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1274 					ifp->prefix_len,
1275 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1276 	if (f6i) {
1277 		if (del_rt)
1278 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1279 		else {
1280 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1281 				fib6_set_expires(f6i, expires);
1282 			fib6_info_release(f6i);
1283 		}
1284 	}
1285 }
1286 
1287 
1288 /* This function wants to get referenced ifp and releases it before return */
1289 
1290 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1291 {
1292 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1293 	struct net *net = dev_net(ifp->idev->dev);
1294 	unsigned long expires;
1295 	int state;
1296 
1297 	ASSERT_RTNL();
1298 
1299 	spin_lock_bh(&ifp->lock);
1300 	state = ifp->state;
1301 	ifp->state = INET6_IFADDR_STATE_DEAD;
1302 	spin_unlock_bh(&ifp->lock);
1303 
1304 	if (state == INET6_IFADDR_STATE_DEAD)
1305 		goto out;
1306 
1307 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1308 	hlist_del_init_rcu(&ifp->addr_lst);
1309 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1310 
1311 	write_lock_bh(&ifp->idev->lock);
1312 
1313 	if (ifp->flags&IFA_F_TEMPORARY) {
1314 		list_del(&ifp->tmp_list);
1315 		if (ifp->ifpub) {
1316 			in6_ifa_put(ifp->ifpub);
1317 			ifp->ifpub = NULL;
1318 		}
1319 		__in6_ifa_put(ifp);
1320 	}
1321 
1322 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1323 		action = check_cleanup_prefix_route(ifp, &expires);
1324 
1325 	list_del_rcu(&ifp->if_list);
1326 	__in6_ifa_put(ifp);
1327 
1328 	write_unlock_bh(&ifp->idev->lock);
1329 
1330 	addrconf_del_dad_work(ifp);
1331 
1332 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1333 
1334 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1335 
1336 	if (action != CLEANUP_PREFIX_RT_NOP) {
1337 		cleanup_prefix_route(ifp, expires,
1338 			action == CLEANUP_PREFIX_RT_DEL, false);
1339 	}
1340 
1341 	/* clean up prefsrc entries */
1342 	rt6_remove_prefsrc(ifp);
1343 out:
1344 	in6_ifa_put(ifp);
1345 }
1346 
1347 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1348 {
1349 	struct inet6_dev *idev = ifp->idev;
1350 	unsigned long tmp_tstamp, age;
1351 	unsigned long regen_advance;
1352 	unsigned long now = jiffies;
1353 	s32 cnf_temp_preferred_lft;
1354 	struct inet6_ifaddr *ift;
1355 	struct ifa6_config cfg;
1356 	long max_desync_factor;
1357 	struct in6_addr addr;
1358 	int ret = 0;
1359 
1360 	write_lock_bh(&idev->lock);
1361 
1362 retry:
1363 	in6_dev_hold(idev);
1364 	if (idev->cnf.use_tempaddr <= 0) {
1365 		write_unlock_bh(&idev->lock);
1366 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1367 		in6_dev_put(idev);
1368 		ret = -1;
1369 		goto out;
1370 	}
1371 	spin_lock_bh(&ifp->lock);
1372 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1373 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1374 		spin_unlock_bh(&ifp->lock);
1375 		write_unlock_bh(&idev->lock);
1376 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1377 			__func__);
1378 		in6_dev_put(idev);
1379 		ret = -1;
1380 		goto out;
1381 	}
1382 	in6_ifa_hold(ifp);
1383 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1384 	ipv6_gen_rnd_iid(&addr);
1385 
1386 	age = (now - ifp->tstamp) / HZ;
1387 
1388 	regen_advance = idev->cnf.regen_max_retry *
1389 			idev->cnf.dad_transmits *
1390 			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1391 
1392 	/* recalculate max_desync_factor each time and update
1393 	 * idev->desync_factor if it's larger
1394 	 */
1395 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1396 	max_desync_factor = min_t(long,
1397 				  idev->cnf.max_desync_factor,
1398 				  cnf_temp_preferred_lft - regen_advance);
1399 
1400 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1401 		if (max_desync_factor > 0) {
1402 			get_random_bytes(&idev->desync_factor,
1403 					 sizeof(idev->desync_factor));
1404 			idev->desync_factor %= max_desync_factor;
1405 		} else {
1406 			idev->desync_factor = 0;
1407 		}
1408 	}
1409 
1410 	memset(&cfg, 0, sizeof(cfg));
1411 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1412 			      idev->cnf.temp_valid_lft + age);
1413 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1414 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1415 
1416 	cfg.plen = ifp->prefix_len;
1417 	tmp_tstamp = ifp->tstamp;
1418 	spin_unlock_bh(&ifp->lock);
1419 
1420 	write_unlock_bh(&idev->lock);
1421 
1422 	/* A temporary address is created only if this calculated Preferred
1423 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1424 	 * an implementation must not create a temporary address with a zero
1425 	 * Preferred Lifetime.
1426 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1427 	 * temporary addresses being generated.
1428 	 */
1429 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1430 	if (cfg.preferred_lft <= regen_advance + age) {
1431 		in6_ifa_put(ifp);
1432 		in6_dev_put(idev);
1433 		ret = -1;
1434 		goto out;
1435 	}
1436 
1437 	cfg.ifa_flags = IFA_F_TEMPORARY;
1438 	/* set in addrconf_prefix_rcv() */
1439 	if (ifp->flags & IFA_F_OPTIMISTIC)
1440 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1441 
1442 	cfg.pfx = &addr;
1443 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1444 
1445 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1446 	if (IS_ERR(ift)) {
1447 		in6_ifa_put(ifp);
1448 		in6_dev_put(idev);
1449 		pr_info("%s: retry temporary address regeneration\n", __func__);
1450 		write_lock_bh(&idev->lock);
1451 		goto retry;
1452 	}
1453 
1454 	spin_lock_bh(&ift->lock);
1455 	ift->ifpub = ifp;
1456 	ift->cstamp = now;
1457 	ift->tstamp = tmp_tstamp;
1458 	spin_unlock_bh(&ift->lock);
1459 
1460 	addrconf_dad_start(ift);
1461 	in6_ifa_put(ift);
1462 	in6_dev_put(idev);
1463 out:
1464 	return ret;
1465 }
1466 
1467 /*
1468  *	Choose an appropriate source address (RFC3484)
1469  */
1470 enum {
1471 	IPV6_SADDR_RULE_INIT = 0,
1472 	IPV6_SADDR_RULE_LOCAL,
1473 	IPV6_SADDR_RULE_SCOPE,
1474 	IPV6_SADDR_RULE_PREFERRED,
1475 #ifdef CONFIG_IPV6_MIP6
1476 	IPV6_SADDR_RULE_HOA,
1477 #endif
1478 	IPV6_SADDR_RULE_OIF,
1479 	IPV6_SADDR_RULE_LABEL,
1480 	IPV6_SADDR_RULE_PRIVACY,
1481 	IPV6_SADDR_RULE_ORCHID,
1482 	IPV6_SADDR_RULE_PREFIX,
1483 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1484 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1485 #endif
1486 	IPV6_SADDR_RULE_MAX
1487 };
1488 
1489 struct ipv6_saddr_score {
1490 	int			rule;
1491 	int			addr_type;
1492 	struct inet6_ifaddr	*ifa;
1493 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1494 	int			scopedist;
1495 	int			matchlen;
1496 };
1497 
1498 struct ipv6_saddr_dst {
1499 	const struct in6_addr *addr;
1500 	int ifindex;
1501 	int scope;
1502 	int label;
1503 	unsigned int prefs;
1504 };
1505 
1506 static inline int ipv6_saddr_preferred(int type)
1507 {
1508 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1509 		return 1;
1510 	return 0;
1511 }
1512 
1513 static bool ipv6_use_optimistic_addr(struct net *net,
1514 				     struct inet6_dev *idev)
1515 {
1516 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1517 	if (!idev)
1518 		return false;
1519 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1520 		return false;
1521 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1522 		return false;
1523 
1524 	return true;
1525 #else
1526 	return false;
1527 #endif
1528 }
1529 
1530 static bool ipv6_allow_optimistic_dad(struct net *net,
1531 				      struct inet6_dev *idev)
1532 {
1533 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1534 	if (!idev)
1535 		return false;
1536 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1537 		return false;
1538 
1539 	return true;
1540 #else
1541 	return false;
1542 #endif
1543 }
1544 
1545 static int ipv6_get_saddr_eval(struct net *net,
1546 			       struct ipv6_saddr_score *score,
1547 			       struct ipv6_saddr_dst *dst,
1548 			       int i)
1549 {
1550 	int ret;
1551 
1552 	if (i <= score->rule) {
1553 		switch (i) {
1554 		case IPV6_SADDR_RULE_SCOPE:
1555 			ret = score->scopedist;
1556 			break;
1557 		case IPV6_SADDR_RULE_PREFIX:
1558 			ret = score->matchlen;
1559 			break;
1560 		default:
1561 			ret = !!test_bit(i, score->scorebits);
1562 		}
1563 		goto out;
1564 	}
1565 
1566 	switch (i) {
1567 	case IPV6_SADDR_RULE_INIT:
1568 		/* Rule 0: remember if hiscore is not ready yet */
1569 		ret = !!score->ifa;
1570 		break;
1571 	case IPV6_SADDR_RULE_LOCAL:
1572 		/* Rule 1: Prefer same address */
1573 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1574 		break;
1575 	case IPV6_SADDR_RULE_SCOPE:
1576 		/* Rule 2: Prefer appropriate scope
1577 		 *
1578 		 *      ret
1579 		 *       ^
1580 		 *    -1 |  d 15
1581 		 *    ---+--+-+---> scope
1582 		 *       |
1583 		 *       |             d is scope of the destination.
1584 		 *  B-d  |  \
1585 		 *       |   \      <- smaller scope is better if
1586 		 *  B-15 |    \        if scope is enough for destination.
1587 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1588 		 * d-C-1 | /
1589 		 *       |/         <- greater is better
1590 		 *   -C  /             if scope is not enough for destination.
1591 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1592 		 *
1593 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1594 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1595 		 * Assume B = 0 and we get C > 29.
1596 		 */
1597 		ret = __ipv6_addr_src_scope(score->addr_type);
1598 		if (ret >= dst->scope)
1599 			ret = -ret;
1600 		else
1601 			ret -= 128;	/* 30 is enough */
1602 		score->scopedist = ret;
1603 		break;
1604 	case IPV6_SADDR_RULE_PREFERRED:
1605 	    {
1606 		/* Rule 3: Avoid deprecated and optimistic addresses */
1607 		u8 avoid = IFA_F_DEPRECATED;
1608 
1609 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1610 			avoid |= IFA_F_OPTIMISTIC;
1611 		ret = ipv6_saddr_preferred(score->addr_type) ||
1612 		      !(score->ifa->flags & avoid);
1613 		break;
1614 	    }
1615 #ifdef CONFIG_IPV6_MIP6
1616 	case IPV6_SADDR_RULE_HOA:
1617 	    {
1618 		/* Rule 4: Prefer home address */
1619 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1620 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1621 		break;
1622 	    }
1623 #endif
1624 	case IPV6_SADDR_RULE_OIF:
1625 		/* Rule 5: Prefer outgoing interface */
1626 		ret = (!dst->ifindex ||
1627 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1628 		break;
1629 	case IPV6_SADDR_RULE_LABEL:
1630 		/* Rule 6: Prefer matching label */
1631 		ret = ipv6_addr_label(net,
1632 				      &score->ifa->addr, score->addr_type,
1633 				      score->ifa->idev->dev->ifindex) == dst->label;
1634 		break;
1635 	case IPV6_SADDR_RULE_PRIVACY:
1636 	    {
1637 		/* Rule 7: Prefer public address
1638 		 * Note: prefer temporary address if use_tempaddr >= 2
1639 		 */
1640 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1641 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1642 				score->ifa->idev->cnf.use_tempaddr >= 2;
1643 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1644 		break;
1645 	    }
1646 	case IPV6_SADDR_RULE_ORCHID:
1647 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1648 		 *	    non-ORCHID vs non-ORCHID
1649 		 */
1650 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1651 			ipv6_addr_orchid(dst->addr));
1652 		break;
1653 	case IPV6_SADDR_RULE_PREFIX:
1654 		/* Rule 8: Use longest matching prefix */
1655 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1656 		if (ret > score->ifa->prefix_len)
1657 			ret = score->ifa->prefix_len;
1658 		score->matchlen = ret;
1659 		break;
1660 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1661 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1662 		/* Optimistic addresses still have lower precedence than other
1663 		 * preferred addresses.
1664 		 */
1665 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1666 		break;
1667 #endif
1668 	default:
1669 		ret = 0;
1670 	}
1671 
1672 	if (ret)
1673 		__set_bit(i, score->scorebits);
1674 	score->rule = i;
1675 out:
1676 	return ret;
1677 }
1678 
1679 static int __ipv6_dev_get_saddr(struct net *net,
1680 				struct ipv6_saddr_dst *dst,
1681 				struct inet6_dev *idev,
1682 				struct ipv6_saddr_score *scores,
1683 				int hiscore_idx)
1684 {
1685 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1686 
1687 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1688 		int i;
1689 
1690 		/*
1691 		 * - Tentative Address (RFC2462 section 5.4)
1692 		 *  - A tentative address is not considered
1693 		 *    "assigned to an interface" in the traditional
1694 		 *    sense, unless it is also flagged as optimistic.
1695 		 * - Candidate Source Address (section 4)
1696 		 *  - In any case, anycast addresses, multicast
1697 		 *    addresses, and the unspecified address MUST
1698 		 *    NOT be included in a candidate set.
1699 		 */
1700 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1701 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1702 			continue;
1703 
1704 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1705 
1706 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1707 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1708 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1709 					    idev->dev->name);
1710 			continue;
1711 		}
1712 
1713 		score->rule = -1;
1714 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1715 
1716 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1717 			int minihiscore, miniscore;
1718 
1719 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1720 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1721 
1722 			if (minihiscore > miniscore) {
1723 				if (i == IPV6_SADDR_RULE_SCOPE &&
1724 				    score->scopedist > 0) {
1725 					/*
1726 					 * special case:
1727 					 * each remaining entry
1728 					 * has too small (not enough)
1729 					 * scope, because ifa entries
1730 					 * are sorted by their scope
1731 					 * values.
1732 					 */
1733 					goto out;
1734 				}
1735 				break;
1736 			} else if (minihiscore < miniscore) {
1737 				swap(hiscore, score);
1738 				hiscore_idx = 1 - hiscore_idx;
1739 
1740 				/* restore our iterator */
1741 				score->ifa = hiscore->ifa;
1742 
1743 				break;
1744 			}
1745 		}
1746 	}
1747 out:
1748 	return hiscore_idx;
1749 }
1750 
1751 static int ipv6_get_saddr_master(struct net *net,
1752 				 const struct net_device *dst_dev,
1753 				 const struct net_device *master,
1754 				 struct ipv6_saddr_dst *dst,
1755 				 struct ipv6_saddr_score *scores,
1756 				 int hiscore_idx)
1757 {
1758 	struct inet6_dev *idev;
1759 
1760 	idev = __in6_dev_get(dst_dev);
1761 	if (idev)
1762 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1763 						   scores, hiscore_idx);
1764 
1765 	idev = __in6_dev_get(master);
1766 	if (idev)
1767 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1768 						   scores, hiscore_idx);
1769 
1770 	return hiscore_idx;
1771 }
1772 
1773 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1774 		       const struct in6_addr *daddr, unsigned int prefs,
1775 		       struct in6_addr *saddr)
1776 {
1777 	struct ipv6_saddr_score scores[2], *hiscore;
1778 	struct ipv6_saddr_dst dst;
1779 	struct inet6_dev *idev;
1780 	struct net_device *dev;
1781 	int dst_type;
1782 	bool use_oif_addr = false;
1783 	int hiscore_idx = 0;
1784 	int ret = 0;
1785 
1786 	dst_type = __ipv6_addr_type(daddr);
1787 	dst.addr = daddr;
1788 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1789 	dst.scope = __ipv6_addr_src_scope(dst_type);
1790 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1791 	dst.prefs = prefs;
1792 
1793 	scores[hiscore_idx].rule = -1;
1794 	scores[hiscore_idx].ifa = NULL;
1795 
1796 	rcu_read_lock();
1797 
1798 	/* Candidate Source Address (section 4)
1799 	 *  - multicast and link-local destination address,
1800 	 *    the set of candidate source address MUST only
1801 	 *    include addresses assigned to interfaces
1802 	 *    belonging to the same link as the outgoing
1803 	 *    interface.
1804 	 * (- For site-local destination addresses, the
1805 	 *    set of candidate source addresses MUST only
1806 	 *    include addresses assigned to interfaces
1807 	 *    belonging to the same site as the outgoing
1808 	 *    interface.)
1809 	 *  - "It is RECOMMENDED that the candidate source addresses
1810 	 *    be the set of unicast addresses assigned to the
1811 	 *    interface that will be used to send to the destination
1812 	 *    (the 'outgoing' interface)." (RFC 6724)
1813 	 */
1814 	if (dst_dev) {
1815 		idev = __in6_dev_get(dst_dev);
1816 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1817 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1818 		    (idev && idev->cnf.use_oif_addrs_only)) {
1819 			use_oif_addr = true;
1820 		}
1821 	}
1822 
1823 	if (use_oif_addr) {
1824 		if (idev)
1825 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1826 	} else {
1827 		const struct net_device *master;
1828 		int master_idx = 0;
1829 
1830 		/* if dst_dev exists and is enslaved to an L3 device, then
1831 		 * prefer addresses from dst_dev and then the master over
1832 		 * any other enslaved devices in the L3 domain.
1833 		 */
1834 		master = l3mdev_master_dev_rcu(dst_dev);
1835 		if (master) {
1836 			master_idx = master->ifindex;
1837 
1838 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1839 							    master, &dst,
1840 							    scores, hiscore_idx);
1841 
1842 			if (scores[hiscore_idx].ifa &&
1843 			    scores[hiscore_idx].scopedist >= 0)
1844 				goto out;
1845 		}
1846 
1847 		for_each_netdev_rcu(net, dev) {
1848 			/* only consider addresses on devices in the
1849 			 * same L3 domain
1850 			 */
1851 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1852 				continue;
1853 			idev = __in6_dev_get(dev);
1854 			if (!idev)
1855 				continue;
1856 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1857 		}
1858 	}
1859 
1860 out:
1861 	hiscore = &scores[hiscore_idx];
1862 	if (!hiscore->ifa)
1863 		ret = -EADDRNOTAVAIL;
1864 	else
1865 		*saddr = hiscore->ifa->addr;
1866 
1867 	rcu_read_unlock();
1868 	return ret;
1869 }
1870 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1871 
1872 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1873 			      u32 banned_flags)
1874 {
1875 	struct inet6_ifaddr *ifp;
1876 	int err = -EADDRNOTAVAIL;
1877 
1878 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1879 		if (ifp->scope > IFA_LINK)
1880 			break;
1881 		if (ifp->scope == IFA_LINK &&
1882 		    !(ifp->flags & banned_flags)) {
1883 			*addr = ifp->addr;
1884 			err = 0;
1885 			break;
1886 		}
1887 	}
1888 	return err;
1889 }
1890 
1891 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1892 		    u32 banned_flags)
1893 {
1894 	struct inet6_dev *idev;
1895 	int err = -EADDRNOTAVAIL;
1896 
1897 	rcu_read_lock();
1898 	idev = __in6_dev_get(dev);
1899 	if (idev) {
1900 		read_lock_bh(&idev->lock);
1901 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1902 		read_unlock_bh(&idev->lock);
1903 	}
1904 	rcu_read_unlock();
1905 	return err;
1906 }
1907 
1908 static int ipv6_count_addresses(const struct inet6_dev *idev)
1909 {
1910 	const struct inet6_ifaddr *ifp;
1911 	int cnt = 0;
1912 
1913 	rcu_read_lock();
1914 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1915 		cnt++;
1916 	rcu_read_unlock();
1917 	return cnt;
1918 }
1919 
1920 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1921 		  const struct net_device *dev, int strict)
1922 {
1923 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1924 				       strict, IFA_F_TENTATIVE);
1925 }
1926 EXPORT_SYMBOL(ipv6_chk_addr);
1927 
1928 /* device argument is used to find the L3 domain of interest. If
1929  * skip_dev_check is set, then the ifp device is not checked against
1930  * the passed in dev argument. So the 2 cases for addresses checks are:
1931  *   1. does the address exist in the L3 domain that dev is part of
1932  *      (skip_dev_check = true), or
1933  *
1934  *   2. does the address exist on the specific device
1935  *      (skip_dev_check = false)
1936  */
1937 static struct net_device *
1938 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1939 			  const struct net_device *dev, bool skip_dev_check,
1940 			  int strict, u32 banned_flags)
1941 {
1942 	unsigned int hash = inet6_addr_hash(net, addr);
1943 	struct net_device *l3mdev, *ndev;
1944 	struct inet6_ifaddr *ifp;
1945 	u32 ifp_flags;
1946 
1947 	rcu_read_lock();
1948 
1949 	l3mdev = l3mdev_master_dev_rcu(dev);
1950 	if (skip_dev_check)
1951 		dev = NULL;
1952 
1953 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1954 		ndev = ifp->idev->dev;
1955 
1956 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1957 			continue;
1958 
1959 		/* Decouple optimistic from tentative for evaluation here.
1960 		 * Ban optimistic addresses explicitly, when required.
1961 		 */
1962 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1963 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1964 			    : ifp->flags;
1965 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1966 		    !(ifp_flags&banned_flags) &&
1967 		    (!dev || ndev == dev ||
1968 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1969 			rcu_read_unlock();
1970 			return ndev;
1971 		}
1972 	}
1973 
1974 	rcu_read_unlock();
1975 	return NULL;
1976 }
1977 
1978 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1979 			    const struct net_device *dev, bool skip_dev_check,
1980 			    int strict, u32 banned_flags)
1981 {
1982 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1983 					 strict, banned_flags) ? 1 : 0;
1984 }
1985 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1986 
1987 
1988 /* Compares an address/prefix_len with addresses on device @dev.
1989  * If one is found it returns true.
1990  */
1991 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1992 	const unsigned int prefix_len, struct net_device *dev)
1993 {
1994 	const struct inet6_ifaddr *ifa;
1995 	const struct inet6_dev *idev;
1996 	bool ret = false;
1997 
1998 	rcu_read_lock();
1999 	idev = __in6_dev_get(dev);
2000 	if (idev) {
2001 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2002 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2003 			if (ret)
2004 				break;
2005 		}
2006 	}
2007 	rcu_read_unlock();
2008 
2009 	return ret;
2010 }
2011 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2012 
2013 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2014 {
2015 	const struct inet6_ifaddr *ifa;
2016 	const struct inet6_dev *idev;
2017 	int	onlink;
2018 
2019 	onlink = 0;
2020 	rcu_read_lock();
2021 	idev = __in6_dev_get(dev);
2022 	if (idev) {
2023 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2024 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
2025 						   ifa->prefix_len);
2026 			if (onlink)
2027 				break;
2028 		}
2029 	}
2030 	rcu_read_unlock();
2031 	return onlink;
2032 }
2033 EXPORT_SYMBOL(ipv6_chk_prefix);
2034 
2035 /**
2036  * ipv6_dev_find - find the first device with a given source address.
2037  * @net: the net namespace
2038  * @addr: the source address
2039  * @dev: used to find the L3 domain of interest
2040  *
2041  * The caller should be protected by RCU, or RTNL.
2042  */
2043 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2044 				 struct net_device *dev)
2045 {
2046 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2047 					 IFA_F_TENTATIVE);
2048 }
2049 EXPORT_SYMBOL(ipv6_dev_find);
2050 
2051 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2052 				     struct net_device *dev, int strict)
2053 {
2054 	unsigned int hash = inet6_addr_hash(net, addr);
2055 	struct inet6_ifaddr *ifp, *result = NULL;
2056 
2057 	rcu_read_lock();
2058 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2059 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2060 			if (!dev || ifp->idev->dev == dev ||
2061 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2062 				if (in6_ifa_hold_safe(ifp)) {
2063 					result = ifp;
2064 					break;
2065 				}
2066 			}
2067 		}
2068 	}
2069 	rcu_read_unlock();
2070 
2071 	return result;
2072 }
2073 
2074 /* Gets referenced address, destroys ifaddr */
2075 
2076 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2077 {
2078 	if (dad_failed)
2079 		ifp->flags |= IFA_F_DADFAILED;
2080 
2081 	if (ifp->flags&IFA_F_TEMPORARY) {
2082 		struct inet6_ifaddr *ifpub;
2083 		spin_lock_bh(&ifp->lock);
2084 		ifpub = ifp->ifpub;
2085 		if (ifpub) {
2086 			in6_ifa_hold(ifpub);
2087 			spin_unlock_bh(&ifp->lock);
2088 			ipv6_create_tempaddr(ifpub, true);
2089 			in6_ifa_put(ifpub);
2090 		} else {
2091 			spin_unlock_bh(&ifp->lock);
2092 		}
2093 		ipv6_del_addr(ifp);
2094 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2095 		spin_lock_bh(&ifp->lock);
2096 		addrconf_del_dad_work(ifp);
2097 		ifp->flags |= IFA_F_TENTATIVE;
2098 		if (dad_failed)
2099 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2100 		spin_unlock_bh(&ifp->lock);
2101 		if (dad_failed)
2102 			ipv6_ifa_notify(0, ifp);
2103 		in6_ifa_put(ifp);
2104 	} else {
2105 		ipv6_del_addr(ifp);
2106 	}
2107 }
2108 
2109 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2110 {
2111 	int err = -ENOENT;
2112 
2113 	spin_lock_bh(&ifp->lock);
2114 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2115 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2116 		err = 0;
2117 	}
2118 	spin_unlock_bh(&ifp->lock);
2119 
2120 	return err;
2121 }
2122 
2123 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2124 {
2125 	struct inet6_dev *idev = ifp->idev;
2126 	struct net *net = dev_net(idev->dev);
2127 
2128 	if (addrconf_dad_end(ifp)) {
2129 		in6_ifa_put(ifp);
2130 		return;
2131 	}
2132 
2133 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2134 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2135 
2136 	spin_lock_bh(&ifp->lock);
2137 
2138 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2139 		struct in6_addr new_addr;
2140 		struct inet6_ifaddr *ifp2;
2141 		int retries = ifp->stable_privacy_retry + 1;
2142 		struct ifa6_config cfg = {
2143 			.pfx = &new_addr,
2144 			.plen = ifp->prefix_len,
2145 			.ifa_flags = ifp->flags,
2146 			.valid_lft = ifp->valid_lft,
2147 			.preferred_lft = ifp->prefered_lft,
2148 			.scope = ifp->scope,
2149 		};
2150 
2151 		if (retries > net->ipv6.sysctl.idgen_retries) {
2152 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2153 					     ifp->idev->dev->name);
2154 			goto errdad;
2155 		}
2156 
2157 		new_addr = ifp->addr;
2158 		if (ipv6_generate_stable_address(&new_addr, retries,
2159 						 idev))
2160 			goto errdad;
2161 
2162 		spin_unlock_bh(&ifp->lock);
2163 
2164 		if (idev->cnf.max_addresses &&
2165 		    ipv6_count_addresses(idev) >=
2166 		    idev->cnf.max_addresses)
2167 			goto lock_errdad;
2168 
2169 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2170 				     ifp->idev->dev->name);
2171 
2172 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2173 		if (IS_ERR(ifp2))
2174 			goto lock_errdad;
2175 
2176 		spin_lock_bh(&ifp2->lock);
2177 		ifp2->stable_privacy_retry = retries;
2178 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2179 		spin_unlock_bh(&ifp2->lock);
2180 
2181 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2182 		in6_ifa_put(ifp2);
2183 lock_errdad:
2184 		spin_lock_bh(&ifp->lock);
2185 	}
2186 
2187 errdad:
2188 	/* transition from _POSTDAD to _ERRDAD */
2189 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2190 	spin_unlock_bh(&ifp->lock);
2191 
2192 	addrconf_mod_dad_work(ifp, 0);
2193 	in6_ifa_put(ifp);
2194 }
2195 
2196 /* Join to solicited addr multicast group.
2197  * caller must hold RTNL */
2198 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2199 {
2200 	struct in6_addr maddr;
2201 
2202 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2203 		return;
2204 
2205 	addrconf_addr_solict_mult(addr, &maddr);
2206 	ipv6_dev_mc_inc(dev, &maddr);
2207 }
2208 
2209 /* caller must hold RTNL */
2210 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2211 {
2212 	struct in6_addr maddr;
2213 
2214 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2215 		return;
2216 
2217 	addrconf_addr_solict_mult(addr, &maddr);
2218 	__ipv6_dev_mc_dec(idev, &maddr);
2219 }
2220 
2221 /* caller must hold RTNL */
2222 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2223 {
2224 	struct in6_addr addr;
2225 
2226 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2227 		return;
2228 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2229 	if (ipv6_addr_any(&addr))
2230 		return;
2231 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2232 }
2233 
2234 /* caller must hold RTNL */
2235 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2236 {
2237 	struct in6_addr addr;
2238 
2239 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2240 		return;
2241 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2242 	if (ipv6_addr_any(&addr))
2243 		return;
2244 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2245 }
2246 
2247 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2248 {
2249 	switch (dev->addr_len) {
2250 	case ETH_ALEN:
2251 		memcpy(eui, dev->dev_addr, 3);
2252 		eui[3] = 0xFF;
2253 		eui[4] = 0xFE;
2254 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2255 		break;
2256 	case EUI64_ADDR_LEN:
2257 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2258 		eui[0] ^= 2;
2259 		break;
2260 	default:
2261 		return -1;
2262 	}
2263 
2264 	return 0;
2265 }
2266 
2267 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2268 {
2269 	const union fwnet_hwaddr *ha;
2270 
2271 	if (dev->addr_len != FWNET_ALEN)
2272 		return -1;
2273 
2274 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2275 
2276 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2277 	eui[0] ^= 2;
2278 	return 0;
2279 }
2280 
2281 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2282 {
2283 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2284 	if (dev->addr_len != ARCNET_ALEN)
2285 		return -1;
2286 	memset(eui, 0, 7);
2287 	eui[7] = *(u8 *)dev->dev_addr;
2288 	return 0;
2289 }
2290 
2291 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2292 {
2293 	if (dev->addr_len != INFINIBAND_ALEN)
2294 		return -1;
2295 	memcpy(eui, dev->dev_addr + 12, 8);
2296 	eui[0] |= 2;
2297 	return 0;
2298 }
2299 
2300 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2301 {
2302 	if (addr == 0)
2303 		return -1;
2304 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2305 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2306 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2307 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2308 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2309 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2310 	eui[1] = 0;
2311 	eui[2] = 0x5E;
2312 	eui[3] = 0xFE;
2313 	memcpy(eui + 4, &addr, 4);
2314 	return 0;
2315 }
2316 
2317 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2318 {
2319 	if (dev->priv_flags & IFF_ISATAP)
2320 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2321 	return -1;
2322 }
2323 
2324 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2325 {
2326 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2327 }
2328 
2329 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2330 {
2331 	memcpy(eui, dev->perm_addr, 3);
2332 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2333 	eui[3] = 0xFF;
2334 	eui[4] = 0xFE;
2335 	eui[0] ^= 2;
2336 	return 0;
2337 }
2338 
2339 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2340 {
2341 	switch (dev->type) {
2342 	case ARPHRD_ETHER:
2343 	case ARPHRD_FDDI:
2344 		return addrconf_ifid_eui48(eui, dev);
2345 	case ARPHRD_ARCNET:
2346 		return addrconf_ifid_arcnet(eui, dev);
2347 	case ARPHRD_INFINIBAND:
2348 		return addrconf_ifid_infiniband(eui, dev);
2349 	case ARPHRD_SIT:
2350 		return addrconf_ifid_sit(eui, dev);
2351 	case ARPHRD_IPGRE:
2352 	case ARPHRD_TUNNEL:
2353 		return addrconf_ifid_gre(eui, dev);
2354 	case ARPHRD_6LOWPAN:
2355 		return addrconf_ifid_6lowpan(eui, dev);
2356 	case ARPHRD_IEEE1394:
2357 		return addrconf_ifid_ieee1394(eui, dev);
2358 	case ARPHRD_TUNNEL6:
2359 	case ARPHRD_IP6GRE:
2360 	case ARPHRD_RAWIP:
2361 		return addrconf_ifid_ip6tnl(eui, dev);
2362 	}
2363 	return -1;
2364 }
2365 
2366 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2367 {
2368 	int err = -1;
2369 	struct inet6_ifaddr *ifp;
2370 
2371 	read_lock_bh(&idev->lock);
2372 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2373 		if (ifp->scope > IFA_LINK)
2374 			break;
2375 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2376 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2377 			err = 0;
2378 			break;
2379 		}
2380 	}
2381 	read_unlock_bh(&idev->lock);
2382 	return err;
2383 }
2384 
2385 /* Generation of a randomized Interface Identifier
2386  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2387  */
2388 
2389 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2390 {
2391 regen:
2392 	get_random_bytes(&addr->s6_addr[8], 8);
2393 
2394 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2395 	 * check if generated address is not inappropriate:
2396 	 *
2397 	 * - Reserved IPv6 Interface Identifiers
2398 	 * - XXX: already assigned to an address on the device
2399 	 */
2400 
2401 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2402 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2403 		goto regen;
2404 
2405 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2406 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2407 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2408 	 */
2409 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2410 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2411 		goto regen;
2412 
2413 	/* Reserved subnet anycast addresses */
2414 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2415 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2416 		goto regen;
2417 }
2418 
2419 /*
2420  *	Add prefix route.
2421  */
2422 
2423 static void
2424 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2425 		      struct net_device *dev, unsigned long expires,
2426 		      u32 flags, gfp_t gfp_flags)
2427 {
2428 	struct fib6_config cfg = {
2429 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2430 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2431 		.fc_ifindex = dev->ifindex,
2432 		.fc_expires = expires,
2433 		.fc_dst_len = plen,
2434 		.fc_flags = RTF_UP | flags,
2435 		.fc_nlinfo.nl_net = dev_net(dev),
2436 		.fc_protocol = RTPROT_KERNEL,
2437 		.fc_type = RTN_UNICAST,
2438 	};
2439 
2440 	cfg.fc_dst = *pfx;
2441 
2442 	/* Prevent useless cloning on PtP SIT.
2443 	   This thing is done here expecting that the whole
2444 	   class of non-broadcast devices need not cloning.
2445 	 */
2446 #if IS_ENABLED(CONFIG_IPV6_SIT)
2447 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2448 		cfg.fc_flags |= RTF_NONEXTHOP;
2449 #endif
2450 
2451 	ip6_route_add(&cfg, gfp_flags, NULL);
2452 }
2453 
2454 
2455 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2456 						  int plen,
2457 						  const struct net_device *dev,
2458 						  u32 flags, u32 noflags,
2459 						  bool no_gw)
2460 {
2461 	struct fib6_node *fn;
2462 	struct fib6_info *rt = NULL;
2463 	struct fib6_table *table;
2464 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2465 
2466 	table = fib6_get_table(dev_net(dev), tb_id);
2467 	if (!table)
2468 		return NULL;
2469 
2470 	rcu_read_lock();
2471 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2472 	if (!fn)
2473 		goto out;
2474 
2475 	for_each_fib6_node_rt_rcu(fn) {
2476 		/* prefix routes only use builtin fib6_nh */
2477 		if (rt->nh)
2478 			continue;
2479 
2480 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2481 			continue;
2482 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2483 			continue;
2484 		if ((rt->fib6_flags & flags) != flags)
2485 			continue;
2486 		if ((rt->fib6_flags & noflags) != 0)
2487 			continue;
2488 		if (!fib6_info_hold_safe(rt))
2489 			continue;
2490 		break;
2491 	}
2492 out:
2493 	rcu_read_unlock();
2494 	return rt;
2495 }
2496 
2497 
2498 /* Create "default" multicast route to the interface */
2499 
2500 static void addrconf_add_mroute(struct net_device *dev)
2501 {
2502 	struct fib6_config cfg = {
2503 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2504 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2505 		.fc_ifindex = dev->ifindex,
2506 		.fc_dst_len = 8,
2507 		.fc_flags = RTF_UP,
2508 		.fc_type = RTN_MULTICAST,
2509 		.fc_nlinfo.nl_net = dev_net(dev),
2510 		.fc_protocol = RTPROT_KERNEL,
2511 	};
2512 
2513 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2514 
2515 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2516 }
2517 
2518 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2519 {
2520 	struct inet6_dev *idev;
2521 
2522 	ASSERT_RTNL();
2523 
2524 	idev = ipv6_find_idev(dev);
2525 	if (IS_ERR(idev))
2526 		return idev;
2527 
2528 	if (idev->cnf.disable_ipv6)
2529 		return ERR_PTR(-EACCES);
2530 
2531 	/* Add default multicast route */
2532 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2533 		addrconf_add_mroute(dev);
2534 
2535 	return idev;
2536 }
2537 
2538 static void delete_tempaddrs(struct inet6_dev *idev,
2539 			     struct inet6_ifaddr *ifp)
2540 {
2541 	struct inet6_ifaddr *ift, *tmp;
2542 
2543 	write_lock_bh(&idev->lock);
2544 	list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) {
2545 		if (ift->ifpub != ifp)
2546 			continue;
2547 
2548 		in6_ifa_hold(ift);
2549 		write_unlock_bh(&idev->lock);
2550 		ipv6_del_addr(ift);
2551 		write_lock_bh(&idev->lock);
2552 	}
2553 	write_unlock_bh(&idev->lock);
2554 }
2555 
2556 static void manage_tempaddrs(struct inet6_dev *idev,
2557 			     struct inet6_ifaddr *ifp,
2558 			     __u32 valid_lft, __u32 prefered_lft,
2559 			     bool create, unsigned long now)
2560 {
2561 	u32 flags;
2562 	struct inet6_ifaddr *ift;
2563 
2564 	read_lock_bh(&idev->lock);
2565 	/* update all temporary addresses in the list */
2566 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2567 		int age, max_valid, max_prefered;
2568 
2569 		if (ifp != ift->ifpub)
2570 			continue;
2571 
2572 		/* RFC 4941 section 3.3:
2573 		 * If a received option will extend the lifetime of a public
2574 		 * address, the lifetimes of temporary addresses should
2575 		 * be extended, subject to the overall constraint that no
2576 		 * temporary addresses should ever remain "valid" or "preferred"
2577 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2578 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2579 		 */
2580 		age = (now - ift->cstamp) / HZ;
2581 		max_valid = idev->cnf.temp_valid_lft - age;
2582 		if (max_valid < 0)
2583 			max_valid = 0;
2584 
2585 		max_prefered = idev->cnf.temp_prefered_lft -
2586 			       idev->desync_factor - age;
2587 		if (max_prefered < 0)
2588 			max_prefered = 0;
2589 
2590 		if (valid_lft > max_valid)
2591 			valid_lft = max_valid;
2592 
2593 		if (prefered_lft > max_prefered)
2594 			prefered_lft = max_prefered;
2595 
2596 		spin_lock(&ift->lock);
2597 		flags = ift->flags;
2598 		ift->valid_lft = valid_lft;
2599 		ift->prefered_lft = prefered_lft;
2600 		ift->tstamp = now;
2601 		if (prefered_lft > 0)
2602 			ift->flags &= ~IFA_F_DEPRECATED;
2603 
2604 		spin_unlock(&ift->lock);
2605 		if (!(flags&IFA_F_TENTATIVE))
2606 			ipv6_ifa_notify(0, ift);
2607 	}
2608 
2609 	/* Also create a temporary address if it's enabled but no temporary
2610 	 * address currently exists.
2611 	 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2612 	 * as part of cleanup (ie. deleting the mngtmpaddr).
2613 	 * We don't want that to result in creating a new temporary ip address.
2614 	 */
2615 	if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2616 		create = true;
2617 
2618 	if (create && idev->cnf.use_tempaddr > 0) {
2619 		/* When a new public address is created as described
2620 		 * in [ADDRCONF], also create a new temporary address.
2621 		 */
2622 		read_unlock_bh(&idev->lock);
2623 		ipv6_create_tempaddr(ifp, false);
2624 	} else {
2625 		read_unlock_bh(&idev->lock);
2626 	}
2627 }
2628 
2629 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2630 {
2631 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2632 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2633 }
2634 
2635 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2636 				 const struct prefix_info *pinfo,
2637 				 struct inet6_dev *in6_dev,
2638 				 const struct in6_addr *addr, int addr_type,
2639 				 u32 addr_flags, bool sllao, bool tokenized,
2640 				 __u32 valid_lft, u32 prefered_lft)
2641 {
2642 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2643 	int create = 0, update_lft = 0;
2644 
2645 	if (!ifp && valid_lft) {
2646 		int max_addresses = in6_dev->cnf.max_addresses;
2647 		struct ifa6_config cfg = {
2648 			.pfx = addr,
2649 			.plen = pinfo->prefix_len,
2650 			.ifa_flags = addr_flags,
2651 			.valid_lft = valid_lft,
2652 			.preferred_lft = prefered_lft,
2653 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2654 			.ifa_proto = IFAPROT_KERNEL_RA
2655 		};
2656 
2657 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2658 		if ((net->ipv6.devconf_all->optimistic_dad ||
2659 		     in6_dev->cnf.optimistic_dad) &&
2660 		    !net->ipv6.devconf_all->forwarding && sllao)
2661 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2662 #endif
2663 
2664 		/* Do not allow to create too much of autoconfigured
2665 		 * addresses; this would be too easy way to crash kernel.
2666 		 */
2667 		if (!max_addresses ||
2668 		    ipv6_count_addresses(in6_dev) < max_addresses)
2669 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2670 
2671 		if (IS_ERR_OR_NULL(ifp))
2672 			return -1;
2673 
2674 		create = 1;
2675 		spin_lock_bh(&ifp->lock);
2676 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2677 		ifp->cstamp = jiffies;
2678 		ifp->tokenized = tokenized;
2679 		spin_unlock_bh(&ifp->lock);
2680 		addrconf_dad_start(ifp);
2681 	}
2682 
2683 	if (ifp) {
2684 		u32 flags;
2685 		unsigned long now;
2686 		u32 stored_lft;
2687 
2688 		/* update lifetime (RFC2462 5.5.3 e) */
2689 		spin_lock_bh(&ifp->lock);
2690 		now = jiffies;
2691 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2692 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2693 		else
2694 			stored_lft = 0;
2695 		if (!create && stored_lft) {
2696 			const u32 minimum_lft = min_t(u32,
2697 				stored_lft, MIN_VALID_LIFETIME);
2698 			valid_lft = max(valid_lft, minimum_lft);
2699 
2700 			/* RFC4862 Section 5.5.3e:
2701 			 * "Note that the preferred lifetime of the
2702 			 *  corresponding address is always reset to
2703 			 *  the Preferred Lifetime in the received
2704 			 *  Prefix Information option, regardless of
2705 			 *  whether the valid lifetime is also reset or
2706 			 *  ignored."
2707 			 *
2708 			 * So we should always update prefered_lft here.
2709 			 */
2710 			update_lft = 1;
2711 		}
2712 
2713 		if (update_lft) {
2714 			ifp->valid_lft = valid_lft;
2715 			ifp->prefered_lft = prefered_lft;
2716 			ifp->tstamp = now;
2717 			flags = ifp->flags;
2718 			ifp->flags &= ~IFA_F_DEPRECATED;
2719 			spin_unlock_bh(&ifp->lock);
2720 
2721 			if (!(flags&IFA_F_TENTATIVE))
2722 				ipv6_ifa_notify(0, ifp);
2723 		} else
2724 			spin_unlock_bh(&ifp->lock);
2725 
2726 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2727 				 create, now);
2728 
2729 		in6_ifa_put(ifp);
2730 		addrconf_verify(net);
2731 	}
2732 
2733 	return 0;
2734 }
2735 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2736 
2737 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2738 {
2739 	struct prefix_info *pinfo;
2740 	__u32 valid_lft;
2741 	__u32 prefered_lft;
2742 	int addr_type, err;
2743 	u32 addr_flags = 0;
2744 	struct inet6_dev *in6_dev;
2745 	struct net *net = dev_net(dev);
2746 
2747 	pinfo = (struct prefix_info *) opt;
2748 
2749 	if (len < sizeof(struct prefix_info)) {
2750 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2751 		return;
2752 	}
2753 
2754 	/*
2755 	 *	Validation checks ([ADDRCONF], page 19)
2756 	 */
2757 
2758 	addr_type = ipv6_addr_type(&pinfo->prefix);
2759 
2760 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2761 		return;
2762 
2763 	valid_lft = ntohl(pinfo->valid);
2764 	prefered_lft = ntohl(pinfo->prefered);
2765 
2766 	if (prefered_lft > valid_lft) {
2767 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2768 		return;
2769 	}
2770 
2771 	in6_dev = in6_dev_get(dev);
2772 
2773 	if (!in6_dev) {
2774 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2775 				    dev->name);
2776 		return;
2777 	}
2778 
2779 	if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2780 		goto put;
2781 
2782 	/*
2783 	 *	Two things going on here:
2784 	 *	1) Add routes for on-link prefixes
2785 	 *	2) Configure prefixes with the auto flag set
2786 	 */
2787 
2788 	if (pinfo->onlink) {
2789 		struct fib6_info *rt;
2790 		unsigned long rt_expires;
2791 
2792 		/* Avoid arithmetic overflow. Really, we could
2793 		 * save rt_expires in seconds, likely valid_lft,
2794 		 * but it would require division in fib gc, that it
2795 		 * not good.
2796 		 */
2797 		if (HZ > USER_HZ)
2798 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2799 		else
2800 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2801 
2802 		if (addrconf_finite_timeout(rt_expires))
2803 			rt_expires *= HZ;
2804 
2805 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2806 					       pinfo->prefix_len,
2807 					       dev,
2808 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2809 					       RTF_DEFAULT, true);
2810 
2811 		if (rt) {
2812 			/* Autoconf prefix route */
2813 			if (valid_lft == 0) {
2814 				ip6_del_rt(net, rt, false);
2815 				rt = NULL;
2816 			} else if (addrconf_finite_timeout(rt_expires)) {
2817 				/* not infinity */
2818 				fib6_set_expires(rt, jiffies + rt_expires);
2819 			} else {
2820 				fib6_clean_expires(rt);
2821 			}
2822 		} else if (valid_lft) {
2823 			clock_t expires = 0;
2824 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2825 			if (addrconf_finite_timeout(rt_expires)) {
2826 				/* not infinity */
2827 				flags |= RTF_EXPIRES;
2828 				expires = jiffies_to_clock_t(rt_expires);
2829 			}
2830 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2831 					      0, dev, expires, flags,
2832 					      GFP_ATOMIC);
2833 		}
2834 		fib6_info_release(rt);
2835 	}
2836 
2837 	/* Try to figure out our local address for this prefix */
2838 
2839 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2840 		struct in6_addr addr;
2841 		bool tokenized = false, dev_addr_generated = false;
2842 
2843 		if (pinfo->prefix_len == 64) {
2844 			memcpy(&addr, &pinfo->prefix, 8);
2845 
2846 			if (!ipv6_addr_any(&in6_dev->token)) {
2847 				read_lock_bh(&in6_dev->lock);
2848 				memcpy(addr.s6_addr + 8,
2849 				       in6_dev->token.s6_addr + 8, 8);
2850 				read_unlock_bh(&in6_dev->lock);
2851 				tokenized = true;
2852 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2853 				   !ipv6_generate_stable_address(&addr, 0,
2854 								 in6_dev)) {
2855 				addr_flags |= IFA_F_STABLE_PRIVACY;
2856 				goto ok;
2857 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2858 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2859 				goto put;
2860 			} else {
2861 				dev_addr_generated = true;
2862 			}
2863 			goto ok;
2864 		}
2865 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2866 				    pinfo->prefix_len);
2867 		goto put;
2868 
2869 ok:
2870 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2871 						   &addr, addr_type,
2872 						   addr_flags, sllao,
2873 						   tokenized, valid_lft,
2874 						   prefered_lft);
2875 		if (err)
2876 			goto put;
2877 
2878 		/* Ignore error case here because previous prefix add addr was
2879 		 * successful which will be notified.
2880 		 */
2881 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2882 					      addr_type, addr_flags, sllao,
2883 					      tokenized, valid_lft,
2884 					      prefered_lft,
2885 					      dev_addr_generated);
2886 	}
2887 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2888 put:
2889 	in6_dev_put(in6_dev);
2890 }
2891 
2892 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2893 		struct in6_ifreq *ireq)
2894 {
2895 	struct ip_tunnel_parm p = { };
2896 	int err;
2897 
2898 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2899 		return -EADDRNOTAVAIL;
2900 
2901 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2902 	p.iph.version = 4;
2903 	p.iph.ihl = 5;
2904 	p.iph.protocol = IPPROTO_IPV6;
2905 	p.iph.ttl = 64;
2906 
2907 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2908 		return -EOPNOTSUPP;
2909 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2910 	if (err)
2911 		return err;
2912 
2913 	dev = __dev_get_by_name(net, p.name);
2914 	if (!dev)
2915 		return -ENOBUFS;
2916 	return dev_open(dev, NULL);
2917 }
2918 
2919 /*
2920  *	Set destination address.
2921  *	Special case for SIT interfaces where we create a new "virtual"
2922  *	device.
2923  */
2924 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2925 {
2926 	struct net_device *dev;
2927 	struct in6_ifreq ireq;
2928 	int err = -ENODEV;
2929 
2930 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2931 		return -ENODEV;
2932 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2933 		return -EFAULT;
2934 
2935 	rtnl_lock();
2936 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2937 	if (dev && dev->type == ARPHRD_SIT)
2938 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2939 	rtnl_unlock();
2940 	return err;
2941 }
2942 
2943 static int ipv6_mc_config(struct sock *sk, bool join,
2944 			  const struct in6_addr *addr, int ifindex)
2945 {
2946 	int ret;
2947 
2948 	ASSERT_RTNL();
2949 
2950 	lock_sock(sk);
2951 	if (join)
2952 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2953 	else
2954 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2955 	release_sock(sk);
2956 
2957 	return ret;
2958 }
2959 
2960 /*
2961  *	Manual configuration of address on an interface
2962  */
2963 static int inet6_addr_add(struct net *net, int ifindex,
2964 			  struct ifa6_config *cfg,
2965 			  struct netlink_ext_ack *extack)
2966 {
2967 	struct inet6_ifaddr *ifp;
2968 	struct inet6_dev *idev;
2969 	struct net_device *dev;
2970 	unsigned long timeout;
2971 	clock_t expires;
2972 	u32 flags;
2973 
2974 	ASSERT_RTNL();
2975 
2976 	if (cfg->plen > 128) {
2977 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
2978 		return -EINVAL;
2979 	}
2980 
2981 	/* check the lifetime */
2982 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
2983 		NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
2984 		return -EINVAL;
2985 	}
2986 
2987 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
2988 		NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
2989 		return -EINVAL;
2990 	}
2991 
2992 	dev = __dev_get_by_index(net, ifindex);
2993 	if (!dev)
2994 		return -ENODEV;
2995 
2996 	idev = addrconf_add_dev(dev);
2997 	if (IS_ERR(idev)) {
2998 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
2999 		return PTR_ERR(idev);
3000 	}
3001 
3002 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3003 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3004 					 true, cfg->pfx, ifindex);
3005 
3006 		if (ret < 0) {
3007 			NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3008 			return ret;
3009 		}
3010 	}
3011 
3012 	cfg->scope = ipv6_addr_scope(cfg->pfx);
3013 
3014 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
3015 	if (addrconf_finite_timeout(timeout)) {
3016 		expires = jiffies_to_clock_t(timeout * HZ);
3017 		cfg->valid_lft = timeout;
3018 		flags = RTF_EXPIRES;
3019 	} else {
3020 		expires = 0;
3021 		flags = 0;
3022 		cfg->ifa_flags |= IFA_F_PERMANENT;
3023 	}
3024 
3025 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3026 	if (addrconf_finite_timeout(timeout)) {
3027 		if (timeout == 0)
3028 			cfg->ifa_flags |= IFA_F_DEPRECATED;
3029 		cfg->preferred_lft = timeout;
3030 	}
3031 
3032 	ifp = ipv6_add_addr(idev, cfg, true, extack);
3033 	if (!IS_ERR(ifp)) {
3034 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3035 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3036 					      ifp->rt_priority, dev, expires,
3037 					      flags, GFP_KERNEL);
3038 		}
3039 
3040 		/* Send a netlink notification if DAD is enabled and
3041 		 * optimistic flag is not set
3042 		 */
3043 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3044 			ipv6_ifa_notify(0, ifp);
3045 		/*
3046 		 * Note that section 3.1 of RFC 4429 indicates
3047 		 * that the Optimistic flag should not be set for
3048 		 * manually configured addresses
3049 		 */
3050 		addrconf_dad_start(ifp);
3051 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3052 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
3053 					 cfg->preferred_lft, true, jiffies);
3054 		in6_ifa_put(ifp);
3055 		addrconf_verify_rtnl(net);
3056 		return 0;
3057 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3058 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3059 			       cfg->pfx, ifindex);
3060 	}
3061 
3062 	return PTR_ERR(ifp);
3063 }
3064 
3065 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3066 			  const struct in6_addr *pfx, unsigned int plen,
3067 			  struct netlink_ext_ack *extack)
3068 {
3069 	struct inet6_ifaddr *ifp;
3070 	struct inet6_dev *idev;
3071 	struct net_device *dev;
3072 
3073 	if (plen > 128) {
3074 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3075 		return -EINVAL;
3076 	}
3077 
3078 	dev = __dev_get_by_index(net, ifindex);
3079 	if (!dev) {
3080 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3081 		return -ENODEV;
3082 	}
3083 
3084 	idev = __in6_dev_get(dev);
3085 	if (!idev) {
3086 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3087 		return -ENXIO;
3088 	}
3089 
3090 	read_lock_bh(&idev->lock);
3091 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3092 		if (ifp->prefix_len == plen &&
3093 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3094 			in6_ifa_hold(ifp);
3095 			read_unlock_bh(&idev->lock);
3096 
3097 			ipv6_del_addr(ifp);
3098 
3099 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3100 			    (ifp->flags & IFA_F_MANAGETEMPADDR))
3101 				delete_tempaddrs(idev, ifp);
3102 
3103 			addrconf_verify_rtnl(net);
3104 			if (ipv6_addr_is_multicast(pfx)) {
3105 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3106 					       false, pfx, dev->ifindex);
3107 			}
3108 			return 0;
3109 		}
3110 	}
3111 	read_unlock_bh(&idev->lock);
3112 
3113 	NL_SET_ERR_MSG_MOD(extack, "address not found");
3114 	return -EADDRNOTAVAIL;
3115 }
3116 
3117 
3118 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3119 {
3120 	struct ifa6_config cfg = {
3121 		.ifa_flags = IFA_F_PERMANENT,
3122 		.preferred_lft = INFINITY_LIFE_TIME,
3123 		.valid_lft = INFINITY_LIFE_TIME,
3124 	};
3125 	struct in6_ifreq ireq;
3126 	int err;
3127 
3128 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3129 		return -EPERM;
3130 
3131 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3132 		return -EFAULT;
3133 
3134 	cfg.pfx = &ireq.ifr6_addr;
3135 	cfg.plen = ireq.ifr6_prefixlen;
3136 
3137 	rtnl_lock();
3138 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3139 	rtnl_unlock();
3140 	return err;
3141 }
3142 
3143 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3144 {
3145 	struct in6_ifreq ireq;
3146 	int err;
3147 
3148 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3149 		return -EPERM;
3150 
3151 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3152 		return -EFAULT;
3153 
3154 	rtnl_lock();
3155 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3156 			     ireq.ifr6_prefixlen, NULL);
3157 	rtnl_unlock();
3158 	return err;
3159 }
3160 
3161 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3162 		     int plen, int scope, u8 proto)
3163 {
3164 	struct inet6_ifaddr *ifp;
3165 	struct ifa6_config cfg = {
3166 		.pfx = addr,
3167 		.plen = plen,
3168 		.ifa_flags = IFA_F_PERMANENT,
3169 		.valid_lft = INFINITY_LIFE_TIME,
3170 		.preferred_lft = INFINITY_LIFE_TIME,
3171 		.scope = scope,
3172 		.ifa_proto = proto
3173 	};
3174 
3175 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3176 	if (!IS_ERR(ifp)) {
3177 		spin_lock_bh(&ifp->lock);
3178 		ifp->flags &= ~IFA_F_TENTATIVE;
3179 		spin_unlock_bh(&ifp->lock);
3180 		rt_genid_bump_ipv6(dev_net(idev->dev));
3181 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3182 		in6_ifa_put(ifp);
3183 	}
3184 }
3185 
3186 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3187 static void add_v4_addrs(struct inet6_dev *idev)
3188 {
3189 	struct in6_addr addr;
3190 	struct net_device *dev;
3191 	struct net *net = dev_net(idev->dev);
3192 	int scope, plen, offset = 0;
3193 	u32 pflags = 0;
3194 
3195 	ASSERT_RTNL();
3196 
3197 	memset(&addr, 0, sizeof(struct in6_addr));
3198 	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3199 	if (idev->dev->addr_len == sizeof(struct in6_addr))
3200 		offset = sizeof(struct in6_addr) - 4;
3201 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3202 
3203 	if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3204 		scope = IPV6_ADDR_COMPATv4;
3205 		plen = 96;
3206 		pflags |= RTF_NONEXTHOP;
3207 	} else {
3208 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3209 			return;
3210 
3211 		addr.s6_addr32[0] = htonl(0xfe800000);
3212 		scope = IFA_LINK;
3213 		plen = 64;
3214 	}
3215 
3216 	if (addr.s6_addr32[3]) {
3217 		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3218 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3219 				      GFP_KERNEL);
3220 		return;
3221 	}
3222 
3223 	for_each_netdev(net, dev) {
3224 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3225 		if (in_dev && (dev->flags & IFF_UP)) {
3226 			struct in_ifaddr *ifa;
3227 			int flag = scope;
3228 
3229 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3230 				addr.s6_addr32[3] = ifa->ifa_local;
3231 
3232 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3233 					continue;
3234 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3235 					if (idev->dev->flags&IFF_POINTOPOINT)
3236 						continue;
3237 					flag |= IFA_HOST;
3238 				}
3239 
3240 				add_addr(idev, &addr, plen, flag,
3241 					 IFAPROT_UNSPEC);
3242 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3243 						      0, pflags, GFP_KERNEL);
3244 			}
3245 		}
3246 	}
3247 }
3248 #endif
3249 
3250 static void init_loopback(struct net_device *dev)
3251 {
3252 	struct inet6_dev  *idev;
3253 
3254 	/* ::1 */
3255 
3256 	ASSERT_RTNL();
3257 
3258 	idev = ipv6_find_idev(dev);
3259 	if (IS_ERR(idev)) {
3260 		pr_debug("%s: add_dev failed\n", __func__);
3261 		return;
3262 	}
3263 
3264 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3265 }
3266 
3267 void addrconf_add_linklocal(struct inet6_dev *idev,
3268 			    const struct in6_addr *addr, u32 flags)
3269 {
3270 	struct ifa6_config cfg = {
3271 		.pfx = addr,
3272 		.plen = 64,
3273 		.ifa_flags = flags | IFA_F_PERMANENT,
3274 		.valid_lft = INFINITY_LIFE_TIME,
3275 		.preferred_lft = INFINITY_LIFE_TIME,
3276 		.scope = IFA_LINK,
3277 		.ifa_proto = IFAPROT_KERNEL_LL
3278 	};
3279 	struct inet6_ifaddr *ifp;
3280 
3281 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3282 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3283 	     idev->cnf.optimistic_dad) &&
3284 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3285 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3286 #endif
3287 
3288 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3289 	if (!IS_ERR(ifp)) {
3290 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3291 				      0, 0, GFP_ATOMIC);
3292 		addrconf_dad_start(ifp);
3293 		in6_ifa_put(ifp);
3294 	}
3295 }
3296 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3297 
3298 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3299 {
3300 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3301 		return true;
3302 
3303 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3304 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3305 		return true;
3306 
3307 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3308 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3309 		return true;
3310 
3311 	return false;
3312 }
3313 
3314 static int ipv6_generate_stable_address(struct in6_addr *address,
3315 					u8 dad_count,
3316 					const struct inet6_dev *idev)
3317 {
3318 	static DEFINE_SPINLOCK(lock);
3319 	static __u32 digest[SHA1_DIGEST_WORDS];
3320 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3321 
3322 	static union {
3323 		char __data[SHA1_BLOCK_SIZE];
3324 		struct {
3325 			struct in6_addr secret;
3326 			__be32 prefix[2];
3327 			unsigned char hwaddr[MAX_ADDR_LEN];
3328 			u8 dad_count;
3329 		} __packed;
3330 	} data;
3331 
3332 	struct in6_addr secret;
3333 	struct in6_addr temp;
3334 	struct net *net = dev_net(idev->dev);
3335 
3336 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3337 
3338 	if (idev->cnf.stable_secret.initialized)
3339 		secret = idev->cnf.stable_secret.secret;
3340 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3341 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3342 	else
3343 		return -1;
3344 
3345 retry:
3346 	spin_lock_bh(&lock);
3347 
3348 	sha1_init(digest);
3349 	memset(&data, 0, sizeof(data));
3350 	memset(workspace, 0, sizeof(workspace));
3351 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3352 	data.prefix[0] = address->s6_addr32[0];
3353 	data.prefix[1] = address->s6_addr32[1];
3354 	data.secret = secret;
3355 	data.dad_count = dad_count;
3356 
3357 	sha1_transform(digest, data.__data, workspace);
3358 
3359 	temp = *address;
3360 	temp.s6_addr32[2] = (__force __be32)digest[0];
3361 	temp.s6_addr32[3] = (__force __be32)digest[1];
3362 
3363 	spin_unlock_bh(&lock);
3364 
3365 	if (ipv6_reserved_interfaceid(temp)) {
3366 		dad_count++;
3367 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3368 			return -1;
3369 		goto retry;
3370 	}
3371 
3372 	*address = temp;
3373 	return 0;
3374 }
3375 
3376 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3377 {
3378 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3379 
3380 	if (s->initialized)
3381 		return;
3382 	s = &idev->cnf.stable_secret;
3383 	get_random_bytes(&s->secret, sizeof(s->secret));
3384 	s->initialized = true;
3385 }
3386 
3387 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3388 {
3389 	struct in6_addr addr;
3390 
3391 	/* no link local addresses on L3 master devices */
3392 	if (netif_is_l3_master(idev->dev))
3393 		return;
3394 
3395 	/* no link local addresses on devices flagged as slaves */
3396 	if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3397 		return;
3398 
3399 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3400 
3401 	switch (idev->cnf.addr_gen_mode) {
3402 	case IN6_ADDR_GEN_MODE_RANDOM:
3403 		ipv6_gen_mode_random_init(idev);
3404 		fallthrough;
3405 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3406 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3407 			addrconf_add_linklocal(idev, &addr,
3408 					       IFA_F_STABLE_PRIVACY);
3409 		else if (prefix_route)
3410 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3411 					      0, 0, GFP_KERNEL);
3412 		break;
3413 	case IN6_ADDR_GEN_MODE_EUI64:
3414 		/* addrconf_add_linklocal also adds a prefix_route and we
3415 		 * only need to care about prefix routes if ipv6_generate_eui64
3416 		 * couldn't generate one.
3417 		 */
3418 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3419 			addrconf_add_linklocal(idev, &addr, 0);
3420 		else if (prefix_route)
3421 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3422 					      0, 0, GFP_KERNEL);
3423 		break;
3424 	case IN6_ADDR_GEN_MODE_NONE:
3425 	default:
3426 		/* will not add any link local address */
3427 		break;
3428 	}
3429 }
3430 
3431 static void addrconf_dev_config(struct net_device *dev)
3432 {
3433 	struct inet6_dev *idev;
3434 
3435 	ASSERT_RTNL();
3436 
3437 	if ((dev->type != ARPHRD_ETHER) &&
3438 	    (dev->type != ARPHRD_FDDI) &&
3439 	    (dev->type != ARPHRD_ARCNET) &&
3440 	    (dev->type != ARPHRD_INFINIBAND) &&
3441 	    (dev->type != ARPHRD_IEEE1394) &&
3442 	    (dev->type != ARPHRD_TUNNEL6) &&
3443 	    (dev->type != ARPHRD_6LOWPAN) &&
3444 	    (dev->type != ARPHRD_TUNNEL) &&
3445 	    (dev->type != ARPHRD_NONE) &&
3446 	    (dev->type != ARPHRD_RAWIP)) {
3447 		/* Alas, we support only Ethernet autoconfiguration. */
3448 		idev = __in6_dev_get(dev);
3449 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3450 		    dev->flags & IFF_MULTICAST)
3451 			ipv6_mc_up(idev);
3452 		return;
3453 	}
3454 
3455 	idev = addrconf_add_dev(dev);
3456 	if (IS_ERR(idev))
3457 		return;
3458 
3459 	/* this device type has no EUI support */
3460 	if (dev->type == ARPHRD_NONE &&
3461 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3462 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3463 
3464 	addrconf_addr_gen(idev, false);
3465 }
3466 
3467 #if IS_ENABLED(CONFIG_IPV6_SIT)
3468 static void addrconf_sit_config(struct net_device *dev)
3469 {
3470 	struct inet6_dev *idev;
3471 
3472 	ASSERT_RTNL();
3473 
3474 	/*
3475 	 * Configure the tunnel with one of our IPv4
3476 	 * addresses... we should configure all of
3477 	 * our v4 addrs in the tunnel
3478 	 */
3479 
3480 	idev = ipv6_find_idev(dev);
3481 	if (IS_ERR(idev)) {
3482 		pr_debug("%s: add_dev failed\n", __func__);
3483 		return;
3484 	}
3485 
3486 	if (dev->priv_flags & IFF_ISATAP) {
3487 		addrconf_addr_gen(idev, false);
3488 		return;
3489 	}
3490 
3491 	add_v4_addrs(idev);
3492 
3493 	if (dev->flags&IFF_POINTOPOINT)
3494 		addrconf_add_mroute(dev);
3495 }
3496 #endif
3497 
3498 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3499 static void addrconf_gre_config(struct net_device *dev)
3500 {
3501 	struct inet6_dev *idev;
3502 
3503 	ASSERT_RTNL();
3504 
3505 	idev = ipv6_find_idev(dev);
3506 	if (IS_ERR(idev)) {
3507 		pr_debug("%s: add_dev failed\n", __func__);
3508 		return;
3509 	}
3510 
3511 	if (dev->type == ARPHRD_ETHER) {
3512 		addrconf_addr_gen(idev, true);
3513 		return;
3514 	}
3515 
3516 	add_v4_addrs(idev);
3517 
3518 	if (dev->flags & IFF_POINTOPOINT)
3519 		addrconf_add_mroute(dev);
3520 }
3521 #endif
3522 
3523 static void addrconf_init_auto_addrs(struct net_device *dev)
3524 {
3525 	switch (dev->type) {
3526 #if IS_ENABLED(CONFIG_IPV6_SIT)
3527 	case ARPHRD_SIT:
3528 		addrconf_sit_config(dev);
3529 		break;
3530 #endif
3531 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3532 	case ARPHRD_IP6GRE:
3533 	case ARPHRD_IPGRE:
3534 		addrconf_gre_config(dev);
3535 		break;
3536 #endif
3537 	case ARPHRD_LOOPBACK:
3538 		init_loopback(dev);
3539 		break;
3540 
3541 	default:
3542 		addrconf_dev_config(dev);
3543 		break;
3544 	}
3545 }
3546 
3547 static int fixup_permanent_addr(struct net *net,
3548 				struct inet6_dev *idev,
3549 				struct inet6_ifaddr *ifp)
3550 {
3551 	/* !fib6_node means the host route was removed from the
3552 	 * FIB, for example, if 'lo' device is taken down. In that
3553 	 * case regenerate the host route.
3554 	 */
3555 	if (!ifp->rt || !ifp->rt->fib6_node) {
3556 		struct fib6_info *f6i, *prev;
3557 
3558 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3559 					 GFP_ATOMIC, NULL);
3560 		if (IS_ERR(f6i))
3561 			return PTR_ERR(f6i);
3562 
3563 		/* ifp->rt can be accessed outside of rtnl */
3564 		spin_lock(&ifp->lock);
3565 		prev = ifp->rt;
3566 		ifp->rt = f6i;
3567 		spin_unlock(&ifp->lock);
3568 
3569 		fib6_info_release(prev);
3570 	}
3571 
3572 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3573 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3574 				      ifp->rt_priority, idev->dev, 0, 0,
3575 				      GFP_ATOMIC);
3576 	}
3577 
3578 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3579 		addrconf_dad_start(ifp);
3580 
3581 	return 0;
3582 }
3583 
3584 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3585 {
3586 	struct inet6_ifaddr *ifp, *tmp;
3587 	struct inet6_dev *idev;
3588 
3589 	idev = __in6_dev_get(dev);
3590 	if (!idev)
3591 		return;
3592 
3593 	write_lock_bh(&idev->lock);
3594 
3595 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3596 		if ((ifp->flags & IFA_F_PERMANENT) &&
3597 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3598 			write_unlock_bh(&idev->lock);
3599 			in6_ifa_hold(ifp);
3600 			ipv6_del_addr(ifp);
3601 			write_lock_bh(&idev->lock);
3602 
3603 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3604 					     idev->dev->name, &ifp->addr);
3605 		}
3606 	}
3607 
3608 	write_unlock_bh(&idev->lock);
3609 }
3610 
3611 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3612 			   void *ptr)
3613 {
3614 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3615 	struct netdev_notifier_change_info *change_info;
3616 	struct netdev_notifier_changeupper_info *info;
3617 	struct inet6_dev *idev = __in6_dev_get(dev);
3618 	struct net *net = dev_net(dev);
3619 	int run_pending = 0;
3620 	int err;
3621 
3622 	switch (event) {
3623 	case NETDEV_REGISTER:
3624 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3625 			idev = ipv6_add_dev(dev);
3626 			if (IS_ERR(idev))
3627 				return notifier_from_errno(PTR_ERR(idev));
3628 		}
3629 		break;
3630 
3631 	case NETDEV_CHANGEMTU:
3632 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3633 		if (dev->mtu < IPV6_MIN_MTU) {
3634 			addrconf_ifdown(dev, dev != net->loopback_dev);
3635 			break;
3636 		}
3637 
3638 		if (idev) {
3639 			rt6_mtu_change(dev, dev->mtu);
3640 			idev->cnf.mtu6 = dev->mtu;
3641 			break;
3642 		}
3643 
3644 		/* allocate new idev */
3645 		idev = ipv6_add_dev(dev);
3646 		if (IS_ERR(idev))
3647 			break;
3648 
3649 		/* device is still not ready */
3650 		if (!(idev->if_flags & IF_READY))
3651 			break;
3652 
3653 		run_pending = 1;
3654 		fallthrough;
3655 	case NETDEV_UP:
3656 	case NETDEV_CHANGE:
3657 		if (idev && idev->cnf.disable_ipv6)
3658 			break;
3659 
3660 		if (dev->priv_flags & IFF_NO_ADDRCONF) {
3661 			if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3662 			    dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3663 				ipv6_mc_up(idev);
3664 			break;
3665 		}
3666 
3667 		if (event == NETDEV_UP) {
3668 			/* restore routes for permanent addresses */
3669 			addrconf_permanent_addr(net, dev);
3670 
3671 			if (!addrconf_link_ready(dev)) {
3672 				/* device is not ready yet. */
3673 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3674 					 dev->name);
3675 				break;
3676 			}
3677 
3678 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3679 				idev = ipv6_add_dev(dev);
3680 
3681 			if (!IS_ERR_OR_NULL(idev)) {
3682 				idev->if_flags |= IF_READY;
3683 				run_pending = 1;
3684 			}
3685 		} else if (event == NETDEV_CHANGE) {
3686 			if (!addrconf_link_ready(dev)) {
3687 				/* device is still not ready. */
3688 				rt6_sync_down_dev(dev, event);
3689 				break;
3690 			}
3691 
3692 			if (!IS_ERR_OR_NULL(idev)) {
3693 				if (idev->if_flags & IF_READY) {
3694 					/* device is already configured -
3695 					 * but resend MLD reports, we might
3696 					 * have roamed and need to update
3697 					 * multicast snooping switches
3698 					 */
3699 					ipv6_mc_up(idev);
3700 					change_info = ptr;
3701 					if (change_info->flags_changed & IFF_NOARP)
3702 						addrconf_dad_run(idev, true);
3703 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3704 					break;
3705 				}
3706 				idev->if_flags |= IF_READY;
3707 			}
3708 
3709 			pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3710 				 dev->name);
3711 
3712 			run_pending = 1;
3713 		}
3714 
3715 		addrconf_init_auto_addrs(dev);
3716 
3717 		if (!IS_ERR_OR_NULL(idev)) {
3718 			if (run_pending)
3719 				addrconf_dad_run(idev, false);
3720 
3721 			/* Device has an address by now */
3722 			rt6_sync_up(dev, RTNH_F_DEAD);
3723 
3724 			/*
3725 			 * If the MTU changed during the interface down,
3726 			 * when the interface up, the changed MTU must be
3727 			 * reflected in the idev as well as routers.
3728 			 */
3729 			if (idev->cnf.mtu6 != dev->mtu &&
3730 			    dev->mtu >= IPV6_MIN_MTU) {
3731 				rt6_mtu_change(dev, dev->mtu);
3732 				idev->cnf.mtu6 = dev->mtu;
3733 			}
3734 			idev->tstamp = jiffies;
3735 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3736 
3737 			/*
3738 			 * If the changed mtu during down is lower than
3739 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3740 			 */
3741 			if (dev->mtu < IPV6_MIN_MTU)
3742 				addrconf_ifdown(dev, dev != net->loopback_dev);
3743 		}
3744 		break;
3745 
3746 	case NETDEV_DOWN:
3747 	case NETDEV_UNREGISTER:
3748 		/*
3749 		 *	Remove all addresses from this interface.
3750 		 */
3751 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3752 		break;
3753 
3754 	case NETDEV_CHANGENAME:
3755 		if (idev) {
3756 			snmp6_unregister_dev(idev);
3757 			addrconf_sysctl_unregister(idev);
3758 			err = addrconf_sysctl_register(idev);
3759 			if (err)
3760 				return notifier_from_errno(err);
3761 			err = snmp6_register_dev(idev);
3762 			if (err) {
3763 				addrconf_sysctl_unregister(idev);
3764 				return notifier_from_errno(err);
3765 			}
3766 		}
3767 		break;
3768 
3769 	case NETDEV_PRE_TYPE_CHANGE:
3770 	case NETDEV_POST_TYPE_CHANGE:
3771 		if (idev)
3772 			addrconf_type_change(dev, event);
3773 		break;
3774 
3775 	case NETDEV_CHANGEUPPER:
3776 		info = ptr;
3777 
3778 		/* flush all routes if dev is linked to or unlinked from
3779 		 * an L3 master device (e.g., VRF)
3780 		 */
3781 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3782 			addrconf_ifdown(dev, false);
3783 	}
3784 
3785 	return NOTIFY_OK;
3786 }
3787 
3788 /*
3789  *	addrconf module should be notified of a device going up
3790  */
3791 static struct notifier_block ipv6_dev_notf = {
3792 	.notifier_call = addrconf_notify,
3793 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3794 };
3795 
3796 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3797 {
3798 	struct inet6_dev *idev;
3799 	ASSERT_RTNL();
3800 
3801 	idev = __in6_dev_get(dev);
3802 
3803 	if (event == NETDEV_POST_TYPE_CHANGE)
3804 		ipv6_mc_remap(idev);
3805 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3806 		ipv6_mc_unmap(idev);
3807 }
3808 
3809 static bool addr_is_local(const struct in6_addr *addr)
3810 {
3811 	return ipv6_addr_type(addr) &
3812 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3813 }
3814 
3815 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3816 {
3817 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3818 	struct net *net = dev_net(dev);
3819 	struct inet6_dev *idev;
3820 	struct inet6_ifaddr *ifa;
3821 	LIST_HEAD(tmp_addr_list);
3822 	bool keep_addr = false;
3823 	bool was_ready;
3824 	int state, i;
3825 
3826 	ASSERT_RTNL();
3827 
3828 	rt6_disable_ip(dev, event);
3829 
3830 	idev = __in6_dev_get(dev);
3831 	if (!idev)
3832 		return -ENODEV;
3833 
3834 	/*
3835 	 * Step 1: remove reference to ipv6 device from parent device.
3836 	 *	   Do not dev_put!
3837 	 */
3838 	if (unregister) {
3839 		idev->dead = 1;
3840 
3841 		/* protected by rtnl_lock */
3842 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3843 
3844 		/* Step 1.5: remove snmp6 entry */
3845 		snmp6_unregister_dev(idev);
3846 
3847 	}
3848 
3849 	/* combine the user config with event to determine if permanent
3850 	 * addresses are to be removed from address hash table
3851 	 */
3852 	if (!unregister && !idev->cnf.disable_ipv6) {
3853 		/* aggregate the system setting and interface setting */
3854 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3855 
3856 		if (!_keep_addr)
3857 			_keep_addr = idev->cnf.keep_addr_on_down;
3858 
3859 		keep_addr = (_keep_addr > 0);
3860 	}
3861 
3862 	/* Step 2: clear hash table */
3863 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3864 		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3865 
3866 		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3867 restart:
3868 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3869 			if (ifa->idev == idev) {
3870 				addrconf_del_dad_work(ifa);
3871 				/* combined flag + permanent flag decide if
3872 				 * address is retained on a down event
3873 				 */
3874 				if (!keep_addr ||
3875 				    !(ifa->flags & IFA_F_PERMANENT) ||
3876 				    addr_is_local(&ifa->addr)) {
3877 					hlist_del_init_rcu(&ifa->addr_lst);
3878 					goto restart;
3879 				}
3880 			}
3881 		}
3882 		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3883 	}
3884 
3885 	write_lock_bh(&idev->lock);
3886 
3887 	addrconf_del_rs_timer(idev);
3888 
3889 	/* Step 2: clear flags for stateless addrconf, repeated down
3890 	 *         detection
3891 	 */
3892 	was_ready = idev->if_flags & IF_READY;
3893 	if (!unregister)
3894 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3895 
3896 	/* Step 3: clear tempaddr list */
3897 	while (!list_empty(&idev->tempaddr_list)) {
3898 		ifa = list_first_entry(&idev->tempaddr_list,
3899 				       struct inet6_ifaddr, tmp_list);
3900 		list_del(&ifa->tmp_list);
3901 		write_unlock_bh(&idev->lock);
3902 		spin_lock_bh(&ifa->lock);
3903 
3904 		if (ifa->ifpub) {
3905 			in6_ifa_put(ifa->ifpub);
3906 			ifa->ifpub = NULL;
3907 		}
3908 		spin_unlock_bh(&ifa->lock);
3909 		in6_ifa_put(ifa);
3910 		write_lock_bh(&idev->lock);
3911 	}
3912 
3913 	list_for_each_entry(ifa, &idev->addr_list, if_list)
3914 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3915 	write_unlock_bh(&idev->lock);
3916 
3917 	while (!list_empty(&tmp_addr_list)) {
3918 		struct fib6_info *rt = NULL;
3919 		bool keep;
3920 
3921 		ifa = list_first_entry(&tmp_addr_list,
3922 				       struct inet6_ifaddr, if_list_aux);
3923 		list_del(&ifa->if_list_aux);
3924 
3925 		addrconf_del_dad_work(ifa);
3926 
3927 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3928 			!addr_is_local(&ifa->addr);
3929 
3930 		spin_lock_bh(&ifa->lock);
3931 
3932 		if (keep) {
3933 			/* set state to skip the notifier below */
3934 			state = INET6_IFADDR_STATE_DEAD;
3935 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3936 			if (!(ifa->flags & IFA_F_NODAD))
3937 				ifa->flags |= IFA_F_TENTATIVE;
3938 
3939 			rt = ifa->rt;
3940 			ifa->rt = NULL;
3941 		} else {
3942 			state = ifa->state;
3943 			ifa->state = INET6_IFADDR_STATE_DEAD;
3944 		}
3945 
3946 		spin_unlock_bh(&ifa->lock);
3947 
3948 		if (rt)
3949 			ip6_del_rt(net, rt, false);
3950 
3951 		if (state != INET6_IFADDR_STATE_DEAD) {
3952 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3953 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3954 		} else {
3955 			if (idev->cnf.forwarding)
3956 				addrconf_leave_anycast(ifa);
3957 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3958 		}
3959 
3960 		if (!keep) {
3961 			write_lock_bh(&idev->lock);
3962 			list_del_rcu(&ifa->if_list);
3963 			write_unlock_bh(&idev->lock);
3964 			in6_ifa_put(ifa);
3965 		}
3966 	}
3967 
3968 	/* Step 5: Discard anycast and multicast list */
3969 	if (unregister) {
3970 		ipv6_ac_destroy_dev(idev);
3971 		ipv6_mc_destroy_dev(idev);
3972 	} else if (was_ready) {
3973 		ipv6_mc_down(idev);
3974 	}
3975 
3976 	idev->tstamp = jiffies;
3977 	idev->ra_mtu = 0;
3978 
3979 	/* Last: Shot the device (if unregistered) */
3980 	if (unregister) {
3981 		addrconf_sysctl_unregister(idev);
3982 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3983 		neigh_ifdown(&nd_tbl, dev);
3984 		in6_dev_put(idev);
3985 	}
3986 	return 0;
3987 }
3988 
3989 static void addrconf_rs_timer(struct timer_list *t)
3990 {
3991 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3992 	struct net_device *dev = idev->dev;
3993 	struct in6_addr lladdr;
3994 
3995 	write_lock(&idev->lock);
3996 	if (idev->dead || !(idev->if_flags & IF_READY))
3997 		goto out;
3998 
3999 	if (!ipv6_accept_ra(idev))
4000 		goto out;
4001 
4002 	/* Announcement received after solicitation was sent */
4003 	if (idev->if_flags & IF_RA_RCVD)
4004 		goto out;
4005 
4006 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
4007 		write_unlock(&idev->lock);
4008 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4009 			ndisc_send_rs(dev, &lladdr,
4010 				      &in6addr_linklocal_allrouters);
4011 		else
4012 			goto put;
4013 
4014 		write_lock(&idev->lock);
4015 		idev->rs_interval = rfc3315_s14_backoff_update(
4016 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
4017 		/* The wait after the last probe can be shorter */
4018 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4019 					     idev->cnf.rtr_solicits) ?
4020 				      idev->cnf.rtr_solicit_delay :
4021 				      idev->rs_interval);
4022 	} else {
4023 		/*
4024 		 * Note: we do not support deprecated "all on-link"
4025 		 * assumption any longer.
4026 		 */
4027 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4028 	}
4029 
4030 out:
4031 	write_unlock(&idev->lock);
4032 put:
4033 	in6_dev_put(idev);
4034 }
4035 
4036 /*
4037  *	Duplicate Address Detection
4038  */
4039 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4040 {
4041 	unsigned long rand_num;
4042 	struct inet6_dev *idev = ifp->idev;
4043 	u64 nonce;
4044 
4045 	if (ifp->flags & IFA_F_OPTIMISTIC)
4046 		rand_num = 0;
4047 	else
4048 		rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
4049 
4050 	nonce = 0;
4051 	if (idev->cnf.enhanced_dad ||
4052 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
4053 		do
4054 			get_random_bytes(&nonce, 6);
4055 		while (nonce == 0);
4056 	}
4057 	ifp->dad_nonce = nonce;
4058 	ifp->dad_probes = idev->cnf.dad_transmits;
4059 	addrconf_mod_dad_work(ifp, rand_num);
4060 }
4061 
4062 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4063 {
4064 	struct inet6_dev *idev = ifp->idev;
4065 	struct net_device *dev = idev->dev;
4066 	bool bump_id, notify = false;
4067 	struct net *net;
4068 
4069 	addrconf_join_solict(dev, &ifp->addr);
4070 
4071 	read_lock_bh(&idev->lock);
4072 	spin_lock(&ifp->lock);
4073 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
4074 		goto out;
4075 
4076 	net = dev_net(dev);
4077 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4078 	    (net->ipv6.devconf_all->accept_dad < 1 &&
4079 	     idev->cnf.accept_dad < 1) ||
4080 	    !(ifp->flags&IFA_F_TENTATIVE) ||
4081 	    ifp->flags & IFA_F_NODAD) {
4082 		bool send_na = false;
4083 
4084 		if (ifp->flags & IFA_F_TENTATIVE &&
4085 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4086 			send_na = true;
4087 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4088 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4089 		spin_unlock(&ifp->lock);
4090 		read_unlock_bh(&idev->lock);
4091 
4092 		addrconf_dad_completed(ifp, bump_id, send_na);
4093 		return;
4094 	}
4095 
4096 	if (!(idev->if_flags & IF_READY)) {
4097 		spin_unlock(&ifp->lock);
4098 		read_unlock_bh(&idev->lock);
4099 		/*
4100 		 * If the device is not ready:
4101 		 * - keep it tentative if it is a permanent address.
4102 		 * - otherwise, kill it.
4103 		 */
4104 		in6_ifa_hold(ifp);
4105 		addrconf_dad_stop(ifp, 0);
4106 		return;
4107 	}
4108 
4109 	/*
4110 	 * Optimistic nodes can start receiving
4111 	 * Frames right away
4112 	 */
4113 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4114 		ip6_ins_rt(net, ifp->rt);
4115 		if (ipv6_use_optimistic_addr(net, idev)) {
4116 			/* Because optimistic nodes can use this address,
4117 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4118 			 */
4119 			notify = true;
4120 		}
4121 	}
4122 
4123 	addrconf_dad_kick(ifp);
4124 out:
4125 	spin_unlock(&ifp->lock);
4126 	read_unlock_bh(&idev->lock);
4127 	if (notify)
4128 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4129 }
4130 
4131 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4132 {
4133 	bool begin_dad = false;
4134 
4135 	spin_lock_bh(&ifp->lock);
4136 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4137 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4138 		begin_dad = true;
4139 	}
4140 	spin_unlock_bh(&ifp->lock);
4141 
4142 	if (begin_dad)
4143 		addrconf_mod_dad_work(ifp, 0);
4144 }
4145 
4146 static void addrconf_dad_work(struct work_struct *w)
4147 {
4148 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4149 						struct inet6_ifaddr,
4150 						dad_work);
4151 	struct inet6_dev *idev = ifp->idev;
4152 	bool bump_id, disable_ipv6 = false;
4153 	struct in6_addr mcaddr;
4154 
4155 	enum {
4156 		DAD_PROCESS,
4157 		DAD_BEGIN,
4158 		DAD_ABORT,
4159 	} action = DAD_PROCESS;
4160 
4161 	rtnl_lock();
4162 
4163 	spin_lock_bh(&ifp->lock);
4164 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4165 		action = DAD_BEGIN;
4166 		ifp->state = INET6_IFADDR_STATE_DAD;
4167 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4168 		action = DAD_ABORT;
4169 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4170 
4171 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4172 		     idev->cnf.accept_dad > 1) &&
4173 		    !idev->cnf.disable_ipv6 &&
4174 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4175 			struct in6_addr addr;
4176 
4177 			addr.s6_addr32[0] = htonl(0xfe800000);
4178 			addr.s6_addr32[1] = 0;
4179 
4180 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4181 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4182 				/* DAD failed for link-local based on MAC */
4183 				WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4184 
4185 				pr_info("%s: IPv6 being disabled!\n",
4186 					ifp->idev->dev->name);
4187 				disable_ipv6 = true;
4188 			}
4189 		}
4190 	}
4191 	spin_unlock_bh(&ifp->lock);
4192 
4193 	if (action == DAD_BEGIN) {
4194 		addrconf_dad_begin(ifp);
4195 		goto out;
4196 	} else if (action == DAD_ABORT) {
4197 		in6_ifa_hold(ifp);
4198 		addrconf_dad_stop(ifp, 1);
4199 		if (disable_ipv6)
4200 			addrconf_ifdown(idev->dev, false);
4201 		goto out;
4202 	}
4203 
4204 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4205 		goto out;
4206 
4207 	write_lock_bh(&idev->lock);
4208 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4209 		write_unlock_bh(&idev->lock);
4210 		goto out;
4211 	}
4212 
4213 	spin_lock(&ifp->lock);
4214 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4215 		spin_unlock(&ifp->lock);
4216 		write_unlock_bh(&idev->lock);
4217 		goto out;
4218 	}
4219 
4220 	if (ifp->dad_probes == 0) {
4221 		bool send_na = false;
4222 
4223 		/*
4224 		 * DAD was successful
4225 		 */
4226 
4227 		if (ifp->flags & IFA_F_TENTATIVE &&
4228 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4229 			send_na = true;
4230 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4231 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4232 		spin_unlock(&ifp->lock);
4233 		write_unlock_bh(&idev->lock);
4234 
4235 		addrconf_dad_completed(ifp, bump_id, send_na);
4236 
4237 		goto out;
4238 	}
4239 
4240 	ifp->dad_probes--;
4241 	addrconf_mod_dad_work(ifp,
4242 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4243 				  HZ/100));
4244 	spin_unlock(&ifp->lock);
4245 	write_unlock_bh(&idev->lock);
4246 
4247 	/* send a neighbour solicitation for our addr */
4248 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4249 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4250 		      ifp->dad_nonce);
4251 out:
4252 	in6_ifa_put(ifp);
4253 	rtnl_unlock();
4254 }
4255 
4256 /* ifp->idev must be at least read locked */
4257 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4258 {
4259 	struct inet6_ifaddr *ifpiter;
4260 	struct inet6_dev *idev = ifp->idev;
4261 
4262 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4263 		if (ifpiter->scope > IFA_LINK)
4264 			break;
4265 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4266 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4267 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4268 		    IFA_F_PERMANENT)
4269 			return false;
4270 	}
4271 	return true;
4272 }
4273 
4274 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4275 				   bool send_na)
4276 {
4277 	struct net_device *dev = ifp->idev->dev;
4278 	struct in6_addr lladdr;
4279 	bool send_rs, send_mld;
4280 
4281 	addrconf_del_dad_work(ifp);
4282 
4283 	/*
4284 	 *	Configure the address for reception. Now it is valid.
4285 	 */
4286 
4287 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4288 
4289 	/* If added prefix is link local and we are prepared to process
4290 	   router advertisements, start sending router solicitations.
4291 	 */
4292 
4293 	read_lock_bh(&ifp->idev->lock);
4294 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4295 	send_rs = send_mld &&
4296 		  ipv6_accept_ra(ifp->idev) &&
4297 		  ifp->idev->cnf.rtr_solicits != 0 &&
4298 		  (dev->flags & IFF_LOOPBACK) == 0 &&
4299 		  (dev->type != ARPHRD_TUNNEL) &&
4300 		  !netif_is_team_port(dev);
4301 	read_unlock_bh(&ifp->idev->lock);
4302 
4303 	/* While dad is in progress mld report's source address is in6_addrany.
4304 	 * Resend with proper ll now.
4305 	 */
4306 	if (send_mld)
4307 		ipv6_mc_dad_complete(ifp->idev);
4308 
4309 	/* send unsolicited NA if enabled */
4310 	if (send_na &&
4311 	    (ifp->idev->cnf.ndisc_notify ||
4312 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4313 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4314 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4315 			      /*solicited=*/ false, /*override=*/ true,
4316 			      /*inc_opt=*/ true);
4317 	}
4318 
4319 	if (send_rs) {
4320 		/*
4321 		 *	If a host as already performed a random delay
4322 		 *	[...] as part of DAD [...] there is no need
4323 		 *	to delay again before sending the first RS
4324 		 */
4325 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4326 			return;
4327 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4328 
4329 		write_lock_bh(&ifp->idev->lock);
4330 		spin_lock(&ifp->lock);
4331 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4332 			ifp->idev->cnf.rtr_solicit_interval);
4333 		ifp->idev->rs_probes = 1;
4334 		ifp->idev->if_flags |= IF_RS_SENT;
4335 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4336 		spin_unlock(&ifp->lock);
4337 		write_unlock_bh(&ifp->idev->lock);
4338 	}
4339 
4340 	if (bump_id)
4341 		rt_genid_bump_ipv6(dev_net(dev));
4342 
4343 	/* Make sure that a new temporary address will be created
4344 	 * before this temporary address becomes deprecated.
4345 	 */
4346 	if (ifp->flags & IFA_F_TEMPORARY)
4347 		addrconf_verify_rtnl(dev_net(dev));
4348 }
4349 
4350 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4351 {
4352 	struct inet6_ifaddr *ifp;
4353 
4354 	read_lock_bh(&idev->lock);
4355 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4356 		spin_lock(&ifp->lock);
4357 		if ((ifp->flags & IFA_F_TENTATIVE &&
4358 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4359 			if (restart)
4360 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4361 			addrconf_dad_kick(ifp);
4362 		}
4363 		spin_unlock(&ifp->lock);
4364 	}
4365 	read_unlock_bh(&idev->lock);
4366 }
4367 
4368 #ifdef CONFIG_PROC_FS
4369 struct if6_iter_state {
4370 	struct seq_net_private p;
4371 	int bucket;
4372 	int offset;
4373 };
4374 
4375 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4376 {
4377 	struct if6_iter_state *state = seq->private;
4378 	struct net *net = seq_file_net(seq);
4379 	struct inet6_ifaddr *ifa = NULL;
4380 	int p = 0;
4381 
4382 	/* initial bucket if pos is 0 */
4383 	if (pos == 0) {
4384 		state->bucket = 0;
4385 		state->offset = 0;
4386 	}
4387 
4388 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4389 		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4390 					 addr_lst) {
4391 			/* sync with offset */
4392 			if (p < state->offset) {
4393 				p++;
4394 				continue;
4395 			}
4396 			return ifa;
4397 		}
4398 
4399 		/* prepare for next bucket */
4400 		state->offset = 0;
4401 		p = 0;
4402 	}
4403 	return NULL;
4404 }
4405 
4406 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4407 					 struct inet6_ifaddr *ifa)
4408 {
4409 	struct if6_iter_state *state = seq->private;
4410 	struct net *net = seq_file_net(seq);
4411 
4412 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4413 		state->offset++;
4414 		return ifa;
4415 	}
4416 
4417 	state->offset = 0;
4418 	while (++state->bucket < IN6_ADDR_HSIZE) {
4419 		hlist_for_each_entry_rcu(ifa,
4420 				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4421 			return ifa;
4422 		}
4423 	}
4424 
4425 	return NULL;
4426 }
4427 
4428 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4429 	__acquires(rcu)
4430 {
4431 	rcu_read_lock();
4432 	return if6_get_first(seq, *pos);
4433 }
4434 
4435 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4436 {
4437 	struct inet6_ifaddr *ifa;
4438 
4439 	ifa = if6_get_next(seq, v);
4440 	++*pos;
4441 	return ifa;
4442 }
4443 
4444 static void if6_seq_stop(struct seq_file *seq, void *v)
4445 	__releases(rcu)
4446 {
4447 	rcu_read_unlock();
4448 }
4449 
4450 static int if6_seq_show(struct seq_file *seq, void *v)
4451 {
4452 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4453 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4454 		   &ifp->addr,
4455 		   ifp->idev->dev->ifindex,
4456 		   ifp->prefix_len,
4457 		   ifp->scope,
4458 		   (u8) ifp->flags,
4459 		   ifp->idev->dev->name);
4460 	return 0;
4461 }
4462 
4463 static const struct seq_operations if6_seq_ops = {
4464 	.start	= if6_seq_start,
4465 	.next	= if6_seq_next,
4466 	.show	= if6_seq_show,
4467 	.stop	= if6_seq_stop,
4468 };
4469 
4470 static int __net_init if6_proc_net_init(struct net *net)
4471 {
4472 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4473 			sizeof(struct if6_iter_state)))
4474 		return -ENOMEM;
4475 	return 0;
4476 }
4477 
4478 static void __net_exit if6_proc_net_exit(struct net *net)
4479 {
4480 	remove_proc_entry("if_inet6", net->proc_net);
4481 }
4482 
4483 static struct pernet_operations if6_proc_net_ops = {
4484 	.init = if6_proc_net_init,
4485 	.exit = if6_proc_net_exit,
4486 };
4487 
4488 int __init if6_proc_init(void)
4489 {
4490 	return register_pernet_subsys(&if6_proc_net_ops);
4491 }
4492 
4493 void if6_proc_exit(void)
4494 {
4495 	unregister_pernet_subsys(&if6_proc_net_ops);
4496 }
4497 #endif	/* CONFIG_PROC_FS */
4498 
4499 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4500 /* Check if address is a home address configured on any interface. */
4501 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4502 {
4503 	unsigned int hash = inet6_addr_hash(net, addr);
4504 	struct inet6_ifaddr *ifp = NULL;
4505 	int ret = 0;
4506 
4507 	rcu_read_lock();
4508 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4509 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4510 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4511 			ret = 1;
4512 			break;
4513 		}
4514 	}
4515 	rcu_read_unlock();
4516 	return ret;
4517 }
4518 #endif
4519 
4520 /* RFC6554 has some algorithm to avoid loops in segment routing by
4521  * checking if the segments contains any of a local interface address.
4522  *
4523  * Quote:
4524  *
4525  * To detect loops in the SRH, a router MUST determine if the SRH
4526  * includes multiple addresses assigned to any interface on that router.
4527  * If such addresses appear more than once and are separated by at least
4528  * one address not assigned to that router.
4529  */
4530 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4531 			  unsigned char nsegs)
4532 {
4533 	const struct in6_addr *addr;
4534 	int i, ret = 0, found = 0;
4535 	struct inet6_ifaddr *ifp;
4536 	bool separated = false;
4537 	unsigned int hash;
4538 	bool hash_found;
4539 
4540 	rcu_read_lock();
4541 	for (i = 0; i < nsegs; i++) {
4542 		addr = &segs[i];
4543 		hash = inet6_addr_hash(net, addr);
4544 
4545 		hash_found = false;
4546 		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4547 
4548 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4549 				hash_found = true;
4550 				break;
4551 			}
4552 		}
4553 
4554 		if (hash_found) {
4555 			if (found > 1 && separated) {
4556 				ret = 1;
4557 				break;
4558 			}
4559 
4560 			separated = false;
4561 			found++;
4562 		} else {
4563 			separated = true;
4564 		}
4565 	}
4566 	rcu_read_unlock();
4567 
4568 	return ret;
4569 }
4570 
4571 /*
4572  *	Periodic address status verification
4573  */
4574 
4575 static void addrconf_verify_rtnl(struct net *net)
4576 {
4577 	unsigned long now, next, next_sec, next_sched;
4578 	struct inet6_ifaddr *ifp;
4579 	int i;
4580 
4581 	ASSERT_RTNL();
4582 
4583 	rcu_read_lock_bh();
4584 	now = jiffies;
4585 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4586 
4587 	cancel_delayed_work(&net->ipv6.addr_chk_work);
4588 
4589 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4590 restart:
4591 		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4592 			unsigned long age;
4593 
4594 			/* When setting preferred_lft to a value not zero or
4595 			 * infinity, while valid_lft is infinity
4596 			 * IFA_F_PERMANENT has a non-infinity life time.
4597 			 */
4598 			if ((ifp->flags & IFA_F_PERMANENT) &&
4599 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4600 				continue;
4601 
4602 			spin_lock(&ifp->lock);
4603 			/* We try to batch several events at once. */
4604 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4605 
4606 			if ((ifp->flags&IFA_F_TEMPORARY) &&
4607 			    !(ifp->flags&IFA_F_TENTATIVE) &&
4608 			    ifp->prefered_lft != INFINITY_LIFE_TIME &&
4609 			    !ifp->regen_count && ifp->ifpub) {
4610 				/* This is a non-regenerated temporary addr. */
4611 
4612 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4613 					ifp->idev->cnf.dad_transmits *
4614 					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4615 
4616 				if (age + regen_advance >= ifp->prefered_lft) {
4617 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4618 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4619 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4620 
4621 					ifp->regen_count++;
4622 					in6_ifa_hold(ifp);
4623 					in6_ifa_hold(ifpub);
4624 					spin_unlock(&ifp->lock);
4625 
4626 					spin_lock(&ifpub->lock);
4627 					ifpub->regen_count = 0;
4628 					spin_unlock(&ifpub->lock);
4629 					rcu_read_unlock_bh();
4630 					ipv6_create_tempaddr(ifpub, true);
4631 					in6_ifa_put(ifpub);
4632 					in6_ifa_put(ifp);
4633 					rcu_read_lock_bh();
4634 					goto restart;
4635 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4636 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4637 			}
4638 
4639 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4640 			    age >= ifp->valid_lft) {
4641 				spin_unlock(&ifp->lock);
4642 				in6_ifa_hold(ifp);
4643 				rcu_read_unlock_bh();
4644 				ipv6_del_addr(ifp);
4645 				rcu_read_lock_bh();
4646 				goto restart;
4647 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4648 				spin_unlock(&ifp->lock);
4649 				continue;
4650 			} else if (age >= ifp->prefered_lft) {
4651 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4652 				int deprecate = 0;
4653 
4654 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4655 					deprecate = 1;
4656 					ifp->flags |= IFA_F_DEPRECATED;
4657 				}
4658 
4659 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4660 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4661 					next = ifp->tstamp + ifp->valid_lft * HZ;
4662 
4663 				spin_unlock(&ifp->lock);
4664 
4665 				if (deprecate) {
4666 					in6_ifa_hold(ifp);
4667 
4668 					ipv6_ifa_notify(0, ifp);
4669 					in6_ifa_put(ifp);
4670 					goto restart;
4671 				}
4672 			} else {
4673 				/* ifp->prefered_lft <= ifp->valid_lft */
4674 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4675 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4676 				spin_unlock(&ifp->lock);
4677 			}
4678 		}
4679 	}
4680 
4681 	next_sec = round_jiffies_up(next);
4682 	next_sched = next;
4683 
4684 	/* If rounded timeout is accurate enough, accept it. */
4685 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4686 		next_sched = next_sec;
4687 
4688 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4689 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4690 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4691 
4692 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4693 		 now, next, next_sec, next_sched);
4694 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4695 	rcu_read_unlock_bh();
4696 }
4697 
4698 static void addrconf_verify_work(struct work_struct *w)
4699 {
4700 	struct net *net = container_of(to_delayed_work(w), struct net,
4701 				       ipv6.addr_chk_work);
4702 
4703 	rtnl_lock();
4704 	addrconf_verify_rtnl(net);
4705 	rtnl_unlock();
4706 }
4707 
4708 static void addrconf_verify(struct net *net)
4709 {
4710 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4711 }
4712 
4713 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4714 				     struct in6_addr **peer_pfx)
4715 {
4716 	struct in6_addr *pfx = NULL;
4717 
4718 	*peer_pfx = NULL;
4719 
4720 	if (addr)
4721 		pfx = nla_data(addr);
4722 
4723 	if (local) {
4724 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4725 			*peer_pfx = pfx;
4726 		pfx = nla_data(local);
4727 	}
4728 
4729 	return pfx;
4730 }
4731 
4732 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4733 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4734 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4735 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4736 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4737 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4738 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4739 	[IFA_PROTO]		= { .type = NLA_U8 },
4740 };
4741 
4742 static int
4743 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4744 		  struct netlink_ext_ack *extack)
4745 {
4746 	struct net *net = sock_net(skb->sk);
4747 	struct ifaddrmsg *ifm;
4748 	struct nlattr *tb[IFA_MAX+1];
4749 	struct in6_addr *pfx, *peer_pfx;
4750 	u32 ifa_flags;
4751 	int err;
4752 
4753 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4754 				     ifa_ipv6_policy, extack);
4755 	if (err < 0)
4756 		return err;
4757 
4758 	ifm = nlmsg_data(nlh);
4759 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4760 	if (!pfx)
4761 		return -EINVAL;
4762 
4763 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4764 
4765 	/* We ignore other flags so far. */
4766 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4767 
4768 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4769 			      ifm->ifa_prefixlen, extack);
4770 }
4771 
4772 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4773 			       unsigned long expires, u32 flags,
4774 			       bool modify_peer)
4775 {
4776 	struct fib6_info *f6i;
4777 	u32 prio;
4778 
4779 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4780 					ifp->prefix_len,
4781 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4782 	if (!f6i)
4783 		return -ENOENT;
4784 
4785 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4786 	if (f6i->fib6_metric != prio) {
4787 		/* delete old one */
4788 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4789 
4790 		/* add new one */
4791 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4792 				      ifp->prefix_len,
4793 				      ifp->rt_priority, ifp->idev->dev,
4794 				      expires, flags, GFP_KERNEL);
4795 	} else {
4796 		if (!expires)
4797 			fib6_clean_expires(f6i);
4798 		else
4799 			fib6_set_expires(f6i, expires);
4800 
4801 		fib6_info_release(f6i);
4802 	}
4803 
4804 	return 0;
4805 }
4806 
4807 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4808 			     struct ifa6_config *cfg)
4809 {
4810 	u32 flags;
4811 	clock_t expires;
4812 	unsigned long timeout;
4813 	bool was_managetempaddr;
4814 	bool had_prefixroute;
4815 	bool new_peer = false;
4816 
4817 	ASSERT_RTNL();
4818 
4819 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4820 		return -EINVAL;
4821 
4822 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4823 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4824 		return -EINVAL;
4825 
4826 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4827 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4828 
4829 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4830 	if (addrconf_finite_timeout(timeout)) {
4831 		expires = jiffies_to_clock_t(timeout * HZ);
4832 		cfg->valid_lft = timeout;
4833 		flags = RTF_EXPIRES;
4834 	} else {
4835 		expires = 0;
4836 		flags = 0;
4837 		cfg->ifa_flags |= IFA_F_PERMANENT;
4838 	}
4839 
4840 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4841 	if (addrconf_finite_timeout(timeout)) {
4842 		if (timeout == 0)
4843 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4844 		cfg->preferred_lft = timeout;
4845 	}
4846 
4847 	if (cfg->peer_pfx &&
4848 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4849 		if (!ipv6_addr_any(&ifp->peer_addr))
4850 			cleanup_prefix_route(ifp, expires, true, true);
4851 		new_peer = true;
4852 	}
4853 
4854 	spin_lock_bh(&ifp->lock);
4855 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4856 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4857 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4858 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4859 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4860 			IFA_F_NOPREFIXROUTE);
4861 	ifp->flags |= cfg->ifa_flags;
4862 	ifp->tstamp = jiffies;
4863 	ifp->valid_lft = cfg->valid_lft;
4864 	ifp->prefered_lft = cfg->preferred_lft;
4865 	ifp->ifa_proto = cfg->ifa_proto;
4866 
4867 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4868 		ifp->rt_priority = cfg->rt_priority;
4869 
4870 	if (new_peer)
4871 		ifp->peer_addr = *cfg->peer_pfx;
4872 
4873 	spin_unlock_bh(&ifp->lock);
4874 	if (!(ifp->flags&IFA_F_TENTATIVE))
4875 		ipv6_ifa_notify(0, ifp);
4876 
4877 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4878 		int rc = -ENOENT;
4879 
4880 		if (had_prefixroute)
4881 			rc = modify_prefix_route(ifp, expires, flags, false);
4882 
4883 		/* prefix route could have been deleted; if so restore it */
4884 		if (rc == -ENOENT) {
4885 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4886 					      ifp->rt_priority, ifp->idev->dev,
4887 					      expires, flags, GFP_KERNEL);
4888 		}
4889 
4890 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4891 			rc = modify_prefix_route(ifp, expires, flags, true);
4892 
4893 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4894 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4895 					      ifp->rt_priority, ifp->idev->dev,
4896 					      expires, flags, GFP_KERNEL);
4897 		}
4898 	} else if (had_prefixroute) {
4899 		enum cleanup_prefix_rt_t action;
4900 		unsigned long rt_expires;
4901 
4902 		write_lock_bh(&ifp->idev->lock);
4903 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4904 		write_unlock_bh(&ifp->idev->lock);
4905 
4906 		if (action != CLEANUP_PREFIX_RT_NOP) {
4907 			cleanup_prefix_route(ifp, rt_expires,
4908 				action == CLEANUP_PREFIX_RT_DEL, false);
4909 		}
4910 	}
4911 
4912 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4913 		if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4914 			delete_tempaddrs(ifp->idev, ifp);
4915 		else
4916 			manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4917 					 cfg->preferred_lft, !was_managetempaddr,
4918 					 jiffies);
4919 	}
4920 
4921 	addrconf_verify_rtnl(net);
4922 
4923 	return 0;
4924 }
4925 
4926 static int
4927 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4928 		  struct netlink_ext_ack *extack)
4929 {
4930 	struct net *net = sock_net(skb->sk);
4931 	struct ifaddrmsg *ifm;
4932 	struct nlattr *tb[IFA_MAX+1];
4933 	struct in6_addr *peer_pfx;
4934 	struct inet6_ifaddr *ifa;
4935 	struct net_device *dev;
4936 	struct inet6_dev *idev;
4937 	struct ifa6_config cfg;
4938 	int err;
4939 
4940 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4941 				     ifa_ipv6_policy, extack);
4942 	if (err < 0)
4943 		return err;
4944 
4945 	memset(&cfg, 0, sizeof(cfg));
4946 
4947 	ifm = nlmsg_data(nlh);
4948 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4949 	if (!cfg.pfx)
4950 		return -EINVAL;
4951 
4952 	cfg.peer_pfx = peer_pfx;
4953 	cfg.plen = ifm->ifa_prefixlen;
4954 	if (tb[IFA_RT_PRIORITY])
4955 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4956 
4957 	if (tb[IFA_PROTO])
4958 		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4959 
4960 	cfg.valid_lft = INFINITY_LIFE_TIME;
4961 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4962 
4963 	if (tb[IFA_CACHEINFO]) {
4964 		struct ifa_cacheinfo *ci;
4965 
4966 		ci = nla_data(tb[IFA_CACHEINFO]);
4967 		cfg.valid_lft = ci->ifa_valid;
4968 		cfg.preferred_lft = ci->ifa_prefered;
4969 	}
4970 
4971 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4972 	if (!dev) {
4973 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
4974 		return -ENODEV;
4975 	}
4976 
4977 	if (tb[IFA_FLAGS])
4978 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4979 	else
4980 		cfg.ifa_flags = ifm->ifa_flags;
4981 
4982 	/* We ignore other flags so far. */
4983 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4984 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4985 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4986 
4987 	idev = ipv6_find_idev(dev);
4988 	if (IS_ERR(idev))
4989 		return PTR_ERR(idev);
4990 
4991 	if (!ipv6_allow_optimistic_dad(net, idev))
4992 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4993 
4994 	if (cfg.ifa_flags & IFA_F_NODAD &&
4995 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4996 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4997 		return -EINVAL;
4998 	}
4999 
5000 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5001 	if (!ifa) {
5002 		/*
5003 		 * It would be best to check for !NLM_F_CREATE here but
5004 		 * userspace already relies on not having to provide this.
5005 		 */
5006 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
5007 	}
5008 
5009 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
5010 	    !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5011 		NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5012 		err = -EEXIST;
5013 	} else {
5014 		err = inet6_addr_modify(net, ifa, &cfg);
5015 	}
5016 
5017 	in6_ifa_put(ifa);
5018 
5019 	return err;
5020 }
5021 
5022 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5023 			  u8 scope, int ifindex)
5024 {
5025 	struct ifaddrmsg *ifm;
5026 
5027 	ifm = nlmsg_data(nlh);
5028 	ifm->ifa_family = AF_INET6;
5029 	ifm->ifa_prefixlen = prefixlen;
5030 	ifm->ifa_flags = flags;
5031 	ifm->ifa_scope = scope;
5032 	ifm->ifa_index = ifindex;
5033 }
5034 
5035 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5036 			 unsigned long tstamp, u32 preferred, u32 valid)
5037 {
5038 	struct ifa_cacheinfo ci;
5039 
5040 	ci.cstamp = cstamp_delta(cstamp);
5041 	ci.tstamp = cstamp_delta(tstamp);
5042 	ci.ifa_prefered = preferred;
5043 	ci.ifa_valid = valid;
5044 
5045 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5046 }
5047 
5048 static inline int rt_scope(int ifa_scope)
5049 {
5050 	if (ifa_scope & IFA_HOST)
5051 		return RT_SCOPE_HOST;
5052 	else if (ifa_scope & IFA_LINK)
5053 		return RT_SCOPE_LINK;
5054 	else if (ifa_scope & IFA_SITE)
5055 		return RT_SCOPE_SITE;
5056 	else
5057 		return RT_SCOPE_UNIVERSE;
5058 }
5059 
5060 static inline int inet6_ifaddr_msgsize(void)
5061 {
5062 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5063 	       + nla_total_size(16) /* IFA_LOCAL */
5064 	       + nla_total_size(16) /* IFA_ADDRESS */
5065 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
5066 	       + nla_total_size(4)  /* IFA_FLAGS */
5067 	       + nla_total_size(1)  /* IFA_PROTO */
5068 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5069 }
5070 
5071 enum addr_type_t {
5072 	UNICAST_ADDR,
5073 	MULTICAST_ADDR,
5074 	ANYCAST_ADDR,
5075 };
5076 
5077 struct inet6_fill_args {
5078 	u32 portid;
5079 	u32 seq;
5080 	int event;
5081 	unsigned int flags;
5082 	int netnsid;
5083 	int ifindex;
5084 	enum addr_type_t type;
5085 };
5086 
5087 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5088 			     struct inet6_fill_args *args)
5089 {
5090 	struct nlmsghdr  *nlh;
5091 	u32 preferred, valid;
5092 
5093 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5094 			sizeof(struct ifaddrmsg), args->flags);
5095 	if (!nlh)
5096 		return -EMSGSIZE;
5097 
5098 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5099 		      ifa->idev->dev->ifindex);
5100 
5101 	if (args->netnsid >= 0 &&
5102 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5103 		goto error;
5104 
5105 	spin_lock_bh(&ifa->lock);
5106 	if (!((ifa->flags&IFA_F_PERMANENT) &&
5107 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5108 		preferred = ifa->prefered_lft;
5109 		valid = ifa->valid_lft;
5110 		if (preferred != INFINITY_LIFE_TIME) {
5111 			long tval = (jiffies - ifa->tstamp)/HZ;
5112 			if (preferred > tval)
5113 				preferred -= tval;
5114 			else
5115 				preferred = 0;
5116 			if (valid != INFINITY_LIFE_TIME) {
5117 				if (valid > tval)
5118 					valid -= tval;
5119 				else
5120 					valid = 0;
5121 			}
5122 		}
5123 	} else {
5124 		preferred = INFINITY_LIFE_TIME;
5125 		valid = INFINITY_LIFE_TIME;
5126 	}
5127 	spin_unlock_bh(&ifa->lock);
5128 
5129 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5130 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5131 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5132 			goto error;
5133 	} else
5134 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5135 			goto error;
5136 
5137 	if (ifa->rt_priority &&
5138 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5139 		goto error;
5140 
5141 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5142 		goto error;
5143 
5144 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5145 		goto error;
5146 
5147 	if (ifa->ifa_proto &&
5148 	    nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5149 		goto error;
5150 
5151 	nlmsg_end(skb, nlh);
5152 	return 0;
5153 
5154 error:
5155 	nlmsg_cancel(skb, nlh);
5156 	return -EMSGSIZE;
5157 }
5158 
5159 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5160 			       struct inet6_fill_args *args)
5161 {
5162 	struct nlmsghdr  *nlh;
5163 	u8 scope = RT_SCOPE_UNIVERSE;
5164 	int ifindex = ifmca->idev->dev->ifindex;
5165 
5166 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5167 		scope = RT_SCOPE_SITE;
5168 
5169 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5170 			sizeof(struct ifaddrmsg), args->flags);
5171 	if (!nlh)
5172 		return -EMSGSIZE;
5173 
5174 	if (args->netnsid >= 0 &&
5175 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5176 		nlmsg_cancel(skb, nlh);
5177 		return -EMSGSIZE;
5178 	}
5179 
5180 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5181 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5182 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5183 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5184 		nlmsg_cancel(skb, nlh);
5185 		return -EMSGSIZE;
5186 	}
5187 
5188 	nlmsg_end(skb, nlh);
5189 	return 0;
5190 }
5191 
5192 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5193 			       struct inet6_fill_args *args)
5194 {
5195 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5196 	int ifindex = dev ? dev->ifindex : 1;
5197 	struct nlmsghdr  *nlh;
5198 	u8 scope = RT_SCOPE_UNIVERSE;
5199 
5200 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5201 		scope = RT_SCOPE_SITE;
5202 
5203 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5204 			sizeof(struct ifaddrmsg), args->flags);
5205 	if (!nlh)
5206 		return -EMSGSIZE;
5207 
5208 	if (args->netnsid >= 0 &&
5209 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5210 		nlmsg_cancel(skb, nlh);
5211 		return -EMSGSIZE;
5212 	}
5213 
5214 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5215 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5216 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5217 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5218 		nlmsg_cancel(skb, nlh);
5219 		return -EMSGSIZE;
5220 	}
5221 
5222 	nlmsg_end(skb, nlh);
5223 	return 0;
5224 }
5225 
5226 /* called with rcu_read_lock() */
5227 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5228 			  struct netlink_callback *cb, int s_ip_idx,
5229 			  struct inet6_fill_args *fillargs)
5230 {
5231 	struct ifmcaddr6 *ifmca;
5232 	struct ifacaddr6 *ifaca;
5233 	int ip_idx = 0;
5234 	int err = 1;
5235 
5236 	read_lock_bh(&idev->lock);
5237 	switch (fillargs->type) {
5238 	case UNICAST_ADDR: {
5239 		struct inet6_ifaddr *ifa;
5240 		fillargs->event = RTM_NEWADDR;
5241 
5242 		/* unicast address incl. temp addr */
5243 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
5244 			if (ip_idx < s_ip_idx)
5245 				goto next;
5246 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5247 			if (err < 0)
5248 				break;
5249 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5250 next:
5251 			ip_idx++;
5252 		}
5253 		break;
5254 	}
5255 	case MULTICAST_ADDR:
5256 		read_unlock_bh(&idev->lock);
5257 		fillargs->event = RTM_GETMULTICAST;
5258 
5259 		/* multicast address */
5260 		for (ifmca = rtnl_dereference(idev->mc_list);
5261 		     ifmca;
5262 		     ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5263 			if (ip_idx < s_ip_idx)
5264 				continue;
5265 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5266 			if (err < 0)
5267 				break;
5268 		}
5269 		read_lock_bh(&idev->lock);
5270 		break;
5271 	case ANYCAST_ADDR:
5272 		fillargs->event = RTM_GETANYCAST;
5273 		/* anycast address */
5274 		for (ifaca = idev->ac_list; ifaca;
5275 		     ifaca = ifaca->aca_next, ip_idx++) {
5276 			if (ip_idx < s_ip_idx)
5277 				continue;
5278 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5279 			if (err < 0)
5280 				break;
5281 		}
5282 		break;
5283 	default:
5284 		break;
5285 	}
5286 	read_unlock_bh(&idev->lock);
5287 	cb->args[2] = ip_idx;
5288 	return err;
5289 }
5290 
5291 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5292 				       struct inet6_fill_args *fillargs,
5293 				       struct net **tgt_net, struct sock *sk,
5294 				       struct netlink_callback *cb)
5295 {
5296 	struct netlink_ext_ack *extack = cb->extack;
5297 	struct nlattr *tb[IFA_MAX+1];
5298 	struct ifaddrmsg *ifm;
5299 	int err, i;
5300 
5301 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5302 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5303 		return -EINVAL;
5304 	}
5305 
5306 	ifm = nlmsg_data(nlh);
5307 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5308 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5309 		return -EINVAL;
5310 	}
5311 
5312 	fillargs->ifindex = ifm->ifa_index;
5313 	if (fillargs->ifindex) {
5314 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5315 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5316 	}
5317 
5318 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5319 					    ifa_ipv6_policy, extack);
5320 	if (err < 0)
5321 		return err;
5322 
5323 	for (i = 0; i <= IFA_MAX; ++i) {
5324 		if (!tb[i])
5325 			continue;
5326 
5327 		if (i == IFA_TARGET_NETNSID) {
5328 			struct net *net;
5329 
5330 			fillargs->netnsid = nla_get_s32(tb[i]);
5331 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5332 			if (IS_ERR(net)) {
5333 				fillargs->netnsid = -1;
5334 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5335 				return PTR_ERR(net);
5336 			}
5337 			*tgt_net = net;
5338 		} else {
5339 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5340 			return -EINVAL;
5341 		}
5342 	}
5343 
5344 	return 0;
5345 }
5346 
5347 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5348 			   enum addr_type_t type)
5349 {
5350 	const struct nlmsghdr *nlh = cb->nlh;
5351 	struct inet6_fill_args fillargs = {
5352 		.portid = NETLINK_CB(cb->skb).portid,
5353 		.seq = cb->nlh->nlmsg_seq,
5354 		.flags = NLM_F_MULTI,
5355 		.netnsid = -1,
5356 		.type = type,
5357 	};
5358 	struct net *tgt_net = sock_net(skb->sk);
5359 	int idx, s_idx, s_ip_idx;
5360 	int h, s_h;
5361 	struct net_device *dev;
5362 	struct inet6_dev *idev;
5363 	struct hlist_head *head;
5364 	int err = 0;
5365 
5366 	s_h = cb->args[0];
5367 	s_idx = idx = cb->args[1];
5368 	s_ip_idx = cb->args[2];
5369 
5370 	if (cb->strict_check) {
5371 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5372 						  skb->sk, cb);
5373 		if (err < 0)
5374 			goto put_tgt_net;
5375 
5376 		err = 0;
5377 		if (fillargs.ifindex) {
5378 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5379 			if (!dev) {
5380 				err = -ENODEV;
5381 				goto put_tgt_net;
5382 			}
5383 			idev = __in6_dev_get(dev);
5384 			if (idev) {
5385 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5386 						     &fillargs);
5387 				if (err > 0)
5388 					err = 0;
5389 			}
5390 			goto put_tgt_net;
5391 		}
5392 	}
5393 
5394 	rcu_read_lock();
5395 	cb->seq = inet6_base_seq(tgt_net);
5396 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5397 		idx = 0;
5398 		head = &tgt_net->dev_index_head[h];
5399 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5400 			if (idx < s_idx)
5401 				goto cont;
5402 			if (h > s_h || idx > s_idx)
5403 				s_ip_idx = 0;
5404 			idev = __in6_dev_get(dev);
5405 			if (!idev)
5406 				goto cont;
5407 
5408 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5409 					   &fillargs) < 0)
5410 				goto done;
5411 cont:
5412 			idx++;
5413 		}
5414 	}
5415 done:
5416 	rcu_read_unlock();
5417 	cb->args[0] = h;
5418 	cb->args[1] = idx;
5419 put_tgt_net:
5420 	if (fillargs.netnsid >= 0)
5421 		put_net(tgt_net);
5422 
5423 	return skb->len ? : err;
5424 }
5425 
5426 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5427 {
5428 	enum addr_type_t type = UNICAST_ADDR;
5429 
5430 	return inet6_dump_addr(skb, cb, type);
5431 }
5432 
5433 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5434 {
5435 	enum addr_type_t type = MULTICAST_ADDR;
5436 
5437 	return inet6_dump_addr(skb, cb, type);
5438 }
5439 
5440 
5441 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5442 {
5443 	enum addr_type_t type = ANYCAST_ADDR;
5444 
5445 	return inet6_dump_addr(skb, cb, type);
5446 }
5447 
5448 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5449 				       const struct nlmsghdr *nlh,
5450 				       struct nlattr **tb,
5451 				       struct netlink_ext_ack *extack)
5452 {
5453 	struct ifaddrmsg *ifm;
5454 	int i, err;
5455 
5456 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5457 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5458 		return -EINVAL;
5459 	}
5460 
5461 	if (!netlink_strict_get_check(skb))
5462 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5463 					      ifa_ipv6_policy, extack);
5464 
5465 	ifm = nlmsg_data(nlh);
5466 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5467 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5468 		return -EINVAL;
5469 	}
5470 
5471 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5472 					    ifa_ipv6_policy, extack);
5473 	if (err)
5474 		return err;
5475 
5476 	for (i = 0; i <= IFA_MAX; i++) {
5477 		if (!tb[i])
5478 			continue;
5479 
5480 		switch (i) {
5481 		case IFA_TARGET_NETNSID:
5482 		case IFA_ADDRESS:
5483 		case IFA_LOCAL:
5484 			break;
5485 		default:
5486 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5487 			return -EINVAL;
5488 		}
5489 	}
5490 
5491 	return 0;
5492 }
5493 
5494 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5495 			     struct netlink_ext_ack *extack)
5496 {
5497 	struct net *tgt_net = sock_net(in_skb->sk);
5498 	struct inet6_fill_args fillargs = {
5499 		.portid = NETLINK_CB(in_skb).portid,
5500 		.seq = nlh->nlmsg_seq,
5501 		.event = RTM_NEWADDR,
5502 		.flags = 0,
5503 		.netnsid = -1,
5504 	};
5505 	struct ifaddrmsg *ifm;
5506 	struct nlattr *tb[IFA_MAX+1];
5507 	struct in6_addr *addr = NULL, *peer;
5508 	struct net_device *dev = NULL;
5509 	struct inet6_ifaddr *ifa;
5510 	struct sk_buff *skb;
5511 	int err;
5512 
5513 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5514 	if (err < 0)
5515 		return err;
5516 
5517 	if (tb[IFA_TARGET_NETNSID]) {
5518 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5519 
5520 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5521 						  fillargs.netnsid);
5522 		if (IS_ERR(tgt_net))
5523 			return PTR_ERR(tgt_net);
5524 	}
5525 
5526 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5527 	if (!addr) {
5528 		err = -EINVAL;
5529 		goto errout;
5530 	}
5531 	ifm = nlmsg_data(nlh);
5532 	if (ifm->ifa_index)
5533 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5534 
5535 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5536 	if (!ifa) {
5537 		err = -EADDRNOTAVAIL;
5538 		goto errout;
5539 	}
5540 
5541 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5542 	if (!skb) {
5543 		err = -ENOBUFS;
5544 		goto errout_ifa;
5545 	}
5546 
5547 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5548 	if (err < 0) {
5549 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5550 		WARN_ON(err == -EMSGSIZE);
5551 		kfree_skb(skb);
5552 		goto errout_ifa;
5553 	}
5554 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5555 errout_ifa:
5556 	in6_ifa_put(ifa);
5557 errout:
5558 	dev_put(dev);
5559 	if (fillargs.netnsid >= 0)
5560 		put_net(tgt_net);
5561 
5562 	return err;
5563 }
5564 
5565 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5566 {
5567 	struct sk_buff *skb;
5568 	struct net *net = dev_net(ifa->idev->dev);
5569 	struct inet6_fill_args fillargs = {
5570 		.portid = 0,
5571 		.seq = 0,
5572 		.event = event,
5573 		.flags = 0,
5574 		.netnsid = -1,
5575 	};
5576 	int err = -ENOBUFS;
5577 
5578 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5579 	if (!skb)
5580 		goto errout;
5581 
5582 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5583 	if (err < 0) {
5584 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5585 		WARN_ON(err == -EMSGSIZE);
5586 		kfree_skb(skb);
5587 		goto errout;
5588 	}
5589 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5590 	return;
5591 errout:
5592 	if (err < 0)
5593 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5594 }
5595 
5596 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5597 				__s32 *array, int bytes)
5598 {
5599 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5600 
5601 	memset(array, 0, bytes);
5602 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5603 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5604 	array[DEVCONF_MTU6] = cnf->mtu6;
5605 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5606 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5607 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5608 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5609 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5610 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5611 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5612 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5613 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5614 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5615 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5616 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5617 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5618 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5619 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5620 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5621 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5622 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5623 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5624 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5625 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5626 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5627 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5628 	array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5629 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5630 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5631 #ifdef CONFIG_IPV6_ROUTER_PREF
5632 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5633 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5634 		jiffies_to_msecs(cnf->rtr_probe_interval);
5635 #ifdef CONFIG_IPV6_ROUTE_INFO
5636 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5637 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5638 #endif
5639 #endif
5640 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5641 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5642 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5643 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5644 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5645 #endif
5646 #ifdef CONFIG_IPV6_MROUTE
5647 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5648 #endif
5649 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5650 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5651 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5652 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5653 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5654 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5655 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5656 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5657 	/* we omit DEVCONF_STABLE_SECRET for now */
5658 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5659 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5660 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5661 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5662 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5663 #ifdef CONFIG_IPV6_SEG6_HMAC
5664 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5665 #endif
5666 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5667 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5668 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5669 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5670 	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5671 	array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5672 	array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5673 	array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5674 	array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5675 	array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5676 	array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5677 }
5678 
5679 static inline size_t inet6_ifla6_size(void)
5680 {
5681 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5682 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5683 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5684 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5685 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5686 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5687 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5688 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5689 	     + 0;
5690 }
5691 
5692 static inline size_t inet6_if_nlmsg_size(void)
5693 {
5694 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5695 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5696 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5697 	       + nla_total_size(4) /* IFLA_MTU */
5698 	       + nla_total_size(4) /* IFLA_LINK */
5699 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5700 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5701 }
5702 
5703 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5704 					int bytes)
5705 {
5706 	int i;
5707 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5708 	BUG_ON(pad < 0);
5709 
5710 	/* Use put_unaligned() because stats may not be aligned for u64. */
5711 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5712 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5713 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5714 
5715 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5716 }
5717 
5718 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5719 					int bytes, size_t syncpoff)
5720 {
5721 	int i, c;
5722 	u64 buff[IPSTATS_MIB_MAX];
5723 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5724 
5725 	BUG_ON(pad < 0);
5726 
5727 	memset(buff, 0, sizeof(buff));
5728 	buff[0] = IPSTATS_MIB_MAX;
5729 
5730 	for_each_possible_cpu(c) {
5731 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5732 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5733 	}
5734 
5735 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5736 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5737 }
5738 
5739 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5740 			     int bytes)
5741 {
5742 	switch (attrtype) {
5743 	case IFLA_INET6_STATS:
5744 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5745 				     offsetof(struct ipstats_mib, syncp));
5746 		break;
5747 	case IFLA_INET6_ICMP6STATS:
5748 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5749 		break;
5750 	}
5751 }
5752 
5753 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5754 				  u32 ext_filter_mask)
5755 {
5756 	struct nlattr *nla;
5757 	struct ifla_cacheinfo ci;
5758 
5759 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5760 		goto nla_put_failure;
5761 	ci.max_reasm_len = IPV6_MAXPLEN;
5762 	ci.tstamp = cstamp_delta(idev->tstamp);
5763 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5764 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5765 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5766 		goto nla_put_failure;
5767 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5768 	if (!nla)
5769 		goto nla_put_failure;
5770 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5771 
5772 	/* XXX - MC not implemented */
5773 
5774 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5775 		return 0;
5776 
5777 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5778 	if (!nla)
5779 		goto nla_put_failure;
5780 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5781 
5782 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5783 	if (!nla)
5784 		goto nla_put_failure;
5785 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5786 
5787 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5788 	if (!nla)
5789 		goto nla_put_failure;
5790 	read_lock_bh(&idev->lock);
5791 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5792 	read_unlock_bh(&idev->lock);
5793 
5794 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5795 		goto nla_put_failure;
5796 
5797 	if (idev->ra_mtu &&
5798 	    nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5799 		goto nla_put_failure;
5800 
5801 	return 0;
5802 
5803 nla_put_failure:
5804 	return -EMSGSIZE;
5805 }
5806 
5807 static size_t inet6_get_link_af_size(const struct net_device *dev,
5808 				     u32 ext_filter_mask)
5809 {
5810 	if (!__in6_dev_get(dev))
5811 		return 0;
5812 
5813 	return inet6_ifla6_size();
5814 }
5815 
5816 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5817 			      u32 ext_filter_mask)
5818 {
5819 	struct inet6_dev *idev = __in6_dev_get(dev);
5820 
5821 	if (!idev)
5822 		return -ENODATA;
5823 
5824 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5825 		return -EMSGSIZE;
5826 
5827 	return 0;
5828 }
5829 
5830 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5831 			     struct netlink_ext_ack *extack)
5832 {
5833 	struct inet6_ifaddr *ifp;
5834 	struct net_device *dev = idev->dev;
5835 	bool clear_token, update_rs = false;
5836 	struct in6_addr ll_addr;
5837 
5838 	ASSERT_RTNL();
5839 
5840 	if (!token)
5841 		return -EINVAL;
5842 
5843 	if (dev->flags & IFF_LOOPBACK) {
5844 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5845 		return -EINVAL;
5846 	}
5847 
5848 	if (dev->flags & IFF_NOARP) {
5849 		NL_SET_ERR_MSG_MOD(extack,
5850 				   "Device does not do neighbour discovery");
5851 		return -EINVAL;
5852 	}
5853 
5854 	if (!ipv6_accept_ra(idev)) {
5855 		NL_SET_ERR_MSG_MOD(extack,
5856 				   "Router advertisement is disabled on device");
5857 		return -EINVAL;
5858 	}
5859 
5860 	if (idev->cnf.rtr_solicits == 0) {
5861 		NL_SET_ERR_MSG(extack,
5862 			       "Router solicitation is disabled on device");
5863 		return -EINVAL;
5864 	}
5865 
5866 	write_lock_bh(&idev->lock);
5867 
5868 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5869 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5870 
5871 	write_unlock_bh(&idev->lock);
5872 
5873 	clear_token = ipv6_addr_any(token);
5874 	if (clear_token)
5875 		goto update_lft;
5876 
5877 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5878 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5879 			     IFA_F_OPTIMISTIC)) {
5880 		/* If we're not ready, then normal ifup will take care
5881 		 * of this. Otherwise, we need to request our rs here.
5882 		 */
5883 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5884 		update_rs = true;
5885 	}
5886 
5887 update_lft:
5888 	write_lock_bh(&idev->lock);
5889 
5890 	if (update_rs) {
5891 		idev->if_flags |= IF_RS_SENT;
5892 		idev->rs_interval = rfc3315_s14_backoff_init(
5893 			idev->cnf.rtr_solicit_interval);
5894 		idev->rs_probes = 1;
5895 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5896 	}
5897 
5898 	/* Well, that's kinda nasty ... */
5899 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5900 		spin_lock(&ifp->lock);
5901 		if (ifp->tokenized) {
5902 			ifp->valid_lft = 0;
5903 			ifp->prefered_lft = 0;
5904 		}
5905 		spin_unlock(&ifp->lock);
5906 	}
5907 
5908 	write_unlock_bh(&idev->lock);
5909 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5910 	addrconf_verify_rtnl(dev_net(dev));
5911 	return 0;
5912 }
5913 
5914 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5915 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5916 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5917 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5918 					    .reject_message =
5919 						"IFLA_INET6_RA_MTU can not be set" },
5920 };
5921 
5922 static int check_addr_gen_mode(int mode)
5923 {
5924 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5925 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5926 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5927 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5928 		return -EINVAL;
5929 	return 1;
5930 }
5931 
5932 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5933 				int mode)
5934 {
5935 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5936 	    !idev->cnf.stable_secret.initialized &&
5937 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5938 		return -EINVAL;
5939 	return 1;
5940 }
5941 
5942 static int inet6_validate_link_af(const struct net_device *dev,
5943 				  const struct nlattr *nla,
5944 				  struct netlink_ext_ack *extack)
5945 {
5946 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5947 	struct inet6_dev *idev = NULL;
5948 	int err;
5949 
5950 	if (dev) {
5951 		idev = __in6_dev_get(dev);
5952 		if (!idev)
5953 			return -EAFNOSUPPORT;
5954 	}
5955 
5956 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5957 					  inet6_af_policy, extack);
5958 	if (err)
5959 		return err;
5960 
5961 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5962 		return -EINVAL;
5963 
5964 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5965 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5966 
5967 		if (check_addr_gen_mode(mode) < 0)
5968 			return -EINVAL;
5969 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5970 			return -EINVAL;
5971 	}
5972 
5973 	return 0;
5974 }
5975 
5976 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5977 			     struct netlink_ext_ack *extack)
5978 {
5979 	struct inet6_dev *idev = __in6_dev_get(dev);
5980 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5981 	int err;
5982 
5983 	if (!idev)
5984 		return -EAFNOSUPPORT;
5985 
5986 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5987 		return -EINVAL;
5988 
5989 	if (tb[IFLA_INET6_TOKEN]) {
5990 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5991 					extack);
5992 		if (err)
5993 			return err;
5994 	}
5995 
5996 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5997 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5998 
5999 		idev->cnf.addr_gen_mode = mode;
6000 	}
6001 
6002 	return 0;
6003 }
6004 
6005 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6006 			     u32 portid, u32 seq, int event, unsigned int flags)
6007 {
6008 	struct net_device *dev = idev->dev;
6009 	struct ifinfomsg *hdr;
6010 	struct nlmsghdr *nlh;
6011 	void *protoinfo;
6012 
6013 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6014 	if (!nlh)
6015 		return -EMSGSIZE;
6016 
6017 	hdr = nlmsg_data(nlh);
6018 	hdr->ifi_family = AF_INET6;
6019 	hdr->__ifi_pad = 0;
6020 	hdr->ifi_type = dev->type;
6021 	hdr->ifi_index = dev->ifindex;
6022 	hdr->ifi_flags = dev_get_flags(dev);
6023 	hdr->ifi_change = 0;
6024 
6025 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6026 	    (dev->addr_len &&
6027 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6028 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
6029 	    (dev->ifindex != dev_get_iflink(dev) &&
6030 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
6031 	    nla_put_u8(skb, IFLA_OPERSTATE,
6032 		       netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6033 		goto nla_put_failure;
6034 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6035 	if (!protoinfo)
6036 		goto nla_put_failure;
6037 
6038 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6039 		goto nla_put_failure;
6040 
6041 	nla_nest_end(skb, protoinfo);
6042 	nlmsg_end(skb, nlh);
6043 	return 0;
6044 
6045 nla_put_failure:
6046 	nlmsg_cancel(skb, nlh);
6047 	return -EMSGSIZE;
6048 }
6049 
6050 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6051 				   struct netlink_ext_ack *extack)
6052 {
6053 	struct ifinfomsg *ifm;
6054 
6055 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6056 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6057 		return -EINVAL;
6058 	}
6059 
6060 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6061 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6062 		return -EINVAL;
6063 	}
6064 
6065 	ifm = nlmsg_data(nlh);
6066 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6067 	    ifm->ifi_change || ifm->ifi_index) {
6068 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6069 		return -EINVAL;
6070 	}
6071 
6072 	return 0;
6073 }
6074 
6075 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6076 {
6077 	struct net *net = sock_net(skb->sk);
6078 	int h, s_h;
6079 	int idx = 0, s_idx;
6080 	struct net_device *dev;
6081 	struct inet6_dev *idev;
6082 	struct hlist_head *head;
6083 
6084 	/* only requests using strict checking can pass data to
6085 	 * influence the dump
6086 	 */
6087 	if (cb->strict_check) {
6088 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6089 
6090 		if (err < 0)
6091 			return err;
6092 	}
6093 
6094 	s_h = cb->args[0];
6095 	s_idx = cb->args[1];
6096 
6097 	rcu_read_lock();
6098 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6099 		idx = 0;
6100 		head = &net->dev_index_head[h];
6101 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
6102 			if (idx < s_idx)
6103 				goto cont;
6104 			idev = __in6_dev_get(dev);
6105 			if (!idev)
6106 				goto cont;
6107 			if (inet6_fill_ifinfo(skb, idev,
6108 					      NETLINK_CB(cb->skb).portid,
6109 					      cb->nlh->nlmsg_seq,
6110 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
6111 				goto out;
6112 cont:
6113 			idx++;
6114 		}
6115 	}
6116 out:
6117 	rcu_read_unlock();
6118 	cb->args[1] = idx;
6119 	cb->args[0] = h;
6120 
6121 	return skb->len;
6122 }
6123 
6124 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6125 {
6126 	struct sk_buff *skb;
6127 	struct net *net = dev_net(idev->dev);
6128 	int err = -ENOBUFS;
6129 
6130 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6131 	if (!skb)
6132 		goto errout;
6133 
6134 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6135 	if (err < 0) {
6136 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6137 		WARN_ON(err == -EMSGSIZE);
6138 		kfree_skb(skb);
6139 		goto errout;
6140 	}
6141 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6142 	return;
6143 errout:
6144 	if (err < 0)
6145 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6146 }
6147 
6148 static inline size_t inet6_prefix_nlmsg_size(void)
6149 {
6150 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6151 	       + nla_total_size(sizeof(struct in6_addr))
6152 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6153 }
6154 
6155 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6156 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6157 			     int event, unsigned int flags)
6158 {
6159 	struct prefixmsg *pmsg;
6160 	struct nlmsghdr *nlh;
6161 	struct prefix_cacheinfo	ci;
6162 
6163 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6164 	if (!nlh)
6165 		return -EMSGSIZE;
6166 
6167 	pmsg = nlmsg_data(nlh);
6168 	pmsg->prefix_family = AF_INET6;
6169 	pmsg->prefix_pad1 = 0;
6170 	pmsg->prefix_pad2 = 0;
6171 	pmsg->prefix_ifindex = idev->dev->ifindex;
6172 	pmsg->prefix_len = pinfo->prefix_len;
6173 	pmsg->prefix_type = pinfo->type;
6174 	pmsg->prefix_pad3 = 0;
6175 	pmsg->prefix_flags = pinfo->flags;
6176 
6177 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6178 		goto nla_put_failure;
6179 	ci.preferred_time = ntohl(pinfo->prefered);
6180 	ci.valid_time = ntohl(pinfo->valid);
6181 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6182 		goto nla_put_failure;
6183 	nlmsg_end(skb, nlh);
6184 	return 0;
6185 
6186 nla_put_failure:
6187 	nlmsg_cancel(skb, nlh);
6188 	return -EMSGSIZE;
6189 }
6190 
6191 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6192 			 struct prefix_info *pinfo)
6193 {
6194 	struct sk_buff *skb;
6195 	struct net *net = dev_net(idev->dev);
6196 	int err = -ENOBUFS;
6197 
6198 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6199 	if (!skb)
6200 		goto errout;
6201 
6202 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6203 	if (err < 0) {
6204 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6205 		WARN_ON(err == -EMSGSIZE);
6206 		kfree_skb(skb);
6207 		goto errout;
6208 	}
6209 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6210 	return;
6211 errout:
6212 	if (err < 0)
6213 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6214 }
6215 
6216 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6217 {
6218 	struct net *net = dev_net(ifp->idev->dev);
6219 
6220 	if (event)
6221 		ASSERT_RTNL();
6222 
6223 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6224 
6225 	switch (event) {
6226 	case RTM_NEWADDR:
6227 		/*
6228 		 * If the address was optimistic we inserted the route at the
6229 		 * start of our DAD process, so we don't need to do it again.
6230 		 * If the device was taken down in the middle of the DAD
6231 		 * cycle there is a race where we could get here without a
6232 		 * host route, so nothing to insert. That will be fixed when
6233 		 * the device is brought up.
6234 		 */
6235 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6236 			ip6_ins_rt(net, ifp->rt);
6237 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6238 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6239 				&ifp->addr, ifp->idev->dev->name);
6240 		}
6241 
6242 		if (ifp->idev->cnf.forwarding)
6243 			addrconf_join_anycast(ifp);
6244 		if (!ipv6_addr_any(&ifp->peer_addr))
6245 			addrconf_prefix_route(&ifp->peer_addr, 128,
6246 					      ifp->rt_priority, ifp->idev->dev,
6247 					      0, 0, GFP_ATOMIC);
6248 		break;
6249 	case RTM_DELADDR:
6250 		if (ifp->idev->cnf.forwarding)
6251 			addrconf_leave_anycast(ifp);
6252 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6253 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6254 			struct fib6_info *rt;
6255 
6256 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6257 						       ifp->idev->dev, 0, 0,
6258 						       false);
6259 			if (rt)
6260 				ip6_del_rt(net, rt, false);
6261 		}
6262 		if (ifp->rt) {
6263 			ip6_del_rt(net, ifp->rt, false);
6264 			ifp->rt = NULL;
6265 		}
6266 		rt_genid_bump_ipv6(net);
6267 		break;
6268 	}
6269 	atomic_inc(&net->ipv6.dev_addr_genid);
6270 }
6271 
6272 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6273 {
6274 	if (likely(ifp->idev->dead == 0))
6275 		__ipv6_ifa_notify(event, ifp);
6276 }
6277 
6278 #ifdef CONFIG_SYSCTL
6279 
6280 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6281 		void *buffer, size_t *lenp, loff_t *ppos)
6282 {
6283 	int *valp = ctl->data;
6284 	int val = *valp;
6285 	loff_t pos = *ppos;
6286 	struct ctl_table lctl;
6287 	int ret;
6288 
6289 	/*
6290 	 * ctl->data points to idev->cnf.forwarding, we should
6291 	 * not modify it until we get the rtnl lock.
6292 	 */
6293 	lctl = *ctl;
6294 	lctl.data = &val;
6295 
6296 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6297 
6298 	if (write)
6299 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6300 	if (ret)
6301 		*ppos = pos;
6302 	return ret;
6303 }
6304 
6305 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6306 		void *buffer, size_t *lenp, loff_t *ppos)
6307 {
6308 	struct inet6_dev *idev = ctl->extra1;
6309 	int min_mtu = IPV6_MIN_MTU;
6310 	struct ctl_table lctl;
6311 
6312 	lctl = *ctl;
6313 	lctl.extra1 = &min_mtu;
6314 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6315 
6316 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6317 }
6318 
6319 static void dev_disable_change(struct inet6_dev *idev)
6320 {
6321 	struct netdev_notifier_info info;
6322 
6323 	if (!idev || !idev->dev)
6324 		return;
6325 
6326 	netdev_notifier_info_init(&info, idev->dev);
6327 	if (idev->cnf.disable_ipv6)
6328 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6329 	else
6330 		addrconf_notify(NULL, NETDEV_UP, &info);
6331 }
6332 
6333 static void addrconf_disable_change(struct net *net, __s32 newf)
6334 {
6335 	struct net_device *dev;
6336 	struct inet6_dev *idev;
6337 
6338 	for_each_netdev(net, dev) {
6339 		idev = __in6_dev_get(dev);
6340 		if (idev) {
6341 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6342 
6343 			WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6344 			if (changed)
6345 				dev_disable_change(idev);
6346 		}
6347 	}
6348 }
6349 
6350 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6351 {
6352 	struct net *net;
6353 	int old;
6354 
6355 	if (!rtnl_trylock())
6356 		return restart_syscall();
6357 
6358 	net = (struct net *)table->extra2;
6359 	old = *p;
6360 	WRITE_ONCE(*p, newf);
6361 
6362 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6363 		rtnl_unlock();
6364 		return 0;
6365 	}
6366 
6367 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6368 		WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6369 		addrconf_disable_change(net, newf);
6370 	} else if ((!newf) ^ (!old))
6371 		dev_disable_change((struct inet6_dev *)table->extra1);
6372 
6373 	rtnl_unlock();
6374 	return 0;
6375 }
6376 
6377 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6378 		void *buffer, size_t *lenp, loff_t *ppos)
6379 {
6380 	int *valp = ctl->data;
6381 	int val = *valp;
6382 	loff_t pos = *ppos;
6383 	struct ctl_table lctl;
6384 	int ret;
6385 
6386 	/*
6387 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6388 	 * not modify it until we get the rtnl lock.
6389 	 */
6390 	lctl = *ctl;
6391 	lctl.data = &val;
6392 
6393 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6394 
6395 	if (write)
6396 		ret = addrconf_disable_ipv6(ctl, valp, val);
6397 	if (ret)
6398 		*ppos = pos;
6399 	return ret;
6400 }
6401 
6402 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6403 		void *buffer, size_t *lenp, loff_t *ppos)
6404 {
6405 	int *valp = ctl->data;
6406 	int ret;
6407 	int old, new;
6408 
6409 	old = *valp;
6410 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6411 	new = *valp;
6412 
6413 	if (write && old != new) {
6414 		struct net *net = ctl->extra2;
6415 
6416 		if (!rtnl_trylock())
6417 			return restart_syscall();
6418 
6419 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6420 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6421 						     NETCONFA_PROXY_NEIGH,
6422 						     NETCONFA_IFINDEX_DEFAULT,
6423 						     net->ipv6.devconf_dflt);
6424 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6425 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6426 						     NETCONFA_PROXY_NEIGH,
6427 						     NETCONFA_IFINDEX_ALL,
6428 						     net->ipv6.devconf_all);
6429 		else {
6430 			struct inet6_dev *idev = ctl->extra1;
6431 
6432 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6433 						     NETCONFA_PROXY_NEIGH,
6434 						     idev->dev->ifindex,
6435 						     &idev->cnf);
6436 		}
6437 		rtnl_unlock();
6438 	}
6439 
6440 	return ret;
6441 }
6442 
6443 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6444 					 void *buffer, size_t *lenp,
6445 					 loff_t *ppos)
6446 {
6447 	int ret = 0;
6448 	u32 new_val;
6449 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6450 	struct net *net = (struct net *)ctl->extra2;
6451 	struct ctl_table tmp = {
6452 		.data = &new_val,
6453 		.maxlen = sizeof(new_val),
6454 		.mode = ctl->mode,
6455 	};
6456 
6457 	if (!rtnl_trylock())
6458 		return restart_syscall();
6459 
6460 	new_val = *((u32 *)ctl->data);
6461 
6462 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6463 	if (ret != 0)
6464 		goto out;
6465 
6466 	if (write) {
6467 		if (check_addr_gen_mode(new_val) < 0) {
6468 			ret = -EINVAL;
6469 			goto out;
6470 		}
6471 
6472 		if (idev) {
6473 			if (check_stable_privacy(idev, net, new_val) < 0) {
6474 				ret = -EINVAL;
6475 				goto out;
6476 			}
6477 
6478 			if (idev->cnf.addr_gen_mode != new_val) {
6479 				idev->cnf.addr_gen_mode = new_val;
6480 				addrconf_init_auto_addrs(idev->dev);
6481 			}
6482 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6483 			struct net_device *dev;
6484 
6485 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6486 			for_each_netdev(net, dev) {
6487 				idev = __in6_dev_get(dev);
6488 				if (idev &&
6489 				    idev->cnf.addr_gen_mode != new_val) {
6490 					idev->cnf.addr_gen_mode = new_val;
6491 					addrconf_init_auto_addrs(idev->dev);
6492 				}
6493 			}
6494 		}
6495 
6496 		*((u32 *)ctl->data) = new_val;
6497 	}
6498 
6499 out:
6500 	rtnl_unlock();
6501 
6502 	return ret;
6503 }
6504 
6505 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6506 					 void *buffer, size_t *lenp,
6507 					 loff_t *ppos)
6508 {
6509 	int err;
6510 	struct in6_addr addr;
6511 	char str[IPV6_MAX_STRLEN];
6512 	struct ctl_table lctl = *ctl;
6513 	struct net *net = ctl->extra2;
6514 	struct ipv6_stable_secret *secret = ctl->data;
6515 
6516 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6517 		return -EIO;
6518 
6519 	lctl.maxlen = IPV6_MAX_STRLEN;
6520 	lctl.data = str;
6521 
6522 	if (!rtnl_trylock())
6523 		return restart_syscall();
6524 
6525 	if (!write && !secret->initialized) {
6526 		err = -EIO;
6527 		goto out;
6528 	}
6529 
6530 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6531 	if (err >= sizeof(str)) {
6532 		err = -EIO;
6533 		goto out;
6534 	}
6535 
6536 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6537 	if (err || !write)
6538 		goto out;
6539 
6540 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6541 		err = -EIO;
6542 		goto out;
6543 	}
6544 
6545 	secret->initialized = true;
6546 	secret->secret = addr;
6547 
6548 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6549 		struct net_device *dev;
6550 
6551 		for_each_netdev(net, dev) {
6552 			struct inet6_dev *idev = __in6_dev_get(dev);
6553 
6554 			if (idev) {
6555 				idev->cnf.addr_gen_mode =
6556 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6557 			}
6558 		}
6559 	} else {
6560 		struct inet6_dev *idev = ctl->extra1;
6561 
6562 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6563 	}
6564 
6565 out:
6566 	rtnl_unlock();
6567 
6568 	return err;
6569 }
6570 
6571 static
6572 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6573 						int write, void *buffer,
6574 						size_t *lenp,
6575 						loff_t *ppos)
6576 {
6577 	int *valp = ctl->data;
6578 	int val = *valp;
6579 	loff_t pos = *ppos;
6580 	struct ctl_table lctl;
6581 	int ret;
6582 
6583 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6584 	 * we should not modify it until we get the rtnl lock.
6585 	 */
6586 	lctl = *ctl;
6587 	lctl.data = &val;
6588 
6589 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6590 
6591 	if (write)
6592 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6593 	if (ret)
6594 		*ppos = pos;
6595 	return ret;
6596 }
6597 
6598 static
6599 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6600 {
6601 	if (rt) {
6602 		if (action)
6603 			rt->dst.flags |= DST_NOPOLICY;
6604 		else
6605 			rt->dst.flags &= ~DST_NOPOLICY;
6606 	}
6607 }
6608 
6609 static
6610 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6611 {
6612 	struct inet6_ifaddr *ifa;
6613 
6614 	read_lock_bh(&idev->lock);
6615 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6616 		spin_lock(&ifa->lock);
6617 		if (ifa->rt) {
6618 			/* host routes only use builtin fib6_nh */
6619 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6620 			int cpu;
6621 
6622 			rcu_read_lock();
6623 			ifa->rt->dst_nopolicy = val ? true : false;
6624 			if (nh->rt6i_pcpu) {
6625 				for_each_possible_cpu(cpu) {
6626 					struct rt6_info **rtp;
6627 
6628 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6629 					addrconf_set_nopolicy(*rtp, val);
6630 				}
6631 			}
6632 			rcu_read_unlock();
6633 		}
6634 		spin_unlock(&ifa->lock);
6635 	}
6636 	read_unlock_bh(&idev->lock);
6637 }
6638 
6639 static
6640 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6641 {
6642 	struct inet6_dev *idev;
6643 	struct net *net;
6644 
6645 	if (!rtnl_trylock())
6646 		return restart_syscall();
6647 
6648 	*valp = val;
6649 
6650 	net = (struct net *)ctl->extra2;
6651 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6652 		rtnl_unlock();
6653 		return 0;
6654 	}
6655 
6656 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6657 		struct net_device *dev;
6658 
6659 		for_each_netdev(net, dev) {
6660 			idev = __in6_dev_get(dev);
6661 			if (idev)
6662 				addrconf_disable_policy_idev(idev, val);
6663 		}
6664 	} else {
6665 		idev = (struct inet6_dev *)ctl->extra1;
6666 		addrconf_disable_policy_idev(idev, val);
6667 	}
6668 
6669 	rtnl_unlock();
6670 	return 0;
6671 }
6672 
6673 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6674 				   void *buffer, size_t *lenp, loff_t *ppos)
6675 {
6676 	int *valp = ctl->data;
6677 	int val = *valp;
6678 	loff_t pos = *ppos;
6679 	struct ctl_table lctl;
6680 	int ret;
6681 
6682 	lctl = *ctl;
6683 	lctl.data = &val;
6684 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6685 
6686 	if (write && (*valp != val))
6687 		ret = addrconf_disable_policy(ctl, valp, val);
6688 
6689 	if (ret)
6690 		*ppos = pos;
6691 
6692 	return ret;
6693 }
6694 
6695 static int minus_one = -1;
6696 static const int two_five_five = 255;
6697 static u32 ioam6_if_id_max = U16_MAX;
6698 
6699 static const struct ctl_table addrconf_sysctl[] = {
6700 	{
6701 		.procname	= "forwarding",
6702 		.data		= &ipv6_devconf.forwarding,
6703 		.maxlen		= sizeof(int),
6704 		.mode		= 0644,
6705 		.proc_handler	= addrconf_sysctl_forward,
6706 	},
6707 	{
6708 		.procname	= "hop_limit",
6709 		.data		= &ipv6_devconf.hop_limit,
6710 		.maxlen		= sizeof(int),
6711 		.mode		= 0644,
6712 		.proc_handler	= proc_dointvec_minmax,
6713 		.extra1		= (void *)SYSCTL_ONE,
6714 		.extra2		= (void *)&two_five_five,
6715 	},
6716 	{
6717 		.procname	= "mtu",
6718 		.data		= &ipv6_devconf.mtu6,
6719 		.maxlen		= sizeof(int),
6720 		.mode		= 0644,
6721 		.proc_handler	= addrconf_sysctl_mtu,
6722 	},
6723 	{
6724 		.procname	= "accept_ra",
6725 		.data		= &ipv6_devconf.accept_ra,
6726 		.maxlen		= sizeof(int),
6727 		.mode		= 0644,
6728 		.proc_handler	= proc_dointvec,
6729 	},
6730 	{
6731 		.procname	= "accept_redirects",
6732 		.data		= &ipv6_devconf.accept_redirects,
6733 		.maxlen		= sizeof(int),
6734 		.mode		= 0644,
6735 		.proc_handler	= proc_dointvec,
6736 	},
6737 	{
6738 		.procname	= "autoconf",
6739 		.data		= &ipv6_devconf.autoconf,
6740 		.maxlen		= sizeof(int),
6741 		.mode		= 0644,
6742 		.proc_handler	= proc_dointvec,
6743 	},
6744 	{
6745 		.procname	= "dad_transmits",
6746 		.data		= &ipv6_devconf.dad_transmits,
6747 		.maxlen		= sizeof(int),
6748 		.mode		= 0644,
6749 		.proc_handler	= proc_dointvec,
6750 	},
6751 	{
6752 		.procname	= "router_solicitations",
6753 		.data		= &ipv6_devconf.rtr_solicits,
6754 		.maxlen		= sizeof(int),
6755 		.mode		= 0644,
6756 		.proc_handler	= proc_dointvec_minmax,
6757 		.extra1		= &minus_one,
6758 	},
6759 	{
6760 		.procname	= "router_solicitation_interval",
6761 		.data		= &ipv6_devconf.rtr_solicit_interval,
6762 		.maxlen		= sizeof(int),
6763 		.mode		= 0644,
6764 		.proc_handler	= proc_dointvec_jiffies,
6765 	},
6766 	{
6767 		.procname	= "router_solicitation_max_interval",
6768 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6769 		.maxlen		= sizeof(int),
6770 		.mode		= 0644,
6771 		.proc_handler	= proc_dointvec_jiffies,
6772 	},
6773 	{
6774 		.procname	= "router_solicitation_delay",
6775 		.data		= &ipv6_devconf.rtr_solicit_delay,
6776 		.maxlen		= sizeof(int),
6777 		.mode		= 0644,
6778 		.proc_handler	= proc_dointvec_jiffies,
6779 	},
6780 	{
6781 		.procname	= "force_mld_version",
6782 		.data		= &ipv6_devconf.force_mld_version,
6783 		.maxlen		= sizeof(int),
6784 		.mode		= 0644,
6785 		.proc_handler	= proc_dointvec,
6786 	},
6787 	{
6788 		.procname	= "mldv1_unsolicited_report_interval",
6789 		.data		=
6790 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6791 		.maxlen		= sizeof(int),
6792 		.mode		= 0644,
6793 		.proc_handler	= proc_dointvec_ms_jiffies,
6794 	},
6795 	{
6796 		.procname	= "mldv2_unsolicited_report_interval",
6797 		.data		=
6798 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6799 		.maxlen		= sizeof(int),
6800 		.mode		= 0644,
6801 		.proc_handler	= proc_dointvec_ms_jiffies,
6802 	},
6803 	{
6804 		.procname	= "use_tempaddr",
6805 		.data		= &ipv6_devconf.use_tempaddr,
6806 		.maxlen		= sizeof(int),
6807 		.mode		= 0644,
6808 		.proc_handler	= proc_dointvec,
6809 	},
6810 	{
6811 		.procname	= "temp_valid_lft",
6812 		.data		= &ipv6_devconf.temp_valid_lft,
6813 		.maxlen		= sizeof(int),
6814 		.mode		= 0644,
6815 		.proc_handler	= proc_dointvec,
6816 	},
6817 	{
6818 		.procname	= "temp_prefered_lft",
6819 		.data		= &ipv6_devconf.temp_prefered_lft,
6820 		.maxlen		= sizeof(int),
6821 		.mode		= 0644,
6822 		.proc_handler	= proc_dointvec,
6823 	},
6824 	{
6825 		.procname	= "regen_max_retry",
6826 		.data		= &ipv6_devconf.regen_max_retry,
6827 		.maxlen		= sizeof(int),
6828 		.mode		= 0644,
6829 		.proc_handler	= proc_dointvec,
6830 	},
6831 	{
6832 		.procname	= "max_desync_factor",
6833 		.data		= &ipv6_devconf.max_desync_factor,
6834 		.maxlen		= sizeof(int),
6835 		.mode		= 0644,
6836 		.proc_handler	= proc_dointvec,
6837 	},
6838 	{
6839 		.procname	= "max_addresses",
6840 		.data		= &ipv6_devconf.max_addresses,
6841 		.maxlen		= sizeof(int),
6842 		.mode		= 0644,
6843 		.proc_handler	= proc_dointvec,
6844 	},
6845 	{
6846 		.procname	= "accept_ra_defrtr",
6847 		.data		= &ipv6_devconf.accept_ra_defrtr,
6848 		.maxlen		= sizeof(int),
6849 		.mode		= 0644,
6850 		.proc_handler	= proc_dointvec,
6851 	},
6852 	{
6853 		.procname	= "ra_defrtr_metric",
6854 		.data		= &ipv6_devconf.ra_defrtr_metric,
6855 		.maxlen		= sizeof(u32),
6856 		.mode		= 0644,
6857 		.proc_handler	= proc_douintvec_minmax,
6858 		.extra1		= (void *)SYSCTL_ONE,
6859 	},
6860 	{
6861 		.procname	= "accept_ra_min_hop_limit",
6862 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6863 		.maxlen		= sizeof(int),
6864 		.mode		= 0644,
6865 		.proc_handler	= proc_dointvec,
6866 	},
6867 	{
6868 		.procname	= "accept_ra_min_lft",
6869 		.data		= &ipv6_devconf.accept_ra_min_lft,
6870 		.maxlen		= sizeof(int),
6871 		.mode		= 0644,
6872 		.proc_handler	= proc_dointvec,
6873 	},
6874 	{
6875 		.procname	= "accept_ra_pinfo",
6876 		.data		= &ipv6_devconf.accept_ra_pinfo,
6877 		.maxlen		= sizeof(int),
6878 		.mode		= 0644,
6879 		.proc_handler	= proc_dointvec,
6880 	},
6881 #ifdef CONFIG_IPV6_ROUTER_PREF
6882 	{
6883 		.procname	= "accept_ra_rtr_pref",
6884 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6885 		.maxlen		= sizeof(int),
6886 		.mode		= 0644,
6887 		.proc_handler	= proc_dointvec,
6888 	},
6889 	{
6890 		.procname	= "router_probe_interval",
6891 		.data		= &ipv6_devconf.rtr_probe_interval,
6892 		.maxlen		= sizeof(int),
6893 		.mode		= 0644,
6894 		.proc_handler	= proc_dointvec_jiffies,
6895 	},
6896 #ifdef CONFIG_IPV6_ROUTE_INFO
6897 	{
6898 		.procname	= "accept_ra_rt_info_min_plen",
6899 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6900 		.maxlen		= sizeof(int),
6901 		.mode		= 0644,
6902 		.proc_handler	= proc_dointvec,
6903 	},
6904 	{
6905 		.procname	= "accept_ra_rt_info_max_plen",
6906 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6907 		.maxlen		= sizeof(int),
6908 		.mode		= 0644,
6909 		.proc_handler	= proc_dointvec,
6910 	},
6911 #endif
6912 #endif
6913 	{
6914 		.procname	= "proxy_ndp",
6915 		.data		= &ipv6_devconf.proxy_ndp,
6916 		.maxlen		= sizeof(int),
6917 		.mode		= 0644,
6918 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6919 	},
6920 	{
6921 		.procname	= "accept_source_route",
6922 		.data		= &ipv6_devconf.accept_source_route,
6923 		.maxlen		= sizeof(int),
6924 		.mode		= 0644,
6925 		.proc_handler	= proc_dointvec,
6926 	},
6927 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6928 	{
6929 		.procname	= "optimistic_dad",
6930 		.data		= &ipv6_devconf.optimistic_dad,
6931 		.maxlen		= sizeof(int),
6932 		.mode		= 0644,
6933 		.proc_handler   = proc_dointvec,
6934 	},
6935 	{
6936 		.procname	= "use_optimistic",
6937 		.data		= &ipv6_devconf.use_optimistic,
6938 		.maxlen		= sizeof(int),
6939 		.mode		= 0644,
6940 		.proc_handler	= proc_dointvec,
6941 	},
6942 #endif
6943 #ifdef CONFIG_IPV6_MROUTE
6944 	{
6945 		.procname	= "mc_forwarding",
6946 		.data		= &ipv6_devconf.mc_forwarding,
6947 		.maxlen		= sizeof(int),
6948 		.mode		= 0444,
6949 		.proc_handler	= proc_dointvec,
6950 	},
6951 #endif
6952 	{
6953 		.procname	= "disable_ipv6",
6954 		.data		= &ipv6_devconf.disable_ipv6,
6955 		.maxlen		= sizeof(int),
6956 		.mode		= 0644,
6957 		.proc_handler	= addrconf_sysctl_disable,
6958 	},
6959 	{
6960 		.procname	= "accept_dad",
6961 		.data		= &ipv6_devconf.accept_dad,
6962 		.maxlen		= sizeof(int),
6963 		.mode		= 0644,
6964 		.proc_handler	= proc_dointvec,
6965 	},
6966 	{
6967 		.procname	= "force_tllao",
6968 		.data		= &ipv6_devconf.force_tllao,
6969 		.maxlen		= sizeof(int),
6970 		.mode		= 0644,
6971 		.proc_handler	= proc_dointvec
6972 	},
6973 	{
6974 		.procname	= "ndisc_notify",
6975 		.data		= &ipv6_devconf.ndisc_notify,
6976 		.maxlen		= sizeof(int),
6977 		.mode		= 0644,
6978 		.proc_handler	= proc_dointvec
6979 	},
6980 	{
6981 		.procname	= "suppress_frag_ndisc",
6982 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6983 		.maxlen		= sizeof(int),
6984 		.mode		= 0644,
6985 		.proc_handler	= proc_dointvec
6986 	},
6987 	{
6988 		.procname	= "accept_ra_from_local",
6989 		.data		= &ipv6_devconf.accept_ra_from_local,
6990 		.maxlen		= sizeof(int),
6991 		.mode		= 0644,
6992 		.proc_handler	= proc_dointvec,
6993 	},
6994 	{
6995 		.procname	= "accept_ra_mtu",
6996 		.data		= &ipv6_devconf.accept_ra_mtu,
6997 		.maxlen		= sizeof(int),
6998 		.mode		= 0644,
6999 		.proc_handler	= proc_dointvec,
7000 	},
7001 	{
7002 		.procname	= "stable_secret",
7003 		.data		= &ipv6_devconf.stable_secret,
7004 		.maxlen		= IPV6_MAX_STRLEN,
7005 		.mode		= 0600,
7006 		.proc_handler	= addrconf_sysctl_stable_secret,
7007 	},
7008 	{
7009 		.procname	= "use_oif_addrs_only",
7010 		.data		= &ipv6_devconf.use_oif_addrs_only,
7011 		.maxlen		= sizeof(int),
7012 		.mode		= 0644,
7013 		.proc_handler	= proc_dointvec,
7014 	},
7015 	{
7016 		.procname	= "ignore_routes_with_linkdown",
7017 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
7018 		.maxlen		= sizeof(int),
7019 		.mode		= 0644,
7020 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
7021 	},
7022 	{
7023 		.procname	= "drop_unicast_in_l2_multicast",
7024 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
7025 		.maxlen		= sizeof(int),
7026 		.mode		= 0644,
7027 		.proc_handler	= proc_dointvec,
7028 	},
7029 	{
7030 		.procname	= "drop_unsolicited_na",
7031 		.data		= &ipv6_devconf.drop_unsolicited_na,
7032 		.maxlen		= sizeof(int),
7033 		.mode		= 0644,
7034 		.proc_handler	= proc_dointvec,
7035 	},
7036 	{
7037 		.procname	= "keep_addr_on_down",
7038 		.data		= &ipv6_devconf.keep_addr_on_down,
7039 		.maxlen		= sizeof(int),
7040 		.mode		= 0644,
7041 		.proc_handler	= proc_dointvec,
7042 
7043 	},
7044 	{
7045 		.procname	= "seg6_enabled",
7046 		.data		= &ipv6_devconf.seg6_enabled,
7047 		.maxlen		= sizeof(int),
7048 		.mode		= 0644,
7049 		.proc_handler	= proc_dointvec,
7050 	},
7051 #ifdef CONFIG_IPV6_SEG6_HMAC
7052 	{
7053 		.procname	= "seg6_require_hmac",
7054 		.data		= &ipv6_devconf.seg6_require_hmac,
7055 		.maxlen		= sizeof(int),
7056 		.mode		= 0644,
7057 		.proc_handler	= proc_dointvec,
7058 	},
7059 #endif
7060 	{
7061 		.procname       = "enhanced_dad",
7062 		.data           = &ipv6_devconf.enhanced_dad,
7063 		.maxlen         = sizeof(int),
7064 		.mode           = 0644,
7065 		.proc_handler   = proc_dointvec,
7066 	},
7067 	{
7068 		.procname	= "addr_gen_mode",
7069 		.data		= &ipv6_devconf.addr_gen_mode,
7070 		.maxlen		= sizeof(int),
7071 		.mode		= 0644,
7072 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
7073 	},
7074 	{
7075 		.procname       = "disable_policy",
7076 		.data           = &ipv6_devconf.disable_policy,
7077 		.maxlen         = sizeof(int),
7078 		.mode           = 0644,
7079 		.proc_handler   = addrconf_sysctl_disable_policy,
7080 	},
7081 	{
7082 		.procname	= "ndisc_tclass",
7083 		.data		= &ipv6_devconf.ndisc_tclass,
7084 		.maxlen		= sizeof(int),
7085 		.mode		= 0644,
7086 		.proc_handler	= proc_dointvec_minmax,
7087 		.extra1		= (void *)SYSCTL_ZERO,
7088 		.extra2		= (void *)&two_five_five,
7089 	},
7090 	{
7091 		.procname	= "rpl_seg_enabled",
7092 		.data		= &ipv6_devconf.rpl_seg_enabled,
7093 		.maxlen		= sizeof(int),
7094 		.mode		= 0644,
7095 		.proc_handler	= proc_dointvec,
7096 	},
7097 	{
7098 		.procname	= "ioam6_enabled",
7099 		.data		= &ipv6_devconf.ioam6_enabled,
7100 		.maxlen		= sizeof(u8),
7101 		.mode		= 0644,
7102 		.proc_handler	= proc_dou8vec_minmax,
7103 		.extra1		= (void *)SYSCTL_ZERO,
7104 		.extra2		= (void *)SYSCTL_ONE,
7105 	},
7106 	{
7107 		.procname	= "ioam6_id",
7108 		.data		= &ipv6_devconf.ioam6_id,
7109 		.maxlen		= sizeof(u32),
7110 		.mode		= 0644,
7111 		.proc_handler	= proc_douintvec_minmax,
7112 		.extra1		= (void *)SYSCTL_ZERO,
7113 		.extra2		= (void *)&ioam6_if_id_max,
7114 	},
7115 	{
7116 		.procname	= "ioam6_id_wide",
7117 		.data		= &ipv6_devconf.ioam6_id_wide,
7118 		.maxlen		= sizeof(u32),
7119 		.mode		= 0644,
7120 		.proc_handler	= proc_douintvec,
7121 	},
7122 	{
7123 		.procname	= "ndisc_evict_nocarrier",
7124 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
7125 		.maxlen		= sizeof(u8),
7126 		.mode		= 0644,
7127 		.proc_handler	= proc_dou8vec_minmax,
7128 		.extra1		= (void *)SYSCTL_ZERO,
7129 		.extra2		= (void *)SYSCTL_ONE,
7130 	},
7131 	{
7132 		.procname	= "accept_untracked_na",
7133 		.data		= &ipv6_devconf.accept_untracked_na,
7134 		.maxlen		= sizeof(int),
7135 		.mode		= 0644,
7136 		.proc_handler	= proc_dointvec_minmax,
7137 		.extra1		= SYSCTL_ZERO,
7138 		.extra2		= SYSCTL_TWO,
7139 	},
7140 	{
7141 		/* sentinel */
7142 	}
7143 };
7144 
7145 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7146 		struct inet6_dev *idev, struct ipv6_devconf *p)
7147 {
7148 	int i, ifindex;
7149 	struct ctl_table *table;
7150 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7151 
7152 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7153 	if (!table)
7154 		goto out;
7155 
7156 	for (i = 0; table[i].data; i++) {
7157 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7158 		/* If one of these is already set, then it is not safe to
7159 		 * overwrite either of them: this makes proc_dointvec_minmax
7160 		 * usable.
7161 		 */
7162 		if (!table[i].extra1 && !table[i].extra2) {
7163 			table[i].extra1 = idev; /* embedded; no ref */
7164 			table[i].extra2 = net;
7165 		}
7166 	}
7167 
7168 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7169 
7170 	p->sysctl_header = register_net_sysctl_sz(net, path, table,
7171 						  ARRAY_SIZE(addrconf_sysctl));
7172 	if (!p->sysctl_header)
7173 		goto free;
7174 
7175 	if (!strcmp(dev_name, "all"))
7176 		ifindex = NETCONFA_IFINDEX_ALL;
7177 	else if (!strcmp(dev_name, "default"))
7178 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7179 	else
7180 		ifindex = idev->dev->ifindex;
7181 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7182 				     ifindex, p);
7183 	return 0;
7184 
7185 free:
7186 	kfree(table);
7187 out:
7188 	return -ENOBUFS;
7189 }
7190 
7191 static void __addrconf_sysctl_unregister(struct net *net,
7192 					 struct ipv6_devconf *p, int ifindex)
7193 {
7194 	struct ctl_table *table;
7195 
7196 	if (!p->sysctl_header)
7197 		return;
7198 
7199 	table = p->sysctl_header->ctl_table_arg;
7200 	unregister_net_sysctl_table(p->sysctl_header);
7201 	p->sysctl_header = NULL;
7202 	kfree(table);
7203 
7204 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7205 }
7206 
7207 static int addrconf_sysctl_register(struct inet6_dev *idev)
7208 {
7209 	int err;
7210 
7211 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7212 		return -EINVAL;
7213 
7214 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7215 				    &ndisc_ifinfo_sysctl_change);
7216 	if (err)
7217 		return err;
7218 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7219 					 idev, &idev->cnf);
7220 	if (err)
7221 		neigh_sysctl_unregister(idev->nd_parms);
7222 
7223 	return err;
7224 }
7225 
7226 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7227 {
7228 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7229 				     idev->dev->ifindex);
7230 	neigh_sysctl_unregister(idev->nd_parms);
7231 }
7232 
7233 
7234 #endif
7235 
7236 static int __net_init addrconf_init_net(struct net *net)
7237 {
7238 	int err = -ENOMEM;
7239 	struct ipv6_devconf *all, *dflt;
7240 
7241 	spin_lock_init(&net->ipv6.addrconf_hash_lock);
7242 	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7243 	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7244 					   sizeof(struct hlist_head),
7245 					   GFP_KERNEL);
7246 	if (!net->ipv6.inet6_addr_lst)
7247 		goto err_alloc_addr;
7248 
7249 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7250 	if (!all)
7251 		goto err_alloc_all;
7252 
7253 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7254 	if (!dflt)
7255 		goto err_alloc_dflt;
7256 
7257 	if (!net_eq(net, &init_net)) {
7258 		switch (net_inherit_devconf()) {
7259 		case 1:  /* copy from init_net */
7260 			memcpy(all, init_net.ipv6.devconf_all,
7261 			       sizeof(ipv6_devconf));
7262 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7263 			       sizeof(ipv6_devconf_dflt));
7264 			break;
7265 		case 3: /* copy from the current netns */
7266 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7267 			       sizeof(ipv6_devconf));
7268 			memcpy(dflt,
7269 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7270 			       sizeof(ipv6_devconf_dflt));
7271 			break;
7272 		case 0:
7273 		case 2:
7274 			/* use compiled values */
7275 			break;
7276 		}
7277 	}
7278 
7279 	/* these will be inherited by all namespaces */
7280 	dflt->autoconf = ipv6_defaults.autoconf;
7281 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7282 
7283 	dflt->stable_secret.initialized = false;
7284 	all->stable_secret.initialized = false;
7285 
7286 	net->ipv6.devconf_all = all;
7287 	net->ipv6.devconf_dflt = dflt;
7288 
7289 #ifdef CONFIG_SYSCTL
7290 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7291 	if (err < 0)
7292 		goto err_reg_all;
7293 
7294 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7295 	if (err < 0)
7296 		goto err_reg_dflt;
7297 #endif
7298 	return 0;
7299 
7300 #ifdef CONFIG_SYSCTL
7301 err_reg_dflt:
7302 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7303 err_reg_all:
7304 	kfree(dflt);
7305 	net->ipv6.devconf_dflt = NULL;
7306 #endif
7307 err_alloc_dflt:
7308 	kfree(all);
7309 	net->ipv6.devconf_all = NULL;
7310 err_alloc_all:
7311 	kfree(net->ipv6.inet6_addr_lst);
7312 err_alloc_addr:
7313 	return err;
7314 }
7315 
7316 static void __net_exit addrconf_exit_net(struct net *net)
7317 {
7318 	int i;
7319 
7320 #ifdef CONFIG_SYSCTL
7321 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7322 				     NETCONFA_IFINDEX_DEFAULT);
7323 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7324 				     NETCONFA_IFINDEX_ALL);
7325 #endif
7326 	kfree(net->ipv6.devconf_dflt);
7327 	net->ipv6.devconf_dflt = NULL;
7328 	kfree(net->ipv6.devconf_all);
7329 	net->ipv6.devconf_all = NULL;
7330 
7331 	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7332 	/*
7333 	 *	Check hash table, then free it.
7334 	 */
7335 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7336 		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7337 
7338 	kfree(net->ipv6.inet6_addr_lst);
7339 	net->ipv6.inet6_addr_lst = NULL;
7340 }
7341 
7342 static struct pernet_operations addrconf_ops = {
7343 	.init = addrconf_init_net,
7344 	.exit = addrconf_exit_net,
7345 };
7346 
7347 static struct rtnl_af_ops inet6_ops __read_mostly = {
7348 	.family		  = AF_INET6,
7349 	.fill_link_af	  = inet6_fill_link_af,
7350 	.get_link_af_size = inet6_get_link_af_size,
7351 	.validate_link_af = inet6_validate_link_af,
7352 	.set_link_af	  = inet6_set_link_af,
7353 };
7354 
7355 /*
7356  *	Init / cleanup code
7357  */
7358 
7359 int __init addrconf_init(void)
7360 {
7361 	struct inet6_dev *idev;
7362 	int err;
7363 
7364 	err = ipv6_addr_label_init();
7365 	if (err < 0) {
7366 		pr_crit("%s: cannot initialize default policy table: %d\n",
7367 			__func__, err);
7368 		goto out;
7369 	}
7370 
7371 	err = register_pernet_subsys(&addrconf_ops);
7372 	if (err < 0)
7373 		goto out_addrlabel;
7374 
7375 	addrconf_wq = create_workqueue("ipv6_addrconf");
7376 	if (!addrconf_wq) {
7377 		err = -ENOMEM;
7378 		goto out_nowq;
7379 	}
7380 
7381 	rtnl_lock();
7382 	idev = ipv6_add_dev(blackhole_netdev);
7383 	rtnl_unlock();
7384 	if (IS_ERR(idev)) {
7385 		err = PTR_ERR(idev);
7386 		goto errlo;
7387 	}
7388 
7389 	ip6_route_init_special_entries();
7390 
7391 	register_netdevice_notifier(&ipv6_dev_notf);
7392 
7393 	addrconf_verify(&init_net);
7394 
7395 	rtnl_af_register(&inet6_ops);
7396 
7397 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7398 				   NULL, inet6_dump_ifinfo, 0);
7399 	if (err < 0)
7400 		goto errout;
7401 
7402 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7403 				   inet6_rtm_newaddr, NULL, 0);
7404 	if (err < 0)
7405 		goto errout;
7406 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7407 				   inet6_rtm_deladdr, NULL, 0);
7408 	if (err < 0)
7409 		goto errout;
7410 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7411 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7412 				   RTNL_FLAG_DOIT_UNLOCKED);
7413 	if (err < 0)
7414 		goto errout;
7415 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7416 				   NULL, inet6_dump_ifmcaddr, 0);
7417 	if (err < 0)
7418 		goto errout;
7419 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7420 				   NULL, inet6_dump_ifacaddr, 0);
7421 	if (err < 0)
7422 		goto errout;
7423 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7424 				   inet6_netconf_get_devconf,
7425 				   inet6_netconf_dump_devconf,
7426 				   RTNL_FLAG_DOIT_UNLOCKED);
7427 	if (err < 0)
7428 		goto errout;
7429 	err = ipv6_addr_label_rtnl_register();
7430 	if (err < 0)
7431 		goto errout;
7432 
7433 	return 0;
7434 errout:
7435 	rtnl_unregister_all(PF_INET6);
7436 	rtnl_af_unregister(&inet6_ops);
7437 	unregister_netdevice_notifier(&ipv6_dev_notf);
7438 errlo:
7439 	destroy_workqueue(addrconf_wq);
7440 out_nowq:
7441 	unregister_pernet_subsys(&addrconf_ops);
7442 out_addrlabel:
7443 	ipv6_addr_label_cleanup();
7444 out:
7445 	return err;
7446 }
7447 
7448 void addrconf_cleanup(void)
7449 {
7450 	struct net_device *dev;
7451 
7452 	unregister_netdevice_notifier(&ipv6_dev_notf);
7453 	unregister_pernet_subsys(&addrconf_ops);
7454 	ipv6_addr_label_cleanup();
7455 
7456 	rtnl_af_unregister(&inet6_ops);
7457 
7458 	rtnl_lock();
7459 
7460 	/* clean dev list */
7461 	for_each_netdev(&init_net, dev) {
7462 		if (__in6_dev_get(dev) == NULL)
7463 			continue;
7464 		addrconf_ifdown(dev, true);
7465 	}
7466 	addrconf_ifdown(init_net.loopback_dev, true);
7467 
7468 	rtnl_unlock();
7469 
7470 	destroy_workqueue(addrconf_wq);
7471 }
7472