xref: /linux/net/ipv6/addrconf.c (revision af2d6148d2a159e1a0862bce5a2c88c1618a2b27)
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/ip_tunnels.h>
67 #include <net/net_namespace.h>
68 #include <net/sock.h>
69 #include <net/snmp.h>
70 
71 #include <net/6lowpan.h>
72 #include <net/firewire.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <net/l3mdev.h>
83 #include <net/netdev_lock.h>
84 #include <linux/if_tunnel.h>
85 #include <linux/rtnetlink.h>
86 #include <linux/netconf.h>
87 #include <linux/random.h>
88 #include <linux/uaccess.h>
89 #include <linux/unaligned.h>
90 
91 #include <linux/proc_fs.h>
92 #include <linux/seq_file.h>
93 #include <linux/export.h>
94 #include <linux/ioam6.h>
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_min_advance	= REGEN_MIN_ADVANCE,
199 	.regen_max_retry	= REGEN_MAX_RETRY,
200 	.max_desync_factor	= MAX_DESYNC_FACTOR,
201 	.max_addresses		= IPV6_MAX_ADDRESSES,
202 	.accept_ra_defrtr	= 1,
203 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
204 	.accept_ra_from_local	= 0,
205 	.accept_ra_min_hop_limit= 1,
206 	.accept_ra_min_lft	= 0,
207 	.accept_ra_pinfo	= 1,
208 #ifdef CONFIG_IPV6_ROUTER_PREF
209 	.accept_ra_rtr_pref	= 1,
210 	.rtr_probe_interval	= 60 * HZ,
211 #ifdef CONFIG_IPV6_ROUTE_INFO
212 	.accept_ra_rt_info_min_plen = 0,
213 	.accept_ra_rt_info_max_plen = 0,
214 #endif
215 #endif
216 	.proxy_ndp		= 0,
217 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
218 	.disable_ipv6		= 0,
219 	.accept_dad		= 0,
220 	.suppress_frag_ndisc	= 1,
221 	.accept_ra_mtu		= 1,
222 	.stable_secret		= {
223 		.initialized = false,
224 	},
225 	.use_oif_addrs_only	= 0,
226 	.ignore_routes_with_linkdown = 0,
227 	.keep_addr_on_down	= 0,
228 	.seg6_enabled		= 0,
229 #ifdef CONFIG_IPV6_SEG6_HMAC
230 	.seg6_require_hmac	= 0,
231 #endif
232 	.enhanced_dad           = 1,
233 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
234 	.disable_policy		= 0,
235 	.rpl_seg_enabled	= 0,
236 	.ioam6_enabled		= 0,
237 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
238 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
239 	.ndisc_evict_nocarrier	= 1,
240 	.ra_honor_pio_life	= 0,
241 	.ra_honor_pio_pflag	= 0,
242 };
243 
244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
245 	.forwarding		= 0,
246 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
247 	.mtu6			= IPV6_MIN_MTU,
248 	.accept_ra		= 1,
249 	.accept_redirects	= 1,
250 	.autoconf		= 1,
251 	.force_mld_version	= 0,
252 	.mldv1_unsolicited_report_interval = 10 * HZ,
253 	.mldv2_unsolicited_report_interval = HZ,
254 	.dad_transmits		= 1,
255 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
256 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
257 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
258 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
259 	.use_tempaddr		= 0,
260 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
261 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
262 	.regen_min_advance	= REGEN_MIN_ADVANCE,
263 	.regen_max_retry	= REGEN_MAX_RETRY,
264 	.max_desync_factor	= MAX_DESYNC_FACTOR,
265 	.max_addresses		= IPV6_MAX_ADDRESSES,
266 	.accept_ra_defrtr	= 1,
267 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
268 	.accept_ra_from_local	= 0,
269 	.accept_ra_min_hop_limit= 1,
270 	.accept_ra_min_lft	= 0,
271 	.accept_ra_pinfo	= 1,
272 #ifdef CONFIG_IPV6_ROUTER_PREF
273 	.accept_ra_rtr_pref	= 1,
274 	.rtr_probe_interval	= 60 * HZ,
275 #ifdef CONFIG_IPV6_ROUTE_INFO
276 	.accept_ra_rt_info_min_plen = 0,
277 	.accept_ra_rt_info_max_plen = 0,
278 #endif
279 #endif
280 	.proxy_ndp		= 0,
281 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
282 	.disable_ipv6		= 0,
283 	.accept_dad		= 1,
284 	.suppress_frag_ndisc	= 1,
285 	.accept_ra_mtu		= 1,
286 	.stable_secret		= {
287 		.initialized = false,
288 	},
289 	.use_oif_addrs_only	= 0,
290 	.ignore_routes_with_linkdown = 0,
291 	.keep_addr_on_down	= 0,
292 	.seg6_enabled		= 0,
293 #ifdef CONFIG_IPV6_SEG6_HMAC
294 	.seg6_require_hmac	= 0,
295 #endif
296 	.enhanced_dad           = 1,
297 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
298 	.disable_policy		= 0,
299 	.rpl_seg_enabled	= 0,
300 	.ioam6_enabled		= 0,
301 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
302 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
303 	.ndisc_evict_nocarrier	= 1,
304 	.ra_honor_pio_life	= 0,
305 	.ra_honor_pio_pflag	= 0,
306 };
307 
308 /* Check if link is ready: is it up and is a valid qdisc available */
309 static inline bool addrconf_link_ready(const struct net_device *dev)
310 {
311 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
312 }
313 
314 static void addrconf_del_rs_timer(struct inet6_dev *idev)
315 {
316 	if (timer_delete(&idev->rs_timer))
317 		__in6_dev_put(idev);
318 }
319 
320 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
321 {
322 	if (cancel_delayed_work(&ifp->dad_work))
323 		__in6_ifa_put(ifp);
324 }
325 
326 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
327 				  unsigned long when)
328 {
329 	if (!mod_timer(&idev->rs_timer, jiffies + when))
330 		in6_dev_hold(idev);
331 }
332 
333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
334 				   unsigned long delay)
335 {
336 	in6_ifa_hold(ifp);
337 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
338 		in6_ifa_put(ifp);
339 }
340 
341 static int snmp6_alloc_dev(struct inet6_dev *idev)
342 {
343 	int i;
344 
345 	idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
346 	if (!idev->stats.ipv6)
347 		goto err_ip;
348 
349 	for_each_possible_cpu(i) {
350 		struct ipstats_mib *addrconf_stats;
351 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
352 		u64_stats_init(&addrconf_stats->syncp);
353 	}
354 
355 
356 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
357 					GFP_KERNEL);
358 	if (!idev->stats.icmpv6dev)
359 		goto err_icmp;
360 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
361 					   GFP_KERNEL_ACCOUNT);
362 	if (!idev->stats.icmpv6msgdev)
363 		goto err_icmpmsg;
364 
365 	return 0;
366 
367 err_icmpmsg:
368 	kfree(idev->stats.icmpv6dev);
369 err_icmp:
370 	free_percpu(idev->stats.ipv6);
371 err_ip:
372 	return -ENOMEM;
373 }
374 
375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
376 {
377 	struct inet6_dev *ndev;
378 	int err = -ENOMEM;
379 
380 	ASSERT_RTNL();
381 	netdev_ops_assert_locked(dev);
382 
383 	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
384 		return ERR_PTR(-EINVAL);
385 
386 	ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
387 	if (!ndev)
388 		return ERR_PTR(err);
389 
390 	rwlock_init(&ndev->lock);
391 	ndev->dev = dev;
392 	INIT_LIST_HEAD(&ndev->addr_list);
393 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
394 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
395 
396 	if (ndev->cnf.stable_secret.initialized)
397 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
398 
399 	ndev->cnf.mtu6 = dev->mtu;
400 	ndev->ra_mtu = 0;
401 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
402 	if (!ndev->nd_parms) {
403 		kfree(ndev);
404 		return ERR_PTR(err);
405 	}
406 	if (ndev->cnf.forwarding)
407 		netif_disable_lro(dev);
408 	/* We refer to the device */
409 	netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
410 
411 	if (snmp6_alloc_dev(ndev) < 0) {
412 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
413 			   __func__);
414 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
415 		netdev_put(dev, &ndev->dev_tracker);
416 		kfree(ndev);
417 		return ERR_PTR(err);
418 	}
419 
420 	if (dev != blackhole_netdev) {
421 		if (snmp6_register_dev(ndev) < 0) {
422 			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
423 				   __func__, dev->name);
424 			goto err_release;
425 		}
426 	}
427 	/* One reference from device. */
428 	refcount_set(&ndev->refcnt, 1);
429 
430 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
431 		ndev->cnf.accept_dad = -1;
432 
433 #if IS_ENABLED(CONFIG_IPV6_SIT)
434 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
435 		pr_info("%s: Disabled Multicast RS\n", dev->name);
436 		ndev->cnf.rtr_solicits = 0;
437 	}
438 #endif
439 
440 	INIT_LIST_HEAD(&ndev->tempaddr_list);
441 	ndev->desync_factor = U32_MAX;
442 	if ((dev->flags&IFF_LOOPBACK) ||
443 	    dev->type == ARPHRD_TUNNEL ||
444 	    dev->type == ARPHRD_TUNNEL6 ||
445 	    dev->type == ARPHRD_SIT ||
446 	    dev->type == ARPHRD_NONE) {
447 		ndev->cnf.use_tempaddr = -1;
448 	}
449 
450 	ndev->token = in6addr_any;
451 
452 	if (netif_running(dev) && addrconf_link_ready(dev))
453 		ndev->if_flags |= IF_READY;
454 
455 	ipv6_mc_init_dev(ndev);
456 	ndev->tstamp = jiffies;
457 	if (dev != blackhole_netdev) {
458 		err = addrconf_sysctl_register(ndev);
459 		if (err) {
460 			ipv6_mc_destroy_dev(ndev);
461 			snmp6_unregister_dev(ndev);
462 			goto err_release;
463 		}
464 	}
465 	/* protected by rtnl_lock */
466 	rcu_assign_pointer(dev->ip6_ptr, ndev);
467 
468 	if (dev != blackhole_netdev) {
469 		/* Join interface-local all-node multicast group */
470 		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
471 
472 		/* Join all-node multicast group */
473 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
474 
475 		/* Join all-router multicast group if forwarding is set */
476 		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
477 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
478 	}
479 	return ndev;
480 
481 err_release:
482 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
483 	ndev->dead = 1;
484 	in6_dev_finish_destroy(ndev);
485 	return ERR_PTR(err);
486 }
487 
488 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
489 {
490 	struct inet6_dev *idev;
491 
492 	ASSERT_RTNL();
493 
494 	idev = __in6_dev_get(dev);
495 	if (!idev) {
496 		idev = ipv6_add_dev(dev);
497 		if (IS_ERR(idev))
498 			return idev;
499 	}
500 
501 	if (dev->flags&IFF_UP)
502 		ipv6_mc_up(idev);
503 	return idev;
504 }
505 
506 static int inet6_netconf_msgsize_devconf(int type)
507 {
508 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
509 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
510 	bool all = false;
511 
512 	if (type == NETCONFA_ALL)
513 		all = true;
514 
515 	if (all || type == NETCONFA_FORWARDING)
516 		size += nla_total_size(4);
517 #ifdef CONFIG_IPV6_MROUTE
518 	if (all || type == NETCONFA_MC_FORWARDING)
519 		size += nla_total_size(4);
520 #endif
521 	if (all || type == NETCONFA_PROXY_NEIGH)
522 		size += nla_total_size(4);
523 
524 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
525 		size += nla_total_size(4);
526 
527 	return size;
528 }
529 
530 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
531 				      struct ipv6_devconf *devconf, u32 portid,
532 				      u32 seq, int event, unsigned int flags,
533 				      int type)
534 {
535 	struct nlmsghdr  *nlh;
536 	struct netconfmsg *ncm;
537 	bool all = false;
538 
539 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
540 			flags);
541 	if (!nlh)
542 		return -EMSGSIZE;
543 
544 	if (type == NETCONFA_ALL)
545 		all = true;
546 
547 	ncm = nlmsg_data(nlh);
548 	ncm->ncm_family = AF_INET6;
549 
550 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
551 		goto nla_put_failure;
552 
553 	if (!devconf)
554 		goto out;
555 
556 	if ((all || type == NETCONFA_FORWARDING) &&
557 	    nla_put_s32(skb, NETCONFA_FORWARDING,
558 			READ_ONCE(devconf->forwarding)) < 0)
559 		goto nla_put_failure;
560 #ifdef CONFIG_IPV6_MROUTE
561 	if ((all || type == NETCONFA_MC_FORWARDING) &&
562 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
563 			atomic_read(&devconf->mc_forwarding)) < 0)
564 		goto nla_put_failure;
565 #endif
566 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
567 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
568 			READ_ONCE(devconf->proxy_ndp)) < 0)
569 		goto nla_put_failure;
570 
571 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
572 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
573 			READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
574 		goto nla_put_failure;
575 
576 out:
577 	nlmsg_end(skb, nlh);
578 	return 0;
579 
580 nla_put_failure:
581 	nlmsg_cancel(skb, nlh);
582 	return -EMSGSIZE;
583 }
584 
585 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
586 				  int ifindex, struct ipv6_devconf *devconf)
587 {
588 	struct sk_buff *skb;
589 	int err = -ENOBUFS;
590 
591 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
592 	if (!skb)
593 		goto errout;
594 
595 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
596 					 event, 0, type);
597 	if (err < 0) {
598 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
599 		WARN_ON(err == -EMSGSIZE);
600 		kfree_skb(skb);
601 		goto errout;
602 	}
603 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
604 	return;
605 errout:
606 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
607 }
608 
609 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
610 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
611 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
612 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
613 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
614 };
615 
616 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
617 				       const struct nlmsghdr *nlh,
618 				       struct nlattr **tb,
619 				       struct netlink_ext_ack *extack)
620 {
621 	int i, err;
622 
623 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
624 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
625 		return -EINVAL;
626 	}
627 
628 	if (!netlink_strict_get_check(skb))
629 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
630 					      tb, NETCONFA_MAX,
631 					      devconf_ipv6_policy, extack);
632 
633 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
634 					    tb, NETCONFA_MAX,
635 					    devconf_ipv6_policy, extack);
636 	if (err)
637 		return err;
638 
639 	for (i = 0; i <= NETCONFA_MAX; i++) {
640 		if (!tb[i])
641 			continue;
642 
643 		switch (i) {
644 		case NETCONFA_IFINDEX:
645 			break;
646 		default:
647 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
648 			return -EINVAL;
649 		}
650 	}
651 
652 	return 0;
653 }
654 
655 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
656 				     struct nlmsghdr *nlh,
657 				     struct netlink_ext_ack *extack)
658 {
659 	struct net *net = sock_net(in_skb->sk);
660 	struct nlattr *tb[NETCONFA_MAX+1];
661 	struct inet6_dev *in6_dev = NULL;
662 	struct net_device *dev = NULL;
663 	struct sk_buff *skb;
664 	struct ipv6_devconf *devconf;
665 	int ifindex;
666 	int err;
667 
668 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
669 	if (err < 0)
670 		return err;
671 
672 	if (!tb[NETCONFA_IFINDEX])
673 		return -EINVAL;
674 
675 	err = -EINVAL;
676 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
677 	switch (ifindex) {
678 	case NETCONFA_IFINDEX_ALL:
679 		devconf = net->ipv6.devconf_all;
680 		break;
681 	case NETCONFA_IFINDEX_DEFAULT:
682 		devconf = net->ipv6.devconf_dflt;
683 		break;
684 	default:
685 		dev = dev_get_by_index(net, ifindex);
686 		if (!dev)
687 			return -EINVAL;
688 		in6_dev = in6_dev_get(dev);
689 		if (!in6_dev)
690 			goto errout;
691 		devconf = &in6_dev->cnf;
692 		break;
693 	}
694 
695 	err = -ENOBUFS;
696 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
697 	if (!skb)
698 		goto errout;
699 
700 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
701 					 NETLINK_CB(in_skb).portid,
702 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
703 					 NETCONFA_ALL);
704 	if (err < 0) {
705 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
706 		WARN_ON(err == -EMSGSIZE);
707 		kfree_skb(skb);
708 		goto errout;
709 	}
710 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
711 errout:
712 	if (in6_dev)
713 		in6_dev_put(in6_dev);
714 	dev_put(dev);
715 	return err;
716 }
717 
718 /* Combine dev_addr_genid and dev_base_seq to detect changes.
719  */
720 static u32 inet6_base_seq(const struct net *net)
721 {
722 	u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
723 		  READ_ONCE(net->dev_base_seq);
724 
725 	/* Must not return 0 (see nl_dump_check_consistent()).
726 	 * Chose a value far away from 0.
727 	 */
728 	if (!res)
729 		res = 0x80000000;
730 	return res;
731 }
732 
733 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
734 				      struct netlink_callback *cb)
735 {
736 	const struct nlmsghdr *nlh = cb->nlh;
737 	struct net *net = sock_net(skb->sk);
738 	struct {
739 		unsigned long ifindex;
740 		unsigned int all_default;
741 	} *ctx = (void *)cb->ctx;
742 	struct net_device *dev;
743 	struct inet6_dev *idev;
744 	int err = 0;
745 
746 	if (cb->strict_check) {
747 		struct netlink_ext_ack *extack = cb->extack;
748 		struct netconfmsg *ncm;
749 
750 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
751 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
752 			return -EINVAL;
753 		}
754 
755 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
756 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
757 			return -EINVAL;
758 		}
759 	}
760 
761 	rcu_read_lock();
762 	for_each_netdev_dump(net, dev, ctx->ifindex) {
763 		idev = __in6_dev_get(dev);
764 		if (!idev)
765 			continue;
766 		err = 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);
773 		if (err < 0)
774 			goto done;
775 	}
776 	if (ctx->all_default == 0) {
777 		err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
778 						 net->ipv6.devconf_all,
779 						 NETLINK_CB(cb->skb).portid,
780 						 nlh->nlmsg_seq,
781 						 RTM_NEWNETCONF, NLM_F_MULTI,
782 						 NETCONFA_ALL);
783 		if (err < 0)
784 			goto done;
785 		ctx->all_default++;
786 	}
787 	if (ctx->all_default == 1) {
788 		err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
789 						 net->ipv6.devconf_dflt,
790 						 NETLINK_CB(cb->skb).portid,
791 						 nlh->nlmsg_seq,
792 						 RTM_NEWNETCONF, NLM_F_MULTI,
793 						 NETCONFA_ALL);
794 		if (err < 0)
795 			goto done;
796 		ctx->all_default++;
797 	}
798 done:
799 	rcu_read_unlock();
800 	return err;
801 }
802 
803 #ifdef CONFIG_SYSCTL
804 static void dev_forward_change(struct inet6_dev *idev)
805 {
806 	struct net_device *dev;
807 	struct inet6_ifaddr *ifa;
808 	LIST_HEAD(tmp_addr_list);
809 
810 	if (!idev)
811 		return;
812 	dev = idev->dev;
813 	if (idev->cnf.forwarding)
814 		dev_disable_lro(dev);
815 	if (dev->flags & IFF_MULTICAST) {
816 		if (idev->cnf.forwarding) {
817 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
818 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
819 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
820 		} else {
821 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
822 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
823 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
824 		}
825 	}
826 
827 	read_lock_bh(&idev->lock);
828 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
829 		if (ifa->flags&IFA_F_TENTATIVE)
830 			continue;
831 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
832 	}
833 	read_unlock_bh(&idev->lock);
834 
835 	while (!list_empty(&tmp_addr_list)) {
836 		ifa = list_first_entry(&tmp_addr_list,
837 				       struct inet6_ifaddr, if_list_aux);
838 		list_del(&ifa->if_list_aux);
839 		if (idev->cnf.forwarding)
840 			addrconf_join_anycast(ifa);
841 		else
842 			addrconf_leave_anycast(ifa);
843 	}
844 
845 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
846 				     NETCONFA_FORWARDING,
847 				     dev->ifindex, &idev->cnf);
848 }
849 
850 
851 static void addrconf_forward_change(struct net *net, __s32 newf)
852 {
853 	struct net_device *dev;
854 	struct inet6_dev *idev;
855 
856 	for_each_netdev(net, dev) {
857 		idev = __in6_dev_get_rtnl_net(dev);
858 		if (idev) {
859 			int changed = (!idev->cnf.forwarding) ^ (!newf);
860 
861 			WRITE_ONCE(idev->cnf.forwarding, newf);
862 			if (changed)
863 				dev_forward_change(idev);
864 		}
865 	}
866 }
867 
868 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf)
869 {
870 	struct net *net = (struct net *)table->extra2;
871 	int old;
872 
873 	if (!rtnl_net_trylock(net))
874 		return restart_syscall();
875 
876 	old = *p;
877 	WRITE_ONCE(*p, newf);
878 
879 	if (p == &net->ipv6.devconf_dflt->forwarding) {
880 		if ((!newf) ^ (!old))
881 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
882 						     NETCONFA_FORWARDING,
883 						     NETCONFA_IFINDEX_DEFAULT,
884 						     net->ipv6.devconf_dflt);
885 		rtnl_net_unlock(net);
886 		return 0;
887 	}
888 
889 	if (p == &net->ipv6.devconf_all->forwarding) {
890 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
891 
892 		WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
893 		if ((!newf) ^ (!old_dflt))
894 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
895 						     NETCONFA_FORWARDING,
896 						     NETCONFA_IFINDEX_DEFAULT,
897 						     net->ipv6.devconf_dflt);
898 
899 		addrconf_forward_change(net, newf);
900 		if ((!newf) ^ (!old))
901 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
902 						     NETCONFA_FORWARDING,
903 						     NETCONFA_IFINDEX_ALL,
904 						     net->ipv6.devconf_all);
905 	} else if ((!newf) ^ (!old))
906 		dev_forward_change((struct inet6_dev *)table->extra1);
907 	rtnl_net_unlock(net);
908 
909 	if (newf)
910 		rt6_purge_dflt_routers(net);
911 	return 1;
912 }
913 
914 static void addrconf_linkdown_change(struct net *net, __s32 newf)
915 {
916 	struct net_device *dev;
917 	struct inet6_dev *idev;
918 
919 	for_each_netdev(net, dev) {
920 		idev = __in6_dev_get_rtnl_net(dev);
921 		if (idev) {
922 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
923 
924 			WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
925 			if (changed)
926 				inet6_netconf_notify_devconf(dev_net(dev),
927 							     RTM_NEWNETCONF,
928 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
929 							     dev->ifindex,
930 							     &idev->cnf);
931 		}
932 	}
933 }
934 
935 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf)
936 {
937 	struct net *net = (struct net *)table->extra2;
938 	int old;
939 
940 	if (!rtnl_net_trylock(net))
941 		return restart_syscall();
942 
943 	old = *p;
944 	WRITE_ONCE(*p, newf);
945 
946 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
947 		if ((!newf) ^ (!old))
948 			inet6_netconf_notify_devconf(net,
949 						     RTM_NEWNETCONF,
950 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
951 						     NETCONFA_IFINDEX_DEFAULT,
952 						     net->ipv6.devconf_dflt);
953 		rtnl_net_unlock(net);
954 		return 0;
955 	}
956 
957 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
958 		WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
959 		addrconf_linkdown_change(net, newf);
960 		if ((!newf) ^ (!old))
961 			inet6_netconf_notify_devconf(net,
962 						     RTM_NEWNETCONF,
963 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
964 						     NETCONFA_IFINDEX_ALL,
965 						     net->ipv6.devconf_all);
966 	}
967 
968 	rtnl_net_unlock(net);
969 
970 	return 1;
971 }
972 
973 #endif
974 
975 /* Nobody refers to this ifaddr, destroy it */
976 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
977 {
978 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
979 
980 #ifdef NET_REFCNT_DEBUG
981 	pr_debug("%s\n", __func__);
982 #endif
983 
984 	in6_dev_put(ifp->idev);
985 
986 	if (cancel_delayed_work(&ifp->dad_work))
987 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
988 			  ifp);
989 
990 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
991 		pr_warn("Freeing alive inet6 address %p\n", ifp);
992 		return;
993 	}
994 
995 	kfree_rcu(ifp, rcu);
996 }
997 
998 static void
999 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1000 {
1001 	struct list_head *p;
1002 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1003 
1004 	/*
1005 	 * Each device address list is sorted in order of scope -
1006 	 * global before linklocal.
1007 	 */
1008 	list_for_each(p, &idev->addr_list) {
1009 		struct inet6_ifaddr *ifa
1010 			= list_entry(p, struct inet6_ifaddr, if_list);
1011 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1012 			break;
1013 	}
1014 
1015 	list_add_tail_rcu(&ifp->if_list, p);
1016 }
1017 
1018 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1019 {
1020 	u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net));
1021 
1022 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1023 }
1024 
1025 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1026 			       struct net_device *dev, unsigned int hash)
1027 {
1028 	struct inet6_ifaddr *ifp;
1029 
1030 	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1031 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1032 			if (!dev || ifp->idev->dev == dev)
1033 				return true;
1034 		}
1035 	}
1036 	return false;
1037 }
1038 
1039 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1040 {
1041 	struct net *net = dev_net(dev);
1042 	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1043 	int err = 0;
1044 
1045 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1046 
1047 	/* Ignore adding duplicate addresses on an interface */
1048 	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1049 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1050 		err = -EEXIST;
1051 	} else {
1052 		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1053 	}
1054 
1055 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1056 
1057 	return err;
1058 }
1059 
1060 /* On success it returns ifp with increased reference count */
1061 
1062 static struct inet6_ifaddr *
1063 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1064 	      bool can_block, struct netlink_ext_ack *extack)
1065 {
1066 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1067 	int addr_type = ipv6_addr_type(cfg->pfx);
1068 	struct net *net = dev_net(idev->dev);
1069 	struct inet6_ifaddr *ifa = NULL;
1070 	struct fib6_info *f6i = NULL;
1071 	int err = 0;
1072 
1073 	if (addr_type == IPV6_ADDR_ANY) {
1074 		NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1075 		return ERR_PTR(-EADDRNOTAVAIL);
1076 	} else if (addr_type & IPV6_ADDR_MULTICAST &&
1077 		   !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1078 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1079 		return ERR_PTR(-EADDRNOTAVAIL);
1080 	} else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1081 		   !netif_is_l3_master(idev->dev) &&
1082 		   addr_type & IPV6_ADDR_LOOPBACK) {
1083 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1084 		return ERR_PTR(-EADDRNOTAVAIL);
1085 	}
1086 
1087 	if (idev->dead) {
1088 		NL_SET_ERR_MSG_MOD(extack, "device is going away");
1089 		err = -ENODEV;
1090 		goto out;
1091 	}
1092 
1093 	if (idev->cnf.disable_ipv6) {
1094 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1095 		err = -EACCES;
1096 		goto out;
1097 	}
1098 
1099 	/* validator notifier needs to be blocking;
1100 	 * do not call in atomic context
1101 	 */
1102 	if (can_block) {
1103 		struct in6_validator_info i6vi = {
1104 			.i6vi_addr = *cfg->pfx,
1105 			.i6vi_dev = idev,
1106 			.extack = extack,
1107 		};
1108 
1109 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1110 		err = notifier_to_errno(err);
1111 		if (err < 0)
1112 			goto out;
1113 	}
1114 
1115 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1116 	if (!ifa) {
1117 		err = -ENOBUFS;
1118 		goto out;
1119 	}
1120 
1121 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1122 	if (IS_ERR(f6i)) {
1123 		err = PTR_ERR(f6i);
1124 		f6i = NULL;
1125 		goto out;
1126 	}
1127 
1128 	neigh_parms_data_state_setall(idev->nd_parms);
1129 
1130 	ifa->addr = *cfg->pfx;
1131 	if (cfg->peer_pfx)
1132 		ifa->peer_addr = *cfg->peer_pfx;
1133 
1134 	spin_lock_init(&ifa->lock);
1135 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1136 	INIT_HLIST_NODE(&ifa->addr_lst);
1137 	ifa->scope = cfg->scope;
1138 	ifa->prefix_len = cfg->plen;
1139 	ifa->rt_priority = cfg->rt_priority;
1140 	ifa->flags = cfg->ifa_flags;
1141 	ifa->ifa_proto = cfg->ifa_proto;
1142 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1143 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1144 		ifa->flags |= IFA_F_TENTATIVE;
1145 	ifa->valid_lft = cfg->valid_lft;
1146 	ifa->prefered_lft = cfg->preferred_lft;
1147 	ifa->cstamp = ifa->tstamp = jiffies;
1148 	ifa->tokenized = false;
1149 
1150 	ifa->rt = f6i;
1151 
1152 	ifa->idev = idev;
1153 	in6_dev_hold(idev);
1154 
1155 	/* For caller */
1156 	refcount_set(&ifa->refcnt, 1);
1157 
1158 	rcu_read_lock();
1159 
1160 	err = ipv6_add_addr_hash(idev->dev, ifa);
1161 	if (err < 0) {
1162 		rcu_read_unlock();
1163 		goto out;
1164 	}
1165 
1166 	write_lock_bh(&idev->lock);
1167 
1168 	/* Add to inet6_dev unicast addr list. */
1169 	ipv6_link_dev_addr(idev, ifa);
1170 
1171 	if (ifa->flags&IFA_F_TEMPORARY) {
1172 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1173 		in6_ifa_hold(ifa);
1174 	}
1175 
1176 	in6_ifa_hold(ifa);
1177 	write_unlock_bh(&idev->lock);
1178 
1179 	rcu_read_unlock();
1180 
1181 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1182 out:
1183 	if (unlikely(err < 0)) {
1184 		fib6_info_release(f6i);
1185 
1186 		if (ifa) {
1187 			if (ifa->idev)
1188 				in6_dev_put(ifa->idev);
1189 			kfree(ifa);
1190 		}
1191 		ifa = ERR_PTR(err);
1192 	}
1193 
1194 	return ifa;
1195 }
1196 
1197 enum cleanup_prefix_rt_t {
1198 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1199 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1200 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1201 };
1202 
1203 /*
1204  * Check, whether the prefix for ifp would still need a prefix route
1205  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1206  * constants.
1207  *
1208  * 1) we don't purge prefix if address was not permanent.
1209  *    prefix is managed by its own lifetime.
1210  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1211  * 3) if there are no addresses, delete prefix.
1212  * 4) if there are still other permanent address(es),
1213  *    corresponding prefix is still permanent.
1214  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1215  *    don't purge the prefix, assume user space is managing it.
1216  * 6) otherwise, update prefix lifetime to the
1217  *    longest valid lifetime among the corresponding
1218  *    addresses on the device.
1219  *    Note: subsequent RA will update lifetime.
1220  **/
1221 static enum cleanup_prefix_rt_t
1222 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1223 {
1224 	struct inet6_ifaddr *ifa;
1225 	struct inet6_dev *idev = ifp->idev;
1226 	unsigned long lifetime;
1227 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1228 
1229 	*expires = jiffies;
1230 
1231 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1232 		if (ifa == ifp)
1233 			continue;
1234 		if (ifa->prefix_len != ifp->prefix_len ||
1235 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1236 				       ifp->prefix_len))
1237 			continue;
1238 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1239 			return CLEANUP_PREFIX_RT_NOP;
1240 
1241 		action = CLEANUP_PREFIX_RT_EXPIRE;
1242 
1243 		spin_lock(&ifa->lock);
1244 
1245 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1246 		/*
1247 		 * Note: Because this address is
1248 		 * not permanent, lifetime <
1249 		 * LONG_MAX / HZ here.
1250 		 */
1251 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1252 			*expires = ifa->tstamp + lifetime * HZ;
1253 		spin_unlock(&ifa->lock);
1254 	}
1255 
1256 	return action;
1257 }
1258 
1259 static void
1260 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1261 		     bool del_rt, bool del_peer)
1262 {
1263 	struct fib6_table *table;
1264 	struct fib6_info *f6i;
1265 
1266 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1267 					ifp->prefix_len,
1268 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1269 	if (f6i) {
1270 		if (del_rt)
1271 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1272 		else {
1273 			if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1274 				table = f6i->fib6_table;
1275 				spin_lock_bh(&table->tb6_lock);
1276 
1277 				fib6_set_expires(f6i, expires);
1278 				fib6_add_gc_list(f6i);
1279 
1280 				spin_unlock_bh(&table->tb6_lock);
1281 			}
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 unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1348 {
1349 	return READ_ONCE(idev->cnf.regen_min_advance) +
1350 		READ_ONCE(idev->cnf.regen_max_retry) *
1351 		READ_ONCE(idev->cnf.dad_transmits) *
1352 		max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1353 }
1354 
1355 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1356 {
1357 	struct inet6_dev *idev = ifp->idev;
1358 	unsigned long tmp_tstamp, age;
1359 	unsigned long regen_advance;
1360 	unsigned long now = jiffies;
1361 	u32 if_public_preferred_lft;
1362 	s32 cnf_temp_preferred_lft;
1363 	struct inet6_ifaddr *ift;
1364 	struct ifa6_config cfg;
1365 	long max_desync_factor;
1366 	struct in6_addr addr;
1367 	int ret = 0;
1368 
1369 	write_lock_bh(&idev->lock);
1370 
1371 retry:
1372 	in6_dev_hold(idev);
1373 	if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1374 		write_unlock_bh(&idev->lock);
1375 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1376 		in6_dev_put(idev);
1377 		ret = -1;
1378 		goto out;
1379 	}
1380 	spin_lock_bh(&ifp->lock);
1381 	if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1382 		WRITE_ONCE(idev->cnf.use_tempaddr, -1);	/*XXX*/
1383 		spin_unlock_bh(&ifp->lock);
1384 		write_unlock_bh(&idev->lock);
1385 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1386 			__func__);
1387 		in6_dev_put(idev);
1388 		ret = -1;
1389 		goto out;
1390 	}
1391 	in6_ifa_hold(ifp);
1392 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1393 	ipv6_gen_rnd_iid(&addr);
1394 
1395 	age = (now - ifp->tstamp) / HZ;
1396 
1397 	regen_advance = ipv6_get_regen_advance(idev);
1398 
1399 	/* recalculate max_desync_factor each time and update
1400 	 * idev->desync_factor if it's larger
1401 	 */
1402 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1403 	max_desync_factor = min_t(long,
1404 				  READ_ONCE(idev->cnf.max_desync_factor),
1405 				  cnf_temp_preferred_lft - regen_advance);
1406 
1407 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1408 		if (max_desync_factor > 0) {
1409 			get_random_bytes(&idev->desync_factor,
1410 					 sizeof(idev->desync_factor));
1411 			idev->desync_factor %= max_desync_factor;
1412 		} else {
1413 			idev->desync_factor = 0;
1414 		}
1415 	}
1416 
1417 	if_public_preferred_lft = ifp->prefered_lft;
1418 
1419 	memset(&cfg, 0, sizeof(cfg));
1420 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1421 			      READ_ONCE(idev->cnf.temp_valid_lft) + age);
1422 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1423 	cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1424 	cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1425 
1426 	cfg.plen = ifp->prefix_len;
1427 	tmp_tstamp = ifp->tstamp;
1428 	spin_unlock_bh(&ifp->lock);
1429 
1430 	write_unlock_bh(&idev->lock);
1431 
1432 	/* From RFC 4941:
1433 	 *
1434 	 *     A temporary address is created only if this calculated Preferred
1435 	 *     Lifetime is greater than REGEN_ADVANCE time units.  In
1436 	 *     particular, an implementation must not create a temporary address
1437 	 *     with a zero Preferred Lifetime.
1438 	 *
1439 	 *     ...
1440 	 *
1441 	 *     When creating a temporary address, the lifetime values MUST be
1442 	 *     derived from the corresponding prefix as follows:
1443 	 *
1444 	 *     ...
1445 	 *
1446 	 *     *  Its Preferred Lifetime is the lower of the Preferred Lifetime
1447 	 *        of the public address or TEMP_PREFERRED_LIFETIME -
1448 	 *        DESYNC_FACTOR.
1449 	 *
1450 	 * To comply with the RFC's requirements, clamp the preferred lifetime
1451 	 * to a minimum of regen_advance, unless that would exceed valid_lft or
1452 	 * ifp->prefered_lft.
1453 	 *
1454 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1455 	 * temporary addresses being generated.
1456 	 */
1457 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1458 	if (cfg.preferred_lft <= regen_advance + age) {
1459 		cfg.preferred_lft = regen_advance + age + 1;
1460 		if (cfg.preferred_lft > cfg.valid_lft ||
1461 		    cfg.preferred_lft > if_public_preferred_lft) {
1462 			in6_ifa_put(ifp);
1463 			in6_dev_put(idev);
1464 			ret = -1;
1465 			goto out;
1466 		}
1467 	}
1468 
1469 	cfg.ifa_flags = IFA_F_TEMPORARY;
1470 	/* set in addrconf_prefix_rcv() */
1471 	if (ifp->flags & IFA_F_OPTIMISTIC)
1472 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1473 
1474 	cfg.pfx = &addr;
1475 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1476 
1477 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1478 	if (IS_ERR(ift)) {
1479 		in6_ifa_put(ifp);
1480 		in6_dev_put(idev);
1481 		pr_info("%s: retry temporary address regeneration\n", __func__);
1482 		write_lock_bh(&idev->lock);
1483 		goto retry;
1484 	}
1485 
1486 	spin_lock_bh(&ift->lock);
1487 	ift->ifpub = ifp;
1488 	ift->cstamp = now;
1489 	ift->tstamp = tmp_tstamp;
1490 	spin_unlock_bh(&ift->lock);
1491 
1492 	addrconf_dad_start(ift);
1493 	in6_ifa_put(ift);
1494 	in6_dev_put(idev);
1495 out:
1496 	return ret;
1497 }
1498 
1499 /*
1500  *	Choose an appropriate source address (RFC3484)
1501  */
1502 enum {
1503 	IPV6_SADDR_RULE_INIT = 0,
1504 	IPV6_SADDR_RULE_LOCAL,
1505 	IPV6_SADDR_RULE_SCOPE,
1506 	IPV6_SADDR_RULE_PREFERRED,
1507 #ifdef CONFIG_IPV6_MIP6
1508 	IPV6_SADDR_RULE_HOA,
1509 #endif
1510 	IPV6_SADDR_RULE_OIF,
1511 	IPV6_SADDR_RULE_LABEL,
1512 	IPV6_SADDR_RULE_PRIVACY,
1513 	IPV6_SADDR_RULE_ORCHID,
1514 	IPV6_SADDR_RULE_PREFIX,
1515 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1516 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1517 #endif
1518 	IPV6_SADDR_RULE_MAX
1519 };
1520 
1521 struct ipv6_saddr_score {
1522 	int			rule;
1523 	int			addr_type;
1524 	struct inet6_ifaddr	*ifa;
1525 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1526 	int			scopedist;
1527 	int			matchlen;
1528 };
1529 
1530 struct ipv6_saddr_dst {
1531 	const struct in6_addr *addr;
1532 	int ifindex;
1533 	int scope;
1534 	int label;
1535 	unsigned int prefs;
1536 };
1537 
1538 static inline int ipv6_saddr_preferred(int type)
1539 {
1540 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1541 		return 1;
1542 	return 0;
1543 }
1544 
1545 static bool ipv6_use_optimistic_addr(const struct net *net,
1546 				     const struct inet6_dev *idev)
1547 {
1548 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1549 	if (!idev)
1550 		return false;
1551 	if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1552 	    !READ_ONCE(idev->cnf.optimistic_dad))
1553 		return false;
1554 	if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1555 	    !READ_ONCE(idev->cnf.use_optimistic))
1556 		return false;
1557 
1558 	return true;
1559 #else
1560 	return false;
1561 #endif
1562 }
1563 
1564 static bool ipv6_allow_optimistic_dad(const struct net *net,
1565 				      const struct inet6_dev *idev)
1566 {
1567 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1568 	if (!idev)
1569 		return false;
1570 	if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1571 	    !READ_ONCE(idev->cnf.optimistic_dad))
1572 		return false;
1573 
1574 	return true;
1575 #else
1576 	return false;
1577 #endif
1578 }
1579 
1580 static int ipv6_get_saddr_eval(struct net *net,
1581 			       struct ipv6_saddr_score *score,
1582 			       struct ipv6_saddr_dst *dst,
1583 			       int i)
1584 {
1585 	int ret;
1586 
1587 	if (i <= score->rule) {
1588 		switch (i) {
1589 		case IPV6_SADDR_RULE_SCOPE:
1590 			ret = score->scopedist;
1591 			break;
1592 		case IPV6_SADDR_RULE_PREFIX:
1593 			ret = score->matchlen;
1594 			break;
1595 		default:
1596 			ret = !!test_bit(i, score->scorebits);
1597 		}
1598 		goto out;
1599 	}
1600 
1601 	switch (i) {
1602 	case IPV6_SADDR_RULE_INIT:
1603 		/* Rule 0: remember if hiscore is not ready yet */
1604 		ret = !!score->ifa;
1605 		break;
1606 	case IPV6_SADDR_RULE_LOCAL:
1607 		/* Rule 1: Prefer same address */
1608 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1609 		break;
1610 	case IPV6_SADDR_RULE_SCOPE:
1611 		/* Rule 2: Prefer appropriate scope
1612 		 *
1613 		 *      ret
1614 		 *       ^
1615 		 *    -1 |  d 15
1616 		 *    ---+--+-+---> scope
1617 		 *       |
1618 		 *       |             d is scope of the destination.
1619 		 *  B-d  |  \
1620 		 *       |   \      <- smaller scope is better if
1621 		 *  B-15 |    \        if scope is enough for destination.
1622 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1623 		 * d-C-1 | /
1624 		 *       |/         <- greater is better
1625 		 *   -C  /             if scope is not enough for destination.
1626 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1627 		 *
1628 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1629 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1630 		 * Assume B = 0 and we get C > 29.
1631 		 */
1632 		ret = __ipv6_addr_src_scope(score->addr_type);
1633 		if (ret >= dst->scope)
1634 			ret = -ret;
1635 		else
1636 			ret -= 128;	/* 30 is enough */
1637 		score->scopedist = ret;
1638 		break;
1639 	case IPV6_SADDR_RULE_PREFERRED:
1640 	    {
1641 		/* Rule 3: Avoid deprecated and optimistic addresses */
1642 		u8 avoid = IFA_F_DEPRECATED;
1643 
1644 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1645 			avoid |= IFA_F_OPTIMISTIC;
1646 		ret = ipv6_saddr_preferred(score->addr_type) ||
1647 		      !(score->ifa->flags & avoid);
1648 		break;
1649 	    }
1650 #ifdef CONFIG_IPV6_MIP6
1651 	case IPV6_SADDR_RULE_HOA:
1652 	    {
1653 		/* Rule 4: Prefer home address */
1654 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1655 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1656 		break;
1657 	    }
1658 #endif
1659 	case IPV6_SADDR_RULE_OIF:
1660 		/* Rule 5: Prefer outgoing interface */
1661 		ret = (!dst->ifindex ||
1662 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1663 		break;
1664 	case IPV6_SADDR_RULE_LABEL:
1665 		/* Rule 6: Prefer matching label */
1666 		ret = ipv6_addr_label(net,
1667 				      &score->ifa->addr, score->addr_type,
1668 				      score->ifa->idev->dev->ifindex) == dst->label;
1669 		break;
1670 	case IPV6_SADDR_RULE_PRIVACY:
1671 	    {
1672 		/* Rule 7: Prefer public address
1673 		 * Note: prefer temporary address if use_tempaddr >= 2
1674 		 */
1675 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1676 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1677 				READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1678 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1679 		break;
1680 	    }
1681 	case IPV6_SADDR_RULE_ORCHID:
1682 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1683 		 *	    non-ORCHID vs non-ORCHID
1684 		 */
1685 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1686 			ipv6_addr_orchid(dst->addr));
1687 		break;
1688 	case IPV6_SADDR_RULE_PREFIX:
1689 		/* Rule 8: Use longest matching prefix */
1690 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1691 		if (ret > score->ifa->prefix_len)
1692 			ret = score->ifa->prefix_len;
1693 		score->matchlen = ret;
1694 		break;
1695 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1696 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1697 		/* Optimistic addresses still have lower precedence than other
1698 		 * preferred addresses.
1699 		 */
1700 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1701 		break;
1702 #endif
1703 	default:
1704 		ret = 0;
1705 	}
1706 
1707 	if (ret)
1708 		__set_bit(i, score->scorebits);
1709 	score->rule = i;
1710 out:
1711 	return ret;
1712 }
1713 
1714 static int __ipv6_dev_get_saddr(struct net *net,
1715 				struct ipv6_saddr_dst *dst,
1716 				struct inet6_dev *idev,
1717 				struct ipv6_saddr_score *scores,
1718 				int hiscore_idx)
1719 {
1720 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1721 
1722 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1723 		int i;
1724 
1725 		/*
1726 		 * - Tentative Address (RFC2462 section 5.4)
1727 		 *  - A tentative address is not considered
1728 		 *    "assigned to an interface" in the traditional
1729 		 *    sense, unless it is also flagged as optimistic.
1730 		 * - Candidate Source Address (section 4)
1731 		 *  - In any case, anycast addresses, multicast
1732 		 *    addresses, and the unspecified address MUST
1733 		 *    NOT be included in a candidate set.
1734 		 */
1735 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1736 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1737 			continue;
1738 
1739 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1740 
1741 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1742 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1743 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1744 					    idev->dev->name);
1745 			continue;
1746 		}
1747 
1748 		score->rule = -1;
1749 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1750 
1751 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1752 			int minihiscore, miniscore;
1753 
1754 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1755 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1756 
1757 			if (minihiscore > miniscore) {
1758 				if (i == IPV6_SADDR_RULE_SCOPE &&
1759 				    score->scopedist > 0) {
1760 					/*
1761 					 * special case:
1762 					 * each remaining entry
1763 					 * has too small (not enough)
1764 					 * scope, because ifa entries
1765 					 * are sorted by their scope
1766 					 * values.
1767 					 */
1768 					goto out;
1769 				}
1770 				break;
1771 			} else if (minihiscore < miniscore) {
1772 				swap(hiscore, score);
1773 				hiscore_idx = 1 - hiscore_idx;
1774 
1775 				/* restore our iterator */
1776 				score->ifa = hiscore->ifa;
1777 
1778 				break;
1779 			}
1780 		}
1781 	}
1782 out:
1783 	return hiscore_idx;
1784 }
1785 
1786 static int ipv6_get_saddr_master(struct net *net,
1787 				 const struct net_device *dst_dev,
1788 				 const struct net_device *master,
1789 				 struct ipv6_saddr_dst *dst,
1790 				 struct ipv6_saddr_score *scores,
1791 				 int hiscore_idx)
1792 {
1793 	struct inet6_dev *idev;
1794 
1795 	idev = __in6_dev_get(dst_dev);
1796 	if (idev)
1797 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1798 						   scores, hiscore_idx);
1799 
1800 	idev = __in6_dev_get(master);
1801 	if (idev)
1802 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1803 						   scores, hiscore_idx);
1804 
1805 	return hiscore_idx;
1806 }
1807 
1808 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1809 		       const struct in6_addr *daddr, unsigned int prefs,
1810 		       struct in6_addr *saddr)
1811 {
1812 	struct ipv6_saddr_score scores[2], *hiscore;
1813 	struct ipv6_saddr_dst dst;
1814 	struct inet6_dev *idev;
1815 	struct net_device *dev;
1816 	int dst_type;
1817 	bool use_oif_addr = false;
1818 	int hiscore_idx = 0;
1819 	int ret = 0;
1820 
1821 	dst_type = __ipv6_addr_type(daddr);
1822 	dst.addr = daddr;
1823 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1824 	dst.scope = __ipv6_addr_src_scope(dst_type);
1825 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1826 	dst.prefs = prefs;
1827 
1828 	scores[hiscore_idx].rule = -1;
1829 	scores[hiscore_idx].ifa = NULL;
1830 
1831 	rcu_read_lock();
1832 
1833 	/* Candidate Source Address (section 4)
1834 	 *  - multicast and link-local destination address,
1835 	 *    the set of candidate source address MUST only
1836 	 *    include addresses assigned to interfaces
1837 	 *    belonging to the same link as the outgoing
1838 	 *    interface.
1839 	 * (- For site-local destination addresses, the
1840 	 *    set of candidate source addresses MUST only
1841 	 *    include addresses assigned to interfaces
1842 	 *    belonging to the same site as the outgoing
1843 	 *    interface.)
1844 	 *  - "It is RECOMMENDED that the candidate source addresses
1845 	 *    be the set of unicast addresses assigned to the
1846 	 *    interface that will be used to send to the destination
1847 	 *    (the 'outgoing' interface)." (RFC 6724)
1848 	 */
1849 	if (dst_dev) {
1850 		idev = __in6_dev_get(dst_dev);
1851 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1852 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1853 		    (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1854 			use_oif_addr = true;
1855 		}
1856 	}
1857 
1858 	if (use_oif_addr) {
1859 		if (idev)
1860 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1861 	} else {
1862 		const struct net_device *master;
1863 		int master_idx = 0;
1864 
1865 		/* if dst_dev exists and is enslaved to an L3 device, then
1866 		 * prefer addresses from dst_dev and then the master over
1867 		 * any other enslaved devices in the L3 domain.
1868 		 */
1869 		master = l3mdev_master_dev_rcu(dst_dev);
1870 		if (master) {
1871 			master_idx = master->ifindex;
1872 
1873 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1874 							    master, &dst,
1875 							    scores, hiscore_idx);
1876 
1877 			if (scores[hiscore_idx].ifa &&
1878 			    scores[hiscore_idx].scopedist >= 0)
1879 				goto out;
1880 		}
1881 
1882 		for_each_netdev_rcu(net, dev) {
1883 			/* only consider addresses on devices in the
1884 			 * same L3 domain
1885 			 */
1886 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1887 				continue;
1888 			idev = __in6_dev_get(dev);
1889 			if (!idev)
1890 				continue;
1891 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1892 		}
1893 	}
1894 
1895 out:
1896 	hiscore = &scores[hiscore_idx];
1897 	if (!hiscore->ifa)
1898 		ret = -EADDRNOTAVAIL;
1899 	else
1900 		*saddr = hiscore->ifa->addr;
1901 
1902 	rcu_read_unlock();
1903 	return ret;
1904 }
1905 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1906 
1907 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1908 			      u32 banned_flags)
1909 {
1910 	struct inet6_ifaddr *ifp;
1911 	int err = -EADDRNOTAVAIL;
1912 
1913 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1914 		if (ifp->scope > IFA_LINK)
1915 			break;
1916 		if (ifp->scope == IFA_LINK &&
1917 		    !(ifp->flags & banned_flags)) {
1918 			*addr = ifp->addr;
1919 			err = 0;
1920 			break;
1921 		}
1922 	}
1923 	return err;
1924 }
1925 
1926 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1927 		    u32 banned_flags)
1928 {
1929 	struct inet6_dev *idev;
1930 	int err = -EADDRNOTAVAIL;
1931 
1932 	rcu_read_lock();
1933 	idev = __in6_dev_get(dev);
1934 	if (idev) {
1935 		read_lock_bh(&idev->lock);
1936 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1937 		read_unlock_bh(&idev->lock);
1938 	}
1939 	rcu_read_unlock();
1940 	return err;
1941 }
1942 
1943 static int ipv6_count_addresses(const struct inet6_dev *idev)
1944 {
1945 	const struct inet6_ifaddr *ifp;
1946 	int cnt = 0;
1947 
1948 	rcu_read_lock();
1949 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1950 		cnt++;
1951 	rcu_read_unlock();
1952 	return cnt;
1953 }
1954 
1955 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1956 		  const struct net_device *dev, int strict)
1957 {
1958 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1959 				       strict, IFA_F_TENTATIVE);
1960 }
1961 EXPORT_SYMBOL(ipv6_chk_addr);
1962 
1963 /* device argument is used to find the L3 domain of interest. If
1964  * skip_dev_check is set, then the ifp device is not checked against
1965  * the passed in dev argument. So the 2 cases for addresses checks are:
1966  *   1. does the address exist in the L3 domain that dev is part of
1967  *      (skip_dev_check = true), or
1968  *
1969  *   2. does the address exist on the specific device
1970  *      (skip_dev_check = false)
1971  */
1972 static struct net_device *
1973 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1974 			  const struct net_device *dev, bool skip_dev_check,
1975 			  int strict, u32 banned_flags)
1976 {
1977 	unsigned int hash = inet6_addr_hash(net, addr);
1978 	struct net_device *l3mdev, *ndev;
1979 	struct inet6_ifaddr *ifp;
1980 	u32 ifp_flags;
1981 
1982 	rcu_read_lock();
1983 
1984 	l3mdev = l3mdev_master_dev_rcu(dev);
1985 	if (skip_dev_check)
1986 		dev = NULL;
1987 
1988 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1989 		ndev = ifp->idev->dev;
1990 
1991 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1992 			continue;
1993 
1994 		/* Decouple optimistic from tentative for evaluation here.
1995 		 * Ban optimistic addresses explicitly, when required.
1996 		 */
1997 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1998 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1999 			    : ifp->flags;
2000 		if (ipv6_addr_equal(&ifp->addr, addr) &&
2001 		    !(ifp_flags&banned_flags) &&
2002 		    (!dev || ndev == dev ||
2003 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2004 			rcu_read_unlock();
2005 			return ndev;
2006 		}
2007 	}
2008 
2009 	rcu_read_unlock();
2010 	return NULL;
2011 }
2012 
2013 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2014 			    const struct net_device *dev, bool skip_dev_check,
2015 			    int strict, u32 banned_flags)
2016 {
2017 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2018 					 strict, banned_flags) ? 1 : 0;
2019 }
2020 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2021 
2022 
2023 /* Compares an address/prefix_len with addresses on device @dev.
2024  * If one is found it returns true.
2025  */
2026 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2027 	const unsigned int prefix_len, struct net_device *dev)
2028 {
2029 	const struct inet6_ifaddr *ifa;
2030 	const struct inet6_dev *idev;
2031 	bool ret = false;
2032 
2033 	rcu_read_lock();
2034 	idev = __in6_dev_get(dev);
2035 	if (idev) {
2036 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2037 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2038 			if (ret)
2039 				break;
2040 		}
2041 	}
2042 	rcu_read_unlock();
2043 
2044 	return ret;
2045 }
2046 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2047 
2048 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2049 {
2050 	const struct inet6_ifaddr *ifa;
2051 	const struct inet6_dev *idev;
2052 	int	onlink;
2053 
2054 	onlink = 0;
2055 	rcu_read_lock();
2056 	idev = __in6_dev_get(dev);
2057 	if (idev) {
2058 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2059 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
2060 						   ifa->prefix_len);
2061 			if (onlink)
2062 				break;
2063 		}
2064 	}
2065 	rcu_read_unlock();
2066 	return onlink;
2067 }
2068 EXPORT_SYMBOL(ipv6_chk_prefix);
2069 
2070 /**
2071  * ipv6_dev_find - find the first device with a given source address.
2072  * @net: the net namespace
2073  * @addr: the source address
2074  * @dev: used to find the L3 domain of interest
2075  *
2076  * The caller should be protected by RCU, or RTNL.
2077  */
2078 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2079 				 struct net_device *dev)
2080 {
2081 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2082 					 IFA_F_TENTATIVE);
2083 }
2084 EXPORT_SYMBOL(ipv6_dev_find);
2085 
2086 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2087 				     struct net_device *dev, int strict)
2088 {
2089 	unsigned int hash = inet6_addr_hash(net, addr);
2090 	struct inet6_ifaddr *ifp, *result = NULL;
2091 
2092 	rcu_read_lock();
2093 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2094 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2095 			if (!dev || ifp->idev->dev == dev ||
2096 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2097 				if (in6_ifa_hold_safe(ifp)) {
2098 					result = ifp;
2099 					break;
2100 				}
2101 			}
2102 		}
2103 	}
2104 	rcu_read_unlock();
2105 
2106 	return result;
2107 }
2108 
2109 /* Gets referenced address, destroys ifaddr */
2110 
2111 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2112 {
2113 	if (dad_failed)
2114 		ifp->flags |= IFA_F_DADFAILED;
2115 
2116 	if (ifp->flags&IFA_F_TEMPORARY) {
2117 		struct inet6_ifaddr *ifpub;
2118 		spin_lock_bh(&ifp->lock);
2119 		ifpub = ifp->ifpub;
2120 		if (ifpub) {
2121 			in6_ifa_hold(ifpub);
2122 			spin_unlock_bh(&ifp->lock);
2123 			ipv6_create_tempaddr(ifpub, true);
2124 			in6_ifa_put(ifpub);
2125 		} else {
2126 			spin_unlock_bh(&ifp->lock);
2127 		}
2128 		ipv6_del_addr(ifp);
2129 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2130 		spin_lock_bh(&ifp->lock);
2131 		addrconf_del_dad_work(ifp);
2132 		ifp->flags |= IFA_F_TENTATIVE;
2133 		if (dad_failed)
2134 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2135 		spin_unlock_bh(&ifp->lock);
2136 		if (dad_failed)
2137 			ipv6_ifa_notify(0, ifp);
2138 		in6_ifa_put(ifp);
2139 	} else {
2140 		ipv6_del_addr(ifp);
2141 	}
2142 }
2143 
2144 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2145 {
2146 	int err = -ENOENT;
2147 
2148 	spin_lock_bh(&ifp->lock);
2149 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2150 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2151 		err = 0;
2152 	}
2153 	spin_unlock_bh(&ifp->lock);
2154 
2155 	return err;
2156 }
2157 
2158 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2159 {
2160 	struct inet6_dev *idev = ifp->idev;
2161 	struct net *net = dev_net(idev->dev);
2162 	int max_addresses;
2163 
2164 	if (addrconf_dad_end(ifp)) {
2165 		in6_ifa_put(ifp);
2166 		return;
2167 	}
2168 
2169 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2170 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2171 
2172 	spin_lock_bh(&ifp->lock);
2173 
2174 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2175 		struct in6_addr new_addr;
2176 		struct inet6_ifaddr *ifp2;
2177 		int retries = ifp->stable_privacy_retry + 1;
2178 		struct ifa6_config cfg = {
2179 			.pfx = &new_addr,
2180 			.plen = ifp->prefix_len,
2181 			.ifa_flags = ifp->flags,
2182 			.valid_lft = ifp->valid_lft,
2183 			.preferred_lft = ifp->prefered_lft,
2184 			.scope = ifp->scope,
2185 		};
2186 
2187 		if (retries > net->ipv6.sysctl.idgen_retries) {
2188 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2189 					     ifp->idev->dev->name);
2190 			goto errdad;
2191 		}
2192 
2193 		new_addr = ifp->addr;
2194 		if (ipv6_generate_stable_address(&new_addr, retries,
2195 						 idev))
2196 			goto errdad;
2197 
2198 		spin_unlock_bh(&ifp->lock);
2199 
2200 		max_addresses = READ_ONCE(idev->cnf.max_addresses);
2201 		if (max_addresses &&
2202 		    ipv6_count_addresses(idev) >= max_addresses)
2203 			goto lock_errdad;
2204 
2205 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2206 				     ifp->idev->dev->name);
2207 
2208 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2209 		if (IS_ERR(ifp2))
2210 			goto lock_errdad;
2211 
2212 		spin_lock_bh(&ifp2->lock);
2213 		ifp2->stable_privacy_retry = retries;
2214 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2215 		spin_unlock_bh(&ifp2->lock);
2216 
2217 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2218 		in6_ifa_put(ifp2);
2219 lock_errdad:
2220 		spin_lock_bh(&ifp->lock);
2221 	}
2222 
2223 errdad:
2224 	/* transition from _POSTDAD to _ERRDAD */
2225 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2226 	spin_unlock_bh(&ifp->lock);
2227 
2228 	addrconf_mod_dad_work(ifp, 0);
2229 	in6_ifa_put(ifp);
2230 }
2231 
2232 /* Join to solicited addr multicast group. */
2233 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2234 {
2235 	struct in6_addr maddr;
2236 
2237 	if (READ_ONCE(dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2238 		return;
2239 
2240 	addrconf_addr_solict_mult(addr, &maddr);
2241 	ipv6_dev_mc_inc(dev, &maddr);
2242 }
2243 
2244 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2245 {
2246 	struct in6_addr maddr;
2247 
2248 	if (READ_ONCE(idev->dev->flags) & (IFF_LOOPBACK | IFF_NOARP))
2249 		return;
2250 
2251 	addrconf_addr_solict_mult(addr, &maddr);
2252 	__ipv6_dev_mc_dec(idev, &maddr);
2253 }
2254 
2255 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2256 {
2257 	struct in6_addr addr;
2258 
2259 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2260 		return;
2261 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2262 	if (ipv6_addr_any(&addr))
2263 		return;
2264 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2265 }
2266 
2267 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2268 {
2269 	struct in6_addr addr;
2270 
2271 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2272 		return;
2273 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2274 	if (ipv6_addr_any(&addr))
2275 		return;
2276 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2277 }
2278 
2279 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2280 {
2281 	switch (dev->addr_len) {
2282 	case ETH_ALEN:
2283 		memcpy(eui, dev->dev_addr, 3);
2284 		eui[3] = 0xFF;
2285 		eui[4] = 0xFE;
2286 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2287 		break;
2288 	case EUI64_ADDR_LEN:
2289 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2290 		eui[0] ^= 2;
2291 		break;
2292 	default:
2293 		return -1;
2294 	}
2295 
2296 	return 0;
2297 }
2298 
2299 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2300 {
2301 	const union fwnet_hwaddr *ha;
2302 
2303 	if (dev->addr_len != FWNET_ALEN)
2304 		return -1;
2305 
2306 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2307 
2308 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2309 	eui[0] ^= 2;
2310 	return 0;
2311 }
2312 
2313 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2314 {
2315 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2316 	if (dev->addr_len != ARCNET_ALEN)
2317 		return -1;
2318 	memset(eui, 0, 7);
2319 	eui[7] = *(u8 *)dev->dev_addr;
2320 	return 0;
2321 }
2322 
2323 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2324 {
2325 	if (dev->addr_len != INFINIBAND_ALEN)
2326 		return -1;
2327 	memcpy(eui, dev->dev_addr + 12, 8);
2328 	eui[0] |= 2;
2329 	return 0;
2330 }
2331 
2332 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2333 {
2334 	if (addr == 0)
2335 		return -1;
2336 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2337 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2338 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2339 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2340 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2341 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2342 	eui[1] = 0;
2343 	eui[2] = 0x5E;
2344 	eui[3] = 0xFE;
2345 	memcpy(eui + 4, &addr, 4);
2346 	return 0;
2347 }
2348 
2349 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2350 {
2351 	if (dev->priv_flags & IFF_ISATAP)
2352 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2353 	return -1;
2354 }
2355 
2356 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2357 {
2358 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2359 }
2360 
2361 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2362 {
2363 	memcpy(eui, dev->perm_addr, 3);
2364 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2365 	eui[3] = 0xFF;
2366 	eui[4] = 0xFE;
2367 	eui[0] ^= 2;
2368 	return 0;
2369 }
2370 
2371 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2372 {
2373 	switch (dev->type) {
2374 	case ARPHRD_ETHER:
2375 	case ARPHRD_FDDI:
2376 		return addrconf_ifid_eui48(eui, dev);
2377 	case ARPHRD_ARCNET:
2378 		return addrconf_ifid_arcnet(eui, dev);
2379 	case ARPHRD_INFINIBAND:
2380 		return addrconf_ifid_infiniband(eui, dev);
2381 	case ARPHRD_SIT:
2382 		return addrconf_ifid_sit(eui, dev);
2383 	case ARPHRD_IPGRE:
2384 	case ARPHRD_TUNNEL:
2385 		return addrconf_ifid_gre(eui, dev);
2386 	case ARPHRD_6LOWPAN:
2387 		return addrconf_ifid_6lowpan(eui, dev);
2388 	case ARPHRD_IEEE1394:
2389 		return addrconf_ifid_ieee1394(eui, dev);
2390 	case ARPHRD_TUNNEL6:
2391 	case ARPHRD_IP6GRE:
2392 	case ARPHRD_RAWIP:
2393 		return addrconf_ifid_ip6tnl(eui, dev);
2394 	}
2395 	return -1;
2396 }
2397 
2398 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2399 {
2400 	int err = -1;
2401 	struct inet6_ifaddr *ifp;
2402 
2403 	read_lock_bh(&idev->lock);
2404 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2405 		if (ifp->scope > IFA_LINK)
2406 			break;
2407 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2408 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2409 			err = 0;
2410 			break;
2411 		}
2412 	}
2413 	read_unlock_bh(&idev->lock);
2414 	return err;
2415 }
2416 
2417 /* Generation of a randomized Interface Identifier
2418  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2419  */
2420 
2421 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2422 {
2423 regen:
2424 	get_random_bytes(&addr->s6_addr[8], 8);
2425 
2426 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2427 	 * check if generated address is not inappropriate:
2428 	 *
2429 	 * - Reserved IPv6 Interface Identifiers
2430 	 * - XXX: already assigned to an address on the device
2431 	 */
2432 
2433 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2434 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2435 		goto regen;
2436 
2437 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2438 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2439 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2440 	 */
2441 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2442 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2443 		goto regen;
2444 
2445 	/* Reserved subnet anycast addresses */
2446 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2447 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2448 		goto regen;
2449 }
2450 
2451 /*
2452  *	Add prefix route.
2453  */
2454 
2455 static void
2456 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2457 		      struct net_device *dev, unsigned long expires,
2458 		      u32 flags, gfp_t gfp_flags)
2459 {
2460 	struct fib6_config cfg = {
2461 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2462 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2463 		.fc_ifindex = dev->ifindex,
2464 		.fc_expires = expires,
2465 		.fc_dst_len = plen,
2466 		.fc_flags = RTF_UP | flags,
2467 		.fc_nlinfo.nl_net = dev_net(dev),
2468 		.fc_protocol = RTPROT_KERNEL,
2469 		.fc_type = RTN_UNICAST,
2470 	};
2471 
2472 	cfg.fc_dst = *pfx;
2473 
2474 	/* Prevent useless cloning on PtP SIT.
2475 	   This thing is done here expecting that the whole
2476 	   class of non-broadcast devices need not cloning.
2477 	 */
2478 #if IS_ENABLED(CONFIG_IPV6_SIT)
2479 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2480 		cfg.fc_flags |= RTF_NONEXTHOP;
2481 #endif
2482 
2483 	ip6_route_add(&cfg, gfp_flags, NULL);
2484 }
2485 
2486 
2487 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2488 						  int plen,
2489 						  const struct net_device *dev,
2490 						  u32 flags, u32 noflags,
2491 						  bool no_gw)
2492 {
2493 	struct fib6_node *fn;
2494 	struct fib6_info *rt = NULL;
2495 	struct fib6_table *table;
2496 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2497 
2498 	table = fib6_get_table(dev_net(dev), tb_id);
2499 	if (!table)
2500 		return NULL;
2501 
2502 	rcu_read_lock();
2503 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2504 	if (!fn)
2505 		goto out;
2506 
2507 	for_each_fib6_node_rt_rcu(fn) {
2508 		/* prefix routes only use builtin fib6_nh */
2509 		if (rt->nh)
2510 			continue;
2511 
2512 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2513 			continue;
2514 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2515 			continue;
2516 		if ((rt->fib6_flags & flags) != flags)
2517 			continue;
2518 		if ((rt->fib6_flags & noflags) != 0)
2519 			continue;
2520 		if (!fib6_info_hold_safe(rt))
2521 			continue;
2522 		break;
2523 	}
2524 out:
2525 	rcu_read_unlock();
2526 	return rt;
2527 }
2528 
2529 
2530 /* Create "default" multicast route to the interface */
2531 
2532 static void addrconf_add_mroute(struct net_device *dev)
2533 {
2534 	struct fib6_config cfg = {
2535 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2536 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2537 		.fc_ifindex = dev->ifindex,
2538 		.fc_dst_len = 8,
2539 		.fc_flags = RTF_UP,
2540 		.fc_type = RTN_MULTICAST,
2541 		.fc_nlinfo.nl_net = dev_net(dev),
2542 		.fc_protocol = RTPROT_KERNEL,
2543 	};
2544 
2545 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2546 
2547 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2548 }
2549 
2550 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2551 {
2552 	struct inet6_dev *idev;
2553 
2554 	ASSERT_RTNL();
2555 
2556 	idev = ipv6_find_idev(dev);
2557 	if (IS_ERR(idev))
2558 		return idev;
2559 
2560 	if (idev->cnf.disable_ipv6)
2561 		return ERR_PTR(-EACCES);
2562 
2563 	/* Add default multicast route */
2564 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2565 		addrconf_add_mroute(dev);
2566 
2567 	return idev;
2568 }
2569 
2570 static void delete_tempaddrs(struct inet6_dev *idev,
2571 			     struct inet6_ifaddr *ifp)
2572 {
2573 	struct inet6_ifaddr *ift, *tmp;
2574 
2575 	write_lock_bh(&idev->lock);
2576 	list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) {
2577 		if (ift->ifpub != ifp)
2578 			continue;
2579 
2580 		in6_ifa_hold(ift);
2581 		write_unlock_bh(&idev->lock);
2582 		ipv6_del_addr(ift);
2583 		write_lock_bh(&idev->lock);
2584 	}
2585 	write_unlock_bh(&idev->lock);
2586 }
2587 
2588 static void manage_tempaddrs(struct inet6_dev *idev,
2589 			     struct inet6_ifaddr *ifp,
2590 			     __u32 valid_lft, __u32 prefered_lft,
2591 			     bool create, unsigned long now)
2592 {
2593 	u32 flags;
2594 	struct inet6_ifaddr *ift;
2595 
2596 	read_lock_bh(&idev->lock);
2597 	/* update all temporary addresses in the list */
2598 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2599 		int age, max_valid, max_prefered;
2600 
2601 		if (ifp != ift->ifpub)
2602 			continue;
2603 
2604 		/* RFC 4941 section 3.3:
2605 		 * If a received option will extend the lifetime of a public
2606 		 * address, the lifetimes of temporary addresses should
2607 		 * be extended, subject to the overall constraint that no
2608 		 * temporary addresses should ever remain "valid" or "preferred"
2609 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2610 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2611 		 */
2612 		age = (now - ift->cstamp) / HZ;
2613 		max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2614 		if (max_valid < 0)
2615 			max_valid = 0;
2616 
2617 		max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2618 			       idev->desync_factor - age;
2619 		if (max_prefered < 0)
2620 			max_prefered = 0;
2621 
2622 		if (valid_lft > max_valid)
2623 			valid_lft = max_valid;
2624 
2625 		if (prefered_lft > max_prefered)
2626 			prefered_lft = max_prefered;
2627 
2628 		spin_lock(&ift->lock);
2629 		flags = ift->flags;
2630 		ift->valid_lft = valid_lft;
2631 		ift->prefered_lft = prefered_lft;
2632 		ift->tstamp = now;
2633 		if (prefered_lft > 0)
2634 			ift->flags &= ~IFA_F_DEPRECATED;
2635 
2636 		spin_unlock(&ift->lock);
2637 		if (!(flags&IFA_F_TENTATIVE))
2638 			ipv6_ifa_notify(0, ift);
2639 	}
2640 
2641 	/* Also create a temporary address if it's enabled but no temporary
2642 	 * address currently exists.
2643 	 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2644 	 * as part of cleanup (ie. deleting the mngtmpaddr).
2645 	 * We don't want that to result in creating a new temporary ip address.
2646 	 */
2647 	if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2648 		create = true;
2649 
2650 	if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2651 		/* When a new public address is created as described
2652 		 * in [ADDRCONF], also create a new temporary address.
2653 		 */
2654 		read_unlock_bh(&idev->lock);
2655 		ipv6_create_tempaddr(ifp, false);
2656 	} else {
2657 		read_unlock_bh(&idev->lock);
2658 	}
2659 }
2660 
2661 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2662 {
2663 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2664 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2665 }
2666 
2667 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2668 				 const struct prefix_info *pinfo,
2669 				 struct inet6_dev *in6_dev,
2670 				 const struct in6_addr *addr, int addr_type,
2671 				 u32 addr_flags, bool sllao, bool tokenized,
2672 				 __u32 valid_lft, u32 prefered_lft)
2673 {
2674 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2675 	int create = 0, update_lft = 0;
2676 
2677 	if (!ifp && valid_lft) {
2678 		int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2679 		struct ifa6_config cfg = {
2680 			.pfx = addr,
2681 			.plen = pinfo->prefix_len,
2682 			.ifa_flags = addr_flags,
2683 			.valid_lft = valid_lft,
2684 			.preferred_lft = prefered_lft,
2685 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2686 			.ifa_proto = IFAPROT_KERNEL_RA
2687 		};
2688 
2689 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2690 		if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2691 		     READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2692 		    !net->ipv6.devconf_all->forwarding && sllao)
2693 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2694 #endif
2695 
2696 		/* Do not allow to create too much of autoconfigured
2697 		 * addresses; this would be too easy way to crash kernel.
2698 		 */
2699 		if (!max_addresses ||
2700 		    ipv6_count_addresses(in6_dev) < max_addresses)
2701 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2702 
2703 		if (IS_ERR_OR_NULL(ifp))
2704 			return -1;
2705 
2706 		create = 1;
2707 		spin_lock_bh(&ifp->lock);
2708 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2709 		ifp->cstamp = jiffies;
2710 		ifp->tokenized = tokenized;
2711 		spin_unlock_bh(&ifp->lock);
2712 		addrconf_dad_start(ifp);
2713 	}
2714 
2715 	if (ifp) {
2716 		u32 flags;
2717 		unsigned long now;
2718 		u32 stored_lft;
2719 
2720 		/* update lifetime (RFC2462 5.5.3 e) */
2721 		spin_lock_bh(&ifp->lock);
2722 		now = jiffies;
2723 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2724 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2725 		else
2726 			stored_lft = 0;
2727 
2728 		/* RFC4862 Section 5.5.3e:
2729 		 * "Note that the preferred lifetime of the
2730 		 *  corresponding address is always reset to
2731 		 *  the Preferred Lifetime in the received
2732 		 *  Prefix Information option, regardless of
2733 		 *  whether the valid lifetime is also reset or
2734 		 *  ignored."
2735 		 *
2736 		 * So we should always update prefered_lft here.
2737 		 */
2738 		update_lft = !create && stored_lft;
2739 
2740 		if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2741 			const u32 minimum_lft = min_t(u32,
2742 				stored_lft, MIN_VALID_LIFETIME);
2743 			valid_lft = max(valid_lft, minimum_lft);
2744 		}
2745 
2746 		if (update_lft) {
2747 			ifp->valid_lft = valid_lft;
2748 			ifp->prefered_lft = prefered_lft;
2749 			WRITE_ONCE(ifp->tstamp, now);
2750 			flags = ifp->flags;
2751 			ifp->flags &= ~IFA_F_DEPRECATED;
2752 			spin_unlock_bh(&ifp->lock);
2753 
2754 			if (!(flags&IFA_F_TENTATIVE))
2755 				ipv6_ifa_notify(0, ifp);
2756 		} else
2757 			spin_unlock_bh(&ifp->lock);
2758 
2759 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2760 				 create, now);
2761 
2762 		in6_ifa_put(ifp);
2763 		addrconf_verify(net);
2764 	}
2765 
2766 	return 0;
2767 }
2768 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2769 
2770 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2771 {
2772 	struct prefix_info *pinfo;
2773 	struct fib6_table *table;
2774 	__u32 valid_lft;
2775 	__u32 prefered_lft;
2776 	int addr_type, err;
2777 	u32 addr_flags = 0;
2778 	struct inet6_dev *in6_dev;
2779 	struct net *net = dev_net(dev);
2780 	bool ignore_autoconf = false;
2781 
2782 	pinfo = (struct prefix_info *) opt;
2783 
2784 	if (len < sizeof(struct prefix_info)) {
2785 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2786 		return;
2787 	}
2788 
2789 	/*
2790 	 *	Validation checks ([ADDRCONF], page 19)
2791 	 */
2792 
2793 	addr_type = ipv6_addr_type(&pinfo->prefix);
2794 
2795 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2796 		return;
2797 
2798 	valid_lft = ntohl(pinfo->valid);
2799 	prefered_lft = ntohl(pinfo->prefered);
2800 
2801 	if (prefered_lft > valid_lft) {
2802 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2803 		return;
2804 	}
2805 
2806 	in6_dev = in6_dev_get(dev);
2807 
2808 	if (!in6_dev) {
2809 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2810 				    dev->name);
2811 		return;
2812 	}
2813 
2814 	if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2815 		goto put;
2816 
2817 	/*
2818 	 *	Two things going on here:
2819 	 *	1) Add routes for on-link prefixes
2820 	 *	2) Configure prefixes with the auto flag set
2821 	 */
2822 
2823 	if (pinfo->onlink) {
2824 		struct fib6_info *rt;
2825 		unsigned long rt_expires;
2826 
2827 		/* Avoid arithmetic overflow. Really, we could
2828 		 * save rt_expires in seconds, likely valid_lft,
2829 		 * but it would require division in fib gc, that it
2830 		 * not good.
2831 		 */
2832 		if (HZ > USER_HZ)
2833 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2834 		else
2835 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2836 
2837 		if (addrconf_finite_timeout(rt_expires))
2838 			rt_expires *= HZ;
2839 
2840 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2841 					       pinfo->prefix_len,
2842 					       dev,
2843 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2844 					       RTF_DEFAULT, true);
2845 
2846 		if (rt) {
2847 			/* Autoconf prefix route */
2848 			if (valid_lft == 0) {
2849 				ip6_del_rt(net, rt, false);
2850 				rt = NULL;
2851 			} else {
2852 				table = rt->fib6_table;
2853 				spin_lock_bh(&table->tb6_lock);
2854 
2855 				if (addrconf_finite_timeout(rt_expires)) {
2856 					/* not infinity */
2857 					fib6_set_expires(rt, jiffies + rt_expires);
2858 					fib6_add_gc_list(rt);
2859 				} else {
2860 					fib6_clean_expires(rt);
2861 					fib6_remove_gc_list(rt);
2862 				}
2863 
2864 				spin_unlock_bh(&table->tb6_lock);
2865 			}
2866 		} else if (valid_lft) {
2867 			clock_t expires = 0;
2868 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2869 			if (addrconf_finite_timeout(rt_expires)) {
2870 				/* not infinity */
2871 				flags |= RTF_EXPIRES;
2872 				expires = jiffies_to_clock_t(rt_expires);
2873 			}
2874 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2875 					      0, dev, expires, flags,
2876 					      GFP_ATOMIC);
2877 		}
2878 		fib6_info_release(rt);
2879 	}
2880 
2881 	/* Try to figure out our local address for this prefix */
2882 
2883 	ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd;
2884 	if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) {
2885 		struct in6_addr addr;
2886 		bool tokenized = false, dev_addr_generated = false;
2887 
2888 		if (pinfo->prefix_len == 64) {
2889 			memcpy(&addr, &pinfo->prefix, 8);
2890 
2891 			if (!ipv6_addr_any(&in6_dev->token)) {
2892 				read_lock_bh(&in6_dev->lock);
2893 				memcpy(addr.s6_addr + 8,
2894 				       in6_dev->token.s6_addr + 8, 8);
2895 				read_unlock_bh(&in6_dev->lock);
2896 				tokenized = true;
2897 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2898 				   !ipv6_generate_stable_address(&addr, 0,
2899 								 in6_dev)) {
2900 				addr_flags |= IFA_F_STABLE_PRIVACY;
2901 				goto ok;
2902 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2903 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2904 				goto put;
2905 			} else {
2906 				dev_addr_generated = true;
2907 			}
2908 			goto ok;
2909 		}
2910 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2911 				    pinfo->prefix_len);
2912 		goto put;
2913 
2914 ok:
2915 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2916 						   &addr, addr_type,
2917 						   addr_flags, sllao,
2918 						   tokenized, valid_lft,
2919 						   prefered_lft);
2920 		if (err)
2921 			goto put;
2922 
2923 		/* Ignore error case here because previous prefix add addr was
2924 		 * successful which will be notified.
2925 		 */
2926 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2927 					      addr_type, addr_flags, sllao,
2928 					      tokenized, valid_lft,
2929 					      prefered_lft,
2930 					      dev_addr_generated);
2931 	}
2932 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2933 put:
2934 	in6_dev_put(in6_dev);
2935 }
2936 
2937 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2938 		struct in6_ifreq *ireq)
2939 {
2940 	struct ip_tunnel_parm_kern p = { };
2941 	int err;
2942 
2943 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2944 		return -EADDRNOTAVAIL;
2945 
2946 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2947 	p.iph.version = 4;
2948 	p.iph.ihl = 5;
2949 	p.iph.protocol = IPPROTO_IPV6;
2950 	p.iph.ttl = 64;
2951 
2952 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2953 		return -EOPNOTSUPP;
2954 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2955 	if (err)
2956 		return err;
2957 
2958 	dev = __dev_get_by_name(net, p.name);
2959 	if (!dev)
2960 		return -ENOBUFS;
2961 	return dev_open(dev, NULL);
2962 }
2963 
2964 /*
2965  *	Set destination address.
2966  *	Special case for SIT interfaces where we create a new "virtual"
2967  *	device.
2968  */
2969 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2970 {
2971 	struct net_device *dev;
2972 	struct in6_ifreq ireq;
2973 	int err = -ENODEV;
2974 
2975 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2976 		return -ENODEV;
2977 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2978 		return -EFAULT;
2979 
2980 	rtnl_net_lock(net);
2981 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2982 	if (dev && dev->type == ARPHRD_SIT)
2983 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2984 	rtnl_net_unlock(net);
2985 	return err;
2986 }
2987 
2988 static int ipv6_mc_config(struct sock *sk, bool join,
2989 			  const struct in6_addr *addr, int ifindex)
2990 {
2991 	int ret;
2992 
2993 	ASSERT_RTNL();
2994 
2995 	lock_sock(sk);
2996 	if (join)
2997 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2998 	else
2999 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
3000 	release_sock(sk);
3001 
3002 	return ret;
3003 }
3004 
3005 /*
3006  *	Manual configuration of address on an interface
3007  */
3008 static int inet6_addr_add(struct net *net, struct net_device *dev,
3009 			  struct ifa6_config *cfg, clock_t expires, u32 flags,
3010 			  struct netlink_ext_ack *extack)
3011 {
3012 	struct inet6_ifaddr *ifp;
3013 	struct inet6_dev *idev;
3014 
3015 	ASSERT_RTNL_NET(net);
3016 
3017 	if (cfg->plen > 128) {
3018 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3019 		return -EINVAL;
3020 	}
3021 
3022 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3023 		NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3024 		return -EINVAL;
3025 	}
3026 
3027 	idev = addrconf_add_dev(dev);
3028 	if (IS_ERR(idev)) {
3029 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3030 		return PTR_ERR(idev);
3031 	}
3032 
3033 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3034 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3035 					 true, cfg->pfx, dev->ifindex);
3036 
3037 		if (ret < 0) {
3038 			NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3039 			return ret;
3040 		}
3041 	}
3042 
3043 	cfg->scope = ipv6_addr_scope(cfg->pfx);
3044 
3045 	ifp = ipv6_add_addr(idev, cfg, true, extack);
3046 	if (!IS_ERR(ifp)) {
3047 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3048 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3049 					      ifp->rt_priority, dev, expires,
3050 					      flags, GFP_KERNEL);
3051 		}
3052 
3053 		/* Send a netlink notification if DAD is enabled and
3054 		 * optimistic flag is not set
3055 		 */
3056 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3057 			ipv6_ifa_notify(0, ifp);
3058 		/*
3059 		 * Note that section 3.1 of RFC 4429 indicates
3060 		 * that the Optimistic flag should not be set for
3061 		 * manually configured addresses
3062 		 */
3063 		addrconf_dad_start(ifp);
3064 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3065 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
3066 					 cfg->preferred_lft, true, jiffies);
3067 		in6_ifa_put(ifp);
3068 		addrconf_verify_rtnl(net);
3069 		return 0;
3070 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3071 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3072 			       cfg->pfx, dev->ifindex);
3073 	}
3074 
3075 	return PTR_ERR(ifp);
3076 }
3077 
3078 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3079 			  const struct in6_addr *pfx, unsigned int plen,
3080 			  struct netlink_ext_ack *extack)
3081 {
3082 	struct inet6_ifaddr *ifp;
3083 	struct inet6_dev *idev;
3084 	struct net_device *dev;
3085 
3086 	if (plen > 128) {
3087 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3088 		return -EINVAL;
3089 	}
3090 
3091 	dev = __dev_get_by_index(net, ifindex);
3092 	if (!dev) {
3093 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3094 		return -ENODEV;
3095 	}
3096 
3097 	idev = __in6_dev_get_rtnl_net(dev);
3098 	if (!idev) {
3099 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3100 		return -ENXIO;
3101 	}
3102 
3103 	read_lock_bh(&idev->lock);
3104 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3105 		if (ifp->prefix_len == plen &&
3106 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3107 			in6_ifa_hold(ifp);
3108 			read_unlock_bh(&idev->lock);
3109 
3110 			ipv6_del_addr(ifp);
3111 
3112 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3113 			    (ifp->flags & IFA_F_MANAGETEMPADDR))
3114 				delete_tempaddrs(idev, ifp);
3115 
3116 			addrconf_verify_rtnl(net);
3117 			if (ipv6_addr_is_multicast(pfx)) {
3118 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3119 					       false, pfx, dev->ifindex);
3120 			}
3121 			return 0;
3122 		}
3123 	}
3124 	read_unlock_bh(&idev->lock);
3125 
3126 	NL_SET_ERR_MSG_MOD(extack, "address not found");
3127 	return -EADDRNOTAVAIL;
3128 }
3129 
3130 
3131 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3132 {
3133 	struct ifa6_config cfg = {
3134 		.ifa_flags = IFA_F_PERMANENT,
3135 		.preferred_lft = INFINITY_LIFE_TIME,
3136 		.valid_lft = INFINITY_LIFE_TIME,
3137 	};
3138 	struct net_device *dev;
3139 	struct in6_ifreq ireq;
3140 	int err;
3141 
3142 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3143 		return -EPERM;
3144 
3145 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3146 		return -EFAULT;
3147 
3148 	cfg.pfx = &ireq.ifr6_addr;
3149 	cfg.plen = ireq.ifr6_prefixlen;
3150 
3151 	rtnl_net_lock(net);
3152 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
3153 	if (dev) {
3154 		netdev_lock_ops(dev);
3155 		err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL);
3156 		netdev_unlock_ops(dev);
3157 	} else {
3158 		err = -ENODEV;
3159 	}
3160 	rtnl_net_unlock(net);
3161 	return err;
3162 }
3163 
3164 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3165 {
3166 	struct in6_ifreq ireq;
3167 	int err;
3168 
3169 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3170 		return -EPERM;
3171 
3172 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3173 		return -EFAULT;
3174 
3175 	rtnl_net_lock(net);
3176 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3177 			     ireq.ifr6_prefixlen, NULL);
3178 	rtnl_net_unlock(net);
3179 	return err;
3180 }
3181 
3182 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3183 		     int plen, int scope, u8 proto)
3184 {
3185 	struct inet6_ifaddr *ifp;
3186 	struct ifa6_config cfg = {
3187 		.pfx = addr,
3188 		.plen = plen,
3189 		.ifa_flags = IFA_F_PERMANENT,
3190 		.valid_lft = INFINITY_LIFE_TIME,
3191 		.preferred_lft = INFINITY_LIFE_TIME,
3192 		.scope = scope,
3193 		.ifa_proto = proto
3194 	};
3195 
3196 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3197 	if (!IS_ERR(ifp)) {
3198 		spin_lock_bh(&ifp->lock);
3199 		ifp->flags &= ~IFA_F_TENTATIVE;
3200 		spin_unlock_bh(&ifp->lock);
3201 		rt_genid_bump_ipv6(dev_net(idev->dev));
3202 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3203 		in6_ifa_put(ifp);
3204 	}
3205 }
3206 
3207 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE)
3208 static void add_v4_addrs(struct inet6_dev *idev)
3209 {
3210 	struct in6_addr addr;
3211 	struct net_device *dev;
3212 	struct net *net = dev_net(idev->dev);
3213 	int scope, plen;
3214 	u32 pflags = 0;
3215 
3216 	ASSERT_RTNL();
3217 
3218 	memset(&addr, 0, sizeof(struct in6_addr));
3219 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3220 
3221 	if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3222 		scope = IPV6_ADDR_COMPATv4;
3223 		plen = 96;
3224 		pflags |= RTF_NONEXTHOP;
3225 	} else {
3226 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3227 			return;
3228 
3229 		addr.s6_addr32[0] = htonl(0xfe800000);
3230 		scope = IFA_LINK;
3231 		plen = 64;
3232 	}
3233 
3234 	if (addr.s6_addr32[3]) {
3235 		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3236 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3237 				      GFP_KERNEL);
3238 		return;
3239 	}
3240 
3241 	for_each_netdev(net, dev) {
3242 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3243 		if (in_dev && (dev->flags & IFF_UP)) {
3244 			struct in_ifaddr *ifa;
3245 			int flag = scope;
3246 
3247 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3248 				addr.s6_addr32[3] = ifa->ifa_local;
3249 
3250 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3251 					continue;
3252 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3253 					if (idev->dev->flags&IFF_POINTOPOINT)
3254 						continue;
3255 					flag |= IFA_HOST;
3256 				}
3257 
3258 				add_addr(idev, &addr, plen, flag,
3259 					 IFAPROT_UNSPEC);
3260 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3261 						      0, pflags, GFP_KERNEL);
3262 			}
3263 		}
3264 	}
3265 }
3266 #endif
3267 
3268 static void init_loopback(struct net_device *dev)
3269 {
3270 	struct inet6_dev  *idev;
3271 
3272 	/* ::1 */
3273 
3274 	ASSERT_RTNL();
3275 
3276 	idev = ipv6_find_idev(dev);
3277 	if (IS_ERR(idev)) {
3278 		pr_debug("%s: add_dev failed\n", __func__);
3279 		return;
3280 	}
3281 
3282 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3283 }
3284 
3285 void addrconf_add_linklocal(struct inet6_dev *idev,
3286 			    const struct in6_addr *addr, u32 flags)
3287 {
3288 	struct ifa6_config cfg = {
3289 		.pfx = addr,
3290 		.plen = 64,
3291 		.ifa_flags = flags | IFA_F_PERMANENT,
3292 		.valid_lft = INFINITY_LIFE_TIME,
3293 		.preferred_lft = INFINITY_LIFE_TIME,
3294 		.scope = IFA_LINK,
3295 		.ifa_proto = IFAPROT_KERNEL_LL
3296 	};
3297 	struct inet6_ifaddr *ifp;
3298 
3299 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3300 	if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3301 	     READ_ONCE(idev->cnf.optimistic_dad)) &&
3302 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3303 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3304 #endif
3305 
3306 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3307 	if (!IS_ERR(ifp)) {
3308 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3309 				      0, 0, GFP_ATOMIC);
3310 		addrconf_dad_start(ifp);
3311 		in6_ifa_put(ifp);
3312 	}
3313 }
3314 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3315 
3316 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3317 {
3318 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3319 		return true;
3320 
3321 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3322 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3323 		return true;
3324 
3325 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3326 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3327 		return true;
3328 
3329 	return false;
3330 }
3331 
3332 static int ipv6_generate_stable_address(struct in6_addr *address,
3333 					u8 dad_count,
3334 					const struct inet6_dev *idev)
3335 {
3336 	static DEFINE_SPINLOCK(lock);
3337 	static __u32 digest[SHA1_DIGEST_WORDS];
3338 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3339 
3340 	static union {
3341 		char __data[SHA1_BLOCK_SIZE];
3342 		struct {
3343 			struct in6_addr secret;
3344 			__be32 prefix[2];
3345 			unsigned char hwaddr[MAX_ADDR_LEN];
3346 			u8 dad_count;
3347 		} __packed;
3348 	} data;
3349 
3350 	struct in6_addr secret;
3351 	struct in6_addr temp;
3352 	struct net *net = dev_net(idev->dev);
3353 
3354 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3355 
3356 	if (idev->cnf.stable_secret.initialized)
3357 		secret = idev->cnf.stable_secret.secret;
3358 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3359 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3360 	else
3361 		return -1;
3362 
3363 retry:
3364 	spin_lock_bh(&lock);
3365 
3366 	sha1_init(digest);
3367 	memset(&data, 0, sizeof(data));
3368 	memset(workspace, 0, sizeof(workspace));
3369 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3370 	data.prefix[0] = address->s6_addr32[0];
3371 	data.prefix[1] = address->s6_addr32[1];
3372 	data.secret = secret;
3373 	data.dad_count = dad_count;
3374 
3375 	sha1_transform(digest, data.__data, workspace);
3376 
3377 	temp = *address;
3378 	temp.s6_addr32[2] = (__force __be32)digest[0];
3379 	temp.s6_addr32[3] = (__force __be32)digest[1];
3380 
3381 	spin_unlock_bh(&lock);
3382 
3383 	if (ipv6_reserved_interfaceid(temp)) {
3384 		dad_count++;
3385 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3386 			return -1;
3387 		goto retry;
3388 	}
3389 
3390 	*address = temp;
3391 	return 0;
3392 }
3393 
3394 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3395 {
3396 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3397 
3398 	if (s->initialized)
3399 		return;
3400 	s = &idev->cnf.stable_secret;
3401 	get_random_bytes(&s->secret, sizeof(s->secret));
3402 	s->initialized = true;
3403 }
3404 
3405 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3406 {
3407 	struct in6_addr addr;
3408 
3409 	/* no link local addresses on L3 master devices */
3410 	if (netif_is_l3_master(idev->dev))
3411 		return;
3412 
3413 	/* no link local addresses on devices flagged as slaves */
3414 	if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3415 		return;
3416 
3417 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3418 
3419 	switch (idev->cnf.addr_gen_mode) {
3420 	case IN6_ADDR_GEN_MODE_RANDOM:
3421 		ipv6_gen_mode_random_init(idev);
3422 		fallthrough;
3423 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3424 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3425 			addrconf_add_linklocal(idev, &addr,
3426 					       IFA_F_STABLE_PRIVACY);
3427 		else if (prefix_route)
3428 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3429 					      0, 0, GFP_KERNEL);
3430 		break;
3431 	case IN6_ADDR_GEN_MODE_EUI64:
3432 		/* addrconf_add_linklocal also adds a prefix_route and we
3433 		 * only need to care about prefix routes if ipv6_generate_eui64
3434 		 * couldn't generate one.
3435 		 */
3436 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3437 			addrconf_add_linklocal(idev, &addr, 0);
3438 		else if (prefix_route)
3439 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3440 					      0, 0, GFP_KERNEL);
3441 		break;
3442 	case IN6_ADDR_GEN_MODE_NONE:
3443 	default:
3444 		/* will not add any link local address */
3445 		break;
3446 	}
3447 }
3448 
3449 static void addrconf_dev_config(struct net_device *dev)
3450 {
3451 	struct inet6_dev *idev;
3452 
3453 	ASSERT_RTNL();
3454 
3455 	if ((dev->type != ARPHRD_ETHER) &&
3456 	    (dev->type != ARPHRD_FDDI) &&
3457 	    (dev->type != ARPHRD_ARCNET) &&
3458 	    (dev->type != ARPHRD_INFINIBAND) &&
3459 	    (dev->type != ARPHRD_IEEE1394) &&
3460 	    (dev->type != ARPHRD_TUNNEL6) &&
3461 	    (dev->type != ARPHRD_6LOWPAN) &&
3462 	    (dev->type != ARPHRD_IP6GRE) &&
3463 	    (dev->type != ARPHRD_TUNNEL) &&
3464 	    (dev->type != ARPHRD_NONE) &&
3465 	    (dev->type != ARPHRD_RAWIP)) {
3466 		/* Alas, we support only Ethernet autoconfiguration. */
3467 		idev = __in6_dev_get(dev);
3468 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3469 		    dev->flags & IFF_MULTICAST)
3470 			ipv6_mc_up(idev);
3471 		return;
3472 	}
3473 
3474 	idev = addrconf_add_dev(dev);
3475 	if (IS_ERR(idev))
3476 		return;
3477 
3478 	/* this device type has no EUI support */
3479 	if (dev->type == ARPHRD_NONE &&
3480 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3481 		WRITE_ONCE(idev->cnf.addr_gen_mode,
3482 			   IN6_ADDR_GEN_MODE_RANDOM);
3483 
3484 	addrconf_addr_gen(idev, false);
3485 }
3486 
3487 #if IS_ENABLED(CONFIG_IPV6_SIT)
3488 static void addrconf_sit_config(struct net_device *dev)
3489 {
3490 	struct inet6_dev *idev;
3491 
3492 	ASSERT_RTNL();
3493 
3494 	/*
3495 	 * Configure the tunnel with one of our IPv4
3496 	 * addresses... we should configure all of
3497 	 * our v4 addrs in the tunnel
3498 	 */
3499 
3500 	idev = ipv6_find_idev(dev);
3501 	if (IS_ERR(idev)) {
3502 		pr_debug("%s: add_dev failed\n", __func__);
3503 		return;
3504 	}
3505 
3506 	if (dev->priv_flags & IFF_ISATAP) {
3507 		addrconf_addr_gen(idev, false);
3508 		return;
3509 	}
3510 
3511 	add_v4_addrs(idev);
3512 
3513 	if (dev->flags&IFF_POINTOPOINT)
3514 		addrconf_add_mroute(dev);
3515 }
3516 #endif
3517 
3518 #if IS_ENABLED(CONFIG_NET_IPGRE)
3519 static void addrconf_gre_config(struct net_device *dev)
3520 {
3521 	struct inet6_dev *idev;
3522 
3523 	ASSERT_RTNL();
3524 
3525 	idev = addrconf_add_dev(dev);
3526 	if (IS_ERR(idev))
3527 		return;
3528 
3529 	/* Generate the IPv6 link-local address using addrconf_addr_gen(),
3530 	 * unless we have an IPv4 GRE device not bound to an IP address and
3531 	 * which is in EUI64 mode (as __ipv6_isatap_ifid() would fail in this
3532 	 * case). Such devices fall back to add_v4_addrs() instead.
3533 	 */
3534 	if (!(*(__be32 *)dev->dev_addr == 0 &&
3535 	      idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)) {
3536 		addrconf_addr_gen(idev, true);
3537 		return;
3538 	}
3539 
3540 	add_v4_addrs(idev);
3541 }
3542 #endif
3543 
3544 static void addrconf_init_auto_addrs(struct net_device *dev)
3545 {
3546 	switch (dev->type) {
3547 #if IS_ENABLED(CONFIG_IPV6_SIT)
3548 	case ARPHRD_SIT:
3549 		addrconf_sit_config(dev);
3550 		break;
3551 #endif
3552 #if IS_ENABLED(CONFIG_NET_IPGRE)
3553 	case ARPHRD_IPGRE:
3554 		addrconf_gre_config(dev);
3555 		break;
3556 #endif
3557 	case ARPHRD_LOOPBACK:
3558 		init_loopback(dev);
3559 		break;
3560 
3561 	default:
3562 		addrconf_dev_config(dev);
3563 		break;
3564 	}
3565 }
3566 
3567 static int fixup_permanent_addr(struct net *net,
3568 				struct inet6_dev *idev,
3569 				struct inet6_ifaddr *ifp)
3570 {
3571 	/* !fib6_node means the host route was removed from the
3572 	 * FIB, for example, if 'lo' device is taken down. In that
3573 	 * case regenerate the host route.
3574 	 */
3575 	if (!ifp->rt || !ifp->rt->fib6_node) {
3576 		struct fib6_info *f6i, *prev;
3577 
3578 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3579 					 GFP_ATOMIC, NULL);
3580 		if (IS_ERR(f6i))
3581 			return PTR_ERR(f6i);
3582 
3583 		/* ifp->rt can be accessed outside of rtnl */
3584 		spin_lock(&ifp->lock);
3585 		prev = ifp->rt;
3586 		ifp->rt = f6i;
3587 		spin_unlock(&ifp->lock);
3588 
3589 		fib6_info_release(prev);
3590 	}
3591 
3592 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3593 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3594 				      ifp->rt_priority, idev->dev, 0, 0,
3595 				      GFP_ATOMIC);
3596 	}
3597 
3598 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3599 		addrconf_dad_start(ifp);
3600 
3601 	return 0;
3602 }
3603 
3604 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3605 {
3606 	struct inet6_ifaddr *ifp, *tmp;
3607 	struct inet6_dev *idev;
3608 
3609 	idev = __in6_dev_get(dev);
3610 	if (!idev)
3611 		return;
3612 
3613 	write_lock_bh(&idev->lock);
3614 
3615 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3616 		if ((ifp->flags & IFA_F_PERMANENT) &&
3617 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3618 			write_unlock_bh(&idev->lock);
3619 			in6_ifa_hold(ifp);
3620 			ipv6_del_addr(ifp);
3621 			write_lock_bh(&idev->lock);
3622 
3623 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3624 					     idev->dev->name, &ifp->addr);
3625 		}
3626 	}
3627 
3628 	write_unlock_bh(&idev->lock);
3629 }
3630 
3631 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3632 			   void *ptr)
3633 {
3634 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3635 	struct netdev_notifier_change_info *change_info;
3636 	struct netdev_notifier_changeupper_info *info;
3637 	struct inet6_dev *idev = __in6_dev_get(dev);
3638 	struct net *net = dev_net(dev);
3639 	int run_pending = 0;
3640 	int err;
3641 
3642 	switch (event) {
3643 	case NETDEV_REGISTER:
3644 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3645 			idev = ipv6_add_dev(dev);
3646 			if (IS_ERR(idev))
3647 				return notifier_from_errno(PTR_ERR(idev));
3648 		}
3649 		break;
3650 
3651 	case NETDEV_CHANGEMTU:
3652 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3653 		if (dev->mtu < IPV6_MIN_MTU) {
3654 			addrconf_ifdown(dev, dev != net->loopback_dev);
3655 			break;
3656 		}
3657 
3658 		if (idev) {
3659 			rt6_mtu_change(dev, dev->mtu);
3660 			WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3661 			break;
3662 		}
3663 
3664 		/* allocate new idev */
3665 		idev = ipv6_add_dev(dev);
3666 		if (IS_ERR(idev))
3667 			break;
3668 
3669 		/* device is still not ready */
3670 		if (!(idev->if_flags & IF_READY))
3671 			break;
3672 
3673 		run_pending = 1;
3674 		fallthrough;
3675 	case NETDEV_UP:
3676 	case NETDEV_CHANGE:
3677 		if (idev && idev->cnf.disable_ipv6)
3678 			break;
3679 
3680 		if (dev->priv_flags & IFF_NO_ADDRCONF) {
3681 			if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3682 			    dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3683 				ipv6_mc_up(idev);
3684 			break;
3685 		}
3686 
3687 		if (event == NETDEV_UP) {
3688 			/* restore routes for permanent addresses */
3689 			addrconf_permanent_addr(net, dev);
3690 
3691 			if (!addrconf_link_ready(dev)) {
3692 				/* device is not ready yet. */
3693 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3694 					 dev->name);
3695 				break;
3696 			}
3697 
3698 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3699 				idev = ipv6_add_dev(dev);
3700 
3701 			if (!IS_ERR_OR_NULL(idev)) {
3702 				idev->if_flags |= IF_READY;
3703 				run_pending = 1;
3704 			}
3705 		} else if (event == NETDEV_CHANGE) {
3706 			if (!addrconf_link_ready(dev)) {
3707 				/* device is still not ready. */
3708 				rt6_sync_down_dev(dev, event);
3709 				break;
3710 			}
3711 
3712 			if (!IS_ERR_OR_NULL(idev)) {
3713 				if (idev->if_flags & IF_READY) {
3714 					/* device is already configured -
3715 					 * but resend MLD reports, we might
3716 					 * have roamed and need to update
3717 					 * multicast snooping switches
3718 					 */
3719 					ipv6_mc_up(idev);
3720 					change_info = ptr;
3721 					if (change_info->flags_changed & IFF_NOARP)
3722 						addrconf_dad_run(idev, true);
3723 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3724 					break;
3725 				}
3726 				idev->if_flags |= IF_READY;
3727 			}
3728 
3729 			pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3730 				 dev->name);
3731 
3732 			run_pending = 1;
3733 		}
3734 
3735 		addrconf_init_auto_addrs(dev);
3736 
3737 		if (!IS_ERR_OR_NULL(idev)) {
3738 			if (run_pending)
3739 				addrconf_dad_run(idev, false);
3740 
3741 			/* Device has an address by now */
3742 			rt6_sync_up(dev, RTNH_F_DEAD);
3743 
3744 			/*
3745 			 * If the MTU changed during the interface down,
3746 			 * when the interface up, the changed MTU must be
3747 			 * reflected in the idev as well as routers.
3748 			 */
3749 			if (idev->cnf.mtu6 != dev->mtu &&
3750 			    dev->mtu >= IPV6_MIN_MTU) {
3751 				rt6_mtu_change(dev, dev->mtu);
3752 				WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3753 			}
3754 			WRITE_ONCE(idev->tstamp, jiffies);
3755 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3756 
3757 			/*
3758 			 * If the changed mtu during down is lower than
3759 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3760 			 */
3761 			if (dev->mtu < IPV6_MIN_MTU)
3762 				addrconf_ifdown(dev, dev != net->loopback_dev);
3763 		}
3764 		break;
3765 
3766 	case NETDEV_DOWN:
3767 	case NETDEV_UNREGISTER:
3768 		/*
3769 		 *	Remove all addresses from this interface.
3770 		 */
3771 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3772 		break;
3773 
3774 	case NETDEV_CHANGENAME:
3775 		if (idev) {
3776 			snmp6_unregister_dev(idev);
3777 			addrconf_sysctl_unregister(idev);
3778 			err = addrconf_sysctl_register(idev);
3779 			if (err)
3780 				return notifier_from_errno(err);
3781 			err = snmp6_register_dev(idev);
3782 			if (err) {
3783 				addrconf_sysctl_unregister(idev);
3784 				return notifier_from_errno(err);
3785 			}
3786 		}
3787 		break;
3788 
3789 	case NETDEV_PRE_TYPE_CHANGE:
3790 	case NETDEV_POST_TYPE_CHANGE:
3791 		if (idev)
3792 			addrconf_type_change(dev, event);
3793 		break;
3794 
3795 	case NETDEV_CHANGEUPPER:
3796 		info = ptr;
3797 
3798 		/* flush all routes if dev is linked to or unlinked from
3799 		 * an L3 master device (e.g., VRF)
3800 		 */
3801 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3802 			addrconf_ifdown(dev, false);
3803 	}
3804 
3805 	return NOTIFY_OK;
3806 }
3807 
3808 /*
3809  *	addrconf module should be notified of a device going up
3810  */
3811 static struct notifier_block ipv6_dev_notf = {
3812 	.notifier_call = addrconf_notify,
3813 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3814 };
3815 
3816 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3817 {
3818 	struct inet6_dev *idev;
3819 	ASSERT_RTNL();
3820 
3821 	idev = __in6_dev_get(dev);
3822 
3823 	if (event == NETDEV_POST_TYPE_CHANGE)
3824 		ipv6_mc_remap(idev);
3825 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3826 		ipv6_mc_unmap(idev);
3827 }
3828 
3829 static bool addr_is_local(const struct in6_addr *addr)
3830 {
3831 	return ipv6_addr_type(addr) &
3832 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3833 }
3834 
3835 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3836 {
3837 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3838 	struct net *net = dev_net(dev);
3839 	struct inet6_dev *idev;
3840 	struct inet6_ifaddr *ifa;
3841 	LIST_HEAD(tmp_addr_list);
3842 	bool keep_addr = false;
3843 	bool was_ready;
3844 	int state, i;
3845 
3846 	ASSERT_RTNL();
3847 
3848 	rt6_disable_ip(dev, event);
3849 
3850 	idev = __in6_dev_get(dev);
3851 	if (!idev)
3852 		return -ENODEV;
3853 
3854 	/*
3855 	 * Step 1: remove reference to ipv6 device from parent device.
3856 	 *	   Do not dev_put!
3857 	 */
3858 	if (unregister) {
3859 		WRITE_ONCE(idev->dead, 1);
3860 
3861 		/* protected by rtnl_lock */
3862 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3863 
3864 		/* Step 1.5: remove snmp6 entry */
3865 		snmp6_unregister_dev(idev);
3866 
3867 	}
3868 
3869 	/* combine the user config with event to determine if permanent
3870 	 * addresses are to be removed from address hash table
3871 	 */
3872 	if (!unregister && !idev->cnf.disable_ipv6) {
3873 		/* aggregate the system setting and interface setting */
3874 		int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3875 
3876 		if (!_keep_addr)
3877 			_keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3878 
3879 		keep_addr = (_keep_addr > 0);
3880 	}
3881 
3882 	/* Step 2: clear hash table */
3883 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3884 		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3885 
3886 		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3887 restart:
3888 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3889 			if (ifa->idev == idev) {
3890 				addrconf_del_dad_work(ifa);
3891 				/* combined flag + permanent flag decide if
3892 				 * address is retained on a down event
3893 				 */
3894 				if (!keep_addr ||
3895 				    !(ifa->flags & IFA_F_PERMANENT) ||
3896 				    addr_is_local(&ifa->addr)) {
3897 					hlist_del_init_rcu(&ifa->addr_lst);
3898 					goto restart;
3899 				}
3900 			}
3901 		}
3902 		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3903 	}
3904 
3905 	write_lock_bh(&idev->lock);
3906 
3907 	addrconf_del_rs_timer(idev);
3908 
3909 	/* Step 2: clear flags for stateless addrconf, repeated down
3910 	 *         detection
3911 	 */
3912 	was_ready = idev->if_flags & IF_READY;
3913 	if (!unregister)
3914 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3915 
3916 	/* Step 3: clear tempaddr list */
3917 	while (!list_empty(&idev->tempaddr_list)) {
3918 		ifa = list_first_entry(&idev->tempaddr_list,
3919 				       struct inet6_ifaddr, tmp_list);
3920 		list_del(&ifa->tmp_list);
3921 		write_unlock_bh(&idev->lock);
3922 		spin_lock_bh(&ifa->lock);
3923 
3924 		if (ifa->ifpub) {
3925 			in6_ifa_put(ifa->ifpub);
3926 			ifa->ifpub = NULL;
3927 		}
3928 		spin_unlock_bh(&ifa->lock);
3929 		in6_ifa_put(ifa);
3930 		write_lock_bh(&idev->lock);
3931 	}
3932 
3933 	list_for_each_entry(ifa, &idev->addr_list, if_list)
3934 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3935 	write_unlock_bh(&idev->lock);
3936 
3937 	while (!list_empty(&tmp_addr_list)) {
3938 		struct fib6_info *rt = NULL;
3939 		bool keep;
3940 
3941 		ifa = list_first_entry(&tmp_addr_list,
3942 				       struct inet6_ifaddr, if_list_aux);
3943 		list_del(&ifa->if_list_aux);
3944 
3945 		addrconf_del_dad_work(ifa);
3946 
3947 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3948 			!addr_is_local(&ifa->addr);
3949 
3950 		spin_lock_bh(&ifa->lock);
3951 
3952 		if (keep) {
3953 			/* set state to skip the notifier below */
3954 			state = INET6_IFADDR_STATE_DEAD;
3955 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3956 			if (!(ifa->flags & IFA_F_NODAD))
3957 				ifa->flags |= IFA_F_TENTATIVE;
3958 
3959 			rt = ifa->rt;
3960 			ifa->rt = NULL;
3961 		} else {
3962 			state = ifa->state;
3963 			ifa->state = INET6_IFADDR_STATE_DEAD;
3964 		}
3965 
3966 		spin_unlock_bh(&ifa->lock);
3967 
3968 		if (rt)
3969 			ip6_del_rt(net, rt, false);
3970 
3971 		if (state != INET6_IFADDR_STATE_DEAD) {
3972 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3973 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3974 		} else {
3975 			if (idev->cnf.forwarding)
3976 				addrconf_leave_anycast(ifa);
3977 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3978 		}
3979 
3980 		if (!keep) {
3981 			write_lock_bh(&idev->lock);
3982 			list_del_rcu(&ifa->if_list);
3983 			write_unlock_bh(&idev->lock);
3984 			in6_ifa_put(ifa);
3985 		}
3986 	}
3987 
3988 	/* Step 5: Discard anycast and multicast list */
3989 	if (unregister) {
3990 		ipv6_ac_destroy_dev(idev);
3991 		ipv6_mc_destroy_dev(idev);
3992 	} else if (was_ready) {
3993 		ipv6_mc_down(idev);
3994 	}
3995 
3996 	WRITE_ONCE(idev->tstamp, jiffies);
3997 	idev->ra_mtu = 0;
3998 
3999 	/* Last: Shot the device (if unregistered) */
4000 	if (unregister) {
4001 		addrconf_sysctl_unregister(idev);
4002 		neigh_parms_release(&nd_tbl, idev->nd_parms);
4003 		neigh_ifdown(&nd_tbl, dev);
4004 		in6_dev_put(idev);
4005 	}
4006 	return 0;
4007 }
4008 
4009 static void addrconf_rs_timer(struct timer_list *t)
4010 {
4011 	struct inet6_dev *idev = timer_container_of(idev, t, rs_timer);
4012 	struct net_device *dev = idev->dev;
4013 	struct in6_addr lladdr;
4014 	int rtr_solicits;
4015 
4016 	write_lock(&idev->lock);
4017 	if (idev->dead || !(idev->if_flags & IF_READY))
4018 		goto out;
4019 
4020 	if (!ipv6_accept_ra(idev))
4021 		goto out;
4022 
4023 	/* Announcement received after solicitation was sent */
4024 	if (idev->if_flags & IF_RA_RCVD)
4025 		goto out;
4026 
4027 	rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4028 
4029 	if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4030 		write_unlock(&idev->lock);
4031 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4032 			ndisc_send_rs(dev, &lladdr,
4033 				      &in6addr_linklocal_allrouters);
4034 		else
4035 			goto put;
4036 
4037 		write_lock(&idev->lock);
4038 		idev->rs_interval = rfc3315_s14_backoff_update(
4039 				idev->rs_interval,
4040 				READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4041 		/* The wait after the last probe can be shorter */
4042 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4043 					     READ_ONCE(idev->cnf.rtr_solicits)) ?
4044 				      READ_ONCE(idev->cnf.rtr_solicit_delay) :
4045 				      idev->rs_interval);
4046 	} else {
4047 		/*
4048 		 * Note: we do not support deprecated "all on-link"
4049 		 * assumption any longer.
4050 		 */
4051 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4052 	}
4053 
4054 out:
4055 	write_unlock(&idev->lock);
4056 put:
4057 	in6_dev_put(idev);
4058 }
4059 
4060 /*
4061  *	Duplicate Address Detection
4062  */
4063 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4064 {
4065 	struct inet6_dev *idev = ifp->idev;
4066 	unsigned long rand_num;
4067 	u64 nonce;
4068 
4069 	if (ifp->flags & IFA_F_OPTIMISTIC)
4070 		rand_num = 0;
4071 	else
4072 		rand_num = get_random_u32_below(
4073 				READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4074 
4075 	nonce = 0;
4076 	if (READ_ONCE(idev->cnf.enhanced_dad) ||
4077 	    READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4078 		do
4079 			get_random_bytes(&nonce, 6);
4080 		while (nonce == 0);
4081 	}
4082 	ifp->dad_nonce = nonce;
4083 	ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4084 	addrconf_mod_dad_work(ifp, rand_num);
4085 }
4086 
4087 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4088 {
4089 	struct inet6_dev *idev = ifp->idev;
4090 	struct net_device *dev = idev->dev;
4091 	bool bump_id, notify = false;
4092 	struct net *net;
4093 
4094 	addrconf_join_solict(dev, &ifp->addr);
4095 
4096 	read_lock_bh(&idev->lock);
4097 	spin_lock(&ifp->lock);
4098 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
4099 		goto out;
4100 
4101 	net = dev_net(dev);
4102 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4103 	    (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4104 	     READ_ONCE(idev->cnf.accept_dad) < 1) ||
4105 	    !(ifp->flags&IFA_F_TENTATIVE) ||
4106 	    ifp->flags & IFA_F_NODAD) {
4107 		bool send_na = false;
4108 
4109 		if (ifp->flags & IFA_F_TENTATIVE &&
4110 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4111 			send_na = true;
4112 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4113 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4114 		spin_unlock(&ifp->lock);
4115 		read_unlock_bh(&idev->lock);
4116 
4117 		addrconf_dad_completed(ifp, bump_id, send_na);
4118 		return;
4119 	}
4120 
4121 	if (!(idev->if_flags & IF_READY)) {
4122 		spin_unlock(&ifp->lock);
4123 		read_unlock_bh(&idev->lock);
4124 		/*
4125 		 * If the device is not ready:
4126 		 * - keep it tentative if it is a permanent address.
4127 		 * - otherwise, kill it.
4128 		 */
4129 		in6_ifa_hold(ifp);
4130 		addrconf_dad_stop(ifp, 0);
4131 		return;
4132 	}
4133 
4134 	/*
4135 	 * Optimistic nodes can start receiving
4136 	 * Frames right away
4137 	 */
4138 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4139 		ip6_ins_rt(net, ifp->rt);
4140 		if (ipv6_use_optimistic_addr(net, idev)) {
4141 			/* Because optimistic nodes can use this address,
4142 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4143 			 */
4144 			notify = true;
4145 		}
4146 	}
4147 
4148 	addrconf_dad_kick(ifp);
4149 out:
4150 	spin_unlock(&ifp->lock);
4151 	read_unlock_bh(&idev->lock);
4152 	if (notify)
4153 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4154 }
4155 
4156 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4157 {
4158 	bool begin_dad = false;
4159 
4160 	spin_lock_bh(&ifp->lock);
4161 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4162 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4163 		begin_dad = true;
4164 	}
4165 	spin_unlock_bh(&ifp->lock);
4166 
4167 	if (begin_dad)
4168 		addrconf_mod_dad_work(ifp, 0);
4169 }
4170 
4171 static void addrconf_dad_work(struct work_struct *w)
4172 {
4173 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4174 						struct inet6_ifaddr,
4175 						dad_work);
4176 	struct inet6_dev *idev = ifp->idev;
4177 	bool bump_id, disable_ipv6 = false;
4178 	struct in6_addr mcaddr;
4179 	struct net *net;
4180 
4181 	enum {
4182 		DAD_PROCESS,
4183 		DAD_BEGIN,
4184 		DAD_ABORT,
4185 	} action = DAD_PROCESS;
4186 
4187 	net = dev_net(idev->dev);
4188 
4189 	rtnl_net_lock(net);
4190 
4191 	spin_lock_bh(&ifp->lock);
4192 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4193 		action = DAD_BEGIN;
4194 		ifp->state = INET6_IFADDR_STATE_DAD;
4195 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4196 		action = DAD_ABORT;
4197 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4198 
4199 		if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 ||
4200 		     READ_ONCE(idev->cnf.accept_dad) > 1) &&
4201 		    !idev->cnf.disable_ipv6 &&
4202 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4203 			struct in6_addr addr;
4204 
4205 			addr.s6_addr32[0] = htonl(0xfe800000);
4206 			addr.s6_addr32[1] = 0;
4207 
4208 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4209 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4210 				/* DAD failed for link-local based on MAC */
4211 				WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4212 
4213 				pr_info("%s: IPv6 being disabled!\n",
4214 					ifp->idev->dev->name);
4215 				disable_ipv6 = true;
4216 			}
4217 		}
4218 	}
4219 	spin_unlock_bh(&ifp->lock);
4220 
4221 	if (action == DAD_BEGIN) {
4222 		addrconf_dad_begin(ifp);
4223 		goto out;
4224 	} else if (action == DAD_ABORT) {
4225 		in6_ifa_hold(ifp);
4226 		addrconf_dad_stop(ifp, 1);
4227 		if (disable_ipv6)
4228 			addrconf_ifdown(idev->dev, false);
4229 		goto out;
4230 	}
4231 
4232 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4233 		goto out;
4234 
4235 	write_lock_bh(&idev->lock);
4236 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4237 		write_unlock_bh(&idev->lock);
4238 		goto out;
4239 	}
4240 
4241 	spin_lock(&ifp->lock);
4242 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4243 		spin_unlock(&ifp->lock);
4244 		write_unlock_bh(&idev->lock);
4245 		goto out;
4246 	}
4247 
4248 	if (ifp->dad_probes == 0) {
4249 		bool send_na = false;
4250 
4251 		/*
4252 		 * DAD was successful
4253 		 */
4254 
4255 		if (ifp->flags & IFA_F_TENTATIVE &&
4256 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4257 			send_na = true;
4258 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4259 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4260 		spin_unlock(&ifp->lock);
4261 		write_unlock_bh(&idev->lock);
4262 
4263 		addrconf_dad_completed(ifp, bump_id, send_na);
4264 
4265 		goto out;
4266 	}
4267 
4268 	ifp->dad_probes--;
4269 	addrconf_mod_dad_work(ifp,
4270 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4271 				  HZ/100));
4272 	spin_unlock(&ifp->lock);
4273 	write_unlock_bh(&idev->lock);
4274 
4275 	/* send a neighbour solicitation for our addr */
4276 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4277 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4278 		      ifp->dad_nonce);
4279 out:
4280 	in6_ifa_put(ifp);
4281 	rtnl_net_unlock(net);
4282 }
4283 
4284 /* ifp->idev must be at least read locked */
4285 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4286 {
4287 	struct inet6_ifaddr *ifpiter;
4288 	struct inet6_dev *idev = ifp->idev;
4289 
4290 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4291 		if (ifpiter->scope > IFA_LINK)
4292 			break;
4293 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4294 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4295 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4296 		    IFA_F_PERMANENT)
4297 			return false;
4298 	}
4299 	return true;
4300 }
4301 
4302 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4303 				   bool send_na)
4304 {
4305 	struct net_device *dev = ifp->idev->dev;
4306 	struct in6_addr lladdr;
4307 	bool send_rs, send_mld;
4308 
4309 	addrconf_del_dad_work(ifp);
4310 
4311 	/*
4312 	 *	Configure the address for reception. Now it is valid.
4313 	 */
4314 
4315 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4316 
4317 	/* If added prefix is link local and we are prepared to process
4318 	   router advertisements, start sending router solicitations.
4319 	 */
4320 
4321 	read_lock_bh(&ifp->idev->lock);
4322 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4323 	send_rs = send_mld &&
4324 		  ipv6_accept_ra(ifp->idev) &&
4325 		  READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4326 		  (dev->flags & IFF_LOOPBACK) == 0 &&
4327 		  (dev->type != ARPHRD_TUNNEL) &&
4328 		  !netif_is_team_port(dev);
4329 	read_unlock_bh(&ifp->idev->lock);
4330 
4331 	/* While dad is in progress mld report's source address is in6_addrany.
4332 	 * Resend with proper ll now.
4333 	 */
4334 	if (send_mld)
4335 		ipv6_mc_dad_complete(ifp->idev);
4336 
4337 	/* send unsolicited NA if enabled */
4338 	if (send_na &&
4339 	    (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4340 	     READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4341 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4342 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4343 			      /*solicited=*/ false, /*override=*/ true,
4344 			      /*inc_opt=*/ true);
4345 	}
4346 
4347 	if (send_rs) {
4348 		/*
4349 		 *	If a host as already performed a random delay
4350 		 *	[...] as part of DAD [...] there is no need
4351 		 *	to delay again before sending the first RS
4352 		 */
4353 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4354 			return;
4355 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4356 
4357 		write_lock_bh(&ifp->idev->lock);
4358 		spin_lock(&ifp->lock);
4359 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4360 			READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4361 		ifp->idev->rs_probes = 1;
4362 		ifp->idev->if_flags |= IF_RS_SENT;
4363 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4364 		spin_unlock(&ifp->lock);
4365 		write_unlock_bh(&ifp->idev->lock);
4366 	}
4367 
4368 	if (bump_id)
4369 		rt_genid_bump_ipv6(dev_net(dev));
4370 
4371 	/* Make sure that a new temporary address will be created
4372 	 * before this temporary address becomes deprecated.
4373 	 */
4374 	if (ifp->flags & IFA_F_TEMPORARY)
4375 		addrconf_verify_rtnl(dev_net(dev));
4376 }
4377 
4378 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4379 {
4380 	struct inet6_ifaddr *ifp;
4381 
4382 	read_lock_bh(&idev->lock);
4383 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4384 		spin_lock(&ifp->lock);
4385 		if ((ifp->flags & IFA_F_TENTATIVE &&
4386 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4387 			if (restart)
4388 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4389 			addrconf_dad_kick(ifp);
4390 		}
4391 		spin_unlock(&ifp->lock);
4392 	}
4393 	read_unlock_bh(&idev->lock);
4394 }
4395 
4396 #ifdef CONFIG_PROC_FS
4397 struct if6_iter_state {
4398 	struct seq_net_private p;
4399 	int bucket;
4400 	int offset;
4401 };
4402 
4403 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4404 {
4405 	struct if6_iter_state *state = seq->private;
4406 	struct net *net = seq_file_net(seq);
4407 	struct inet6_ifaddr *ifa = NULL;
4408 	int p = 0;
4409 
4410 	/* initial bucket if pos is 0 */
4411 	if (pos == 0) {
4412 		state->bucket = 0;
4413 		state->offset = 0;
4414 	}
4415 
4416 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4417 		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4418 					 addr_lst) {
4419 			/* sync with offset */
4420 			if (p < state->offset) {
4421 				p++;
4422 				continue;
4423 			}
4424 			return ifa;
4425 		}
4426 
4427 		/* prepare for next bucket */
4428 		state->offset = 0;
4429 		p = 0;
4430 	}
4431 	return NULL;
4432 }
4433 
4434 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4435 					 struct inet6_ifaddr *ifa)
4436 {
4437 	struct if6_iter_state *state = seq->private;
4438 	struct net *net = seq_file_net(seq);
4439 
4440 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4441 		state->offset++;
4442 		return ifa;
4443 	}
4444 
4445 	state->offset = 0;
4446 	while (++state->bucket < IN6_ADDR_HSIZE) {
4447 		hlist_for_each_entry_rcu(ifa,
4448 				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4449 			return ifa;
4450 		}
4451 	}
4452 
4453 	return NULL;
4454 }
4455 
4456 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4457 	__acquires(rcu)
4458 {
4459 	rcu_read_lock();
4460 	return if6_get_first(seq, *pos);
4461 }
4462 
4463 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4464 {
4465 	struct inet6_ifaddr *ifa;
4466 
4467 	ifa = if6_get_next(seq, v);
4468 	++*pos;
4469 	return ifa;
4470 }
4471 
4472 static void if6_seq_stop(struct seq_file *seq, void *v)
4473 	__releases(rcu)
4474 {
4475 	rcu_read_unlock();
4476 }
4477 
4478 static int if6_seq_show(struct seq_file *seq, void *v)
4479 {
4480 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4481 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4482 		   &ifp->addr,
4483 		   ifp->idev->dev->ifindex,
4484 		   ifp->prefix_len,
4485 		   ifp->scope,
4486 		   (u8) ifp->flags,
4487 		   ifp->idev->dev->name);
4488 	return 0;
4489 }
4490 
4491 static const struct seq_operations if6_seq_ops = {
4492 	.start	= if6_seq_start,
4493 	.next	= if6_seq_next,
4494 	.show	= if6_seq_show,
4495 	.stop	= if6_seq_stop,
4496 };
4497 
4498 static int __net_init if6_proc_net_init(struct net *net)
4499 {
4500 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4501 			sizeof(struct if6_iter_state)))
4502 		return -ENOMEM;
4503 	return 0;
4504 }
4505 
4506 static void __net_exit if6_proc_net_exit(struct net *net)
4507 {
4508 	remove_proc_entry("if_inet6", net->proc_net);
4509 }
4510 
4511 static struct pernet_operations if6_proc_net_ops = {
4512 	.init = if6_proc_net_init,
4513 	.exit = if6_proc_net_exit,
4514 };
4515 
4516 int __init if6_proc_init(void)
4517 {
4518 	return register_pernet_subsys(&if6_proc_net_ops);
4519 }
4520 
4521 void if6_proc_exit(void)
4522 {
4523 	unregister_pernet_subsys(&if6_proc_net_ops);
4524 }
4525 #endif	/* CONFIG_PROC_FS */
4526 
4527 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4528 /* Check if address is a home address configured on any interface. */
4529 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4530 {
4531 	unsigned int hash = inet6_addr_hash(net, addr);
4532 	struct inet6_ifaddr *ifp = NULL;
4533 	int ret = 0;
4534 
4535 	rcu_read_lock();
4536 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4537 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4538 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4539 			ret = 1;
4540 			break;
4541 		}
4542 	}
4543 	rcu_read_unlock();
4544 	return ret;
4545 }
4546 #endif
4547 
4548 /* RFC6554 has some algorithm to avoid loops in segment routing by
4549  * checking if the segments contains any of a local interface address.
4550  *
4551  * Quote:
4552  *
4553  * To detect loops in the SRH, a router MUST determine if the SRH
4554  * includes multiple addresses assigned to any interface on that router.
4555  * If such addresses appear more than once and are separated by at least
4556  * one address not assigned to that router.
4557  */
4558 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4559 			  unsigned char nsegs)
4560 {
4561 	const struct in6_addr *addr;
4562 	int i, ret = 0, found = 0;
4563 	struct inet6_ifaddr *ifp;
4564 	bool separated = false;
4565 	unsigned int hash;
4566 	bool hash_found;
4567 
4568 	rcu_read_lock();
4569 	for (i = 0; i < nsegs; i++) {
4570 		addr = &segs[i];
4571 		hash = inet6_addr_hash(net, addr);
4572 
4573 		hash_found = false;
4574 		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4575 
4576 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4577 				hash_found = true;
4578 				break;
4579 			}
4580 		}
4581 
4582 		if (hash_found) {
4583 			if (found > 1 && separated) {
4584 				ret = 1;
4585 				break;
4586 			}
4587 
4588 			separated = false;
4589 			found++;
4590 		} else {
4591 			separated = true;
4592 		}
4593 	}
4594 	rcu_read_unlock();
4595 
4596 	return ret;
4597 }
4598 
4599 /*
4600  *	Periodic address status verification
4601  */
4602 
4603 static void addrconf_verify_rtnl(struct net *net)
4604 {
4605 	unsigned long now, next, next_sec, next_sched;
4606 	struct inet6_ifaddr *ifp;
4607 	int i;
4608 
4609 	ASSERT_RTNL();
4610 
4611 	rcu_read_lock_bh();
4612 	now = jiffies;
4613 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4614 
4615 	cancel_delayed_work(&net->ipv6.addr_chk_work);
4616 
4617 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4618 restart:
4619 		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4620 			unsigned long age;
4621 
4622 			/* When setting preferred_lft to a value not zero or
4623 			 * infinity, while valid_lft is infinity
4624 			 * IFA_F_PERMANENT has a non-infinity life time.
4625 			 */
4626 			if ((ifp->flags & IFA_F_PERMANENT) &&
4627 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4628 				continue;
4629 
4630 			spin_lock(&ifp->lock);
4631 			/* We try to batch several events at once. */
4632 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4633 
4634 			if ((ifp->flags&IFA_F_TEMPORARY) &&
4635 			    !(ifp->flags&IFA_F_TENTATIVE) &&
4636 			    ifp->prefered_lft != INFINITY_LIFE_TIME &&
4637 			    !ifp->regen_count && ifp->ifpub) {
4638 				/* This is a non-regenerated temporary addr. */
4639 
4640 				unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4641 
4642 				if (age + regen_advance >= ifp->prefered_lft) {
4643 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4644 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4645 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4646 
4647 					ifp->regen_count++;
4648 					in6_ifa_hold(ifp);
4649 					in6_ifa_hold(ifpub);
4650 					spin_unlock(&ifp->lock);
4651 
4652 					spin_lock(&ifpub->lock);
4653 					ifpub->regen_count = 0;
4654 					spin_unlock(&ifpub->lock);
4655 					rcu_read_unlock_bh();
4656 					ipv6_create_tempaddr(ifpub, true);
4657 					in6_ifa_put(ifpub);
4658 					in6_ifa_put(ifp);
4659 					rcu_read_lock_bh();
4660 					goto restart;
4661 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4662 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4663 			}
4664 
4665 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4666 			    age >= ifp->valid_lft) {
4667 				spin_unlock(&ifp->lock);
4668 				in6_ifa_hold(ifp);
4669 				rcu_read_unlock_bh();
4670 				ipv6_del_addr(ifp);
4671 				rcu_read_lock_bh();
4672 				goto restart;
4673 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4674 				spin_unlock(&ifp->lock);
4675 				continue;
4676 			} else if (age >= ifp->prefered_lft) {
4677 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4678 				int deprecate = 0;
4679 
4680 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4681 					deprecate = 1;
4682 					ifp->flags |= IFA_F_DEPRECATED;
4683 				}
4684 
4685 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4686 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4687 					next = ifp->tstamp + ifp->valid_lft * HZ;
4688 
4689 				spin_unlock(&ifp->lock);
4690 
4691 				if (deprecate) {
4692 					in6_ifa_hold(ifp);
4693 
4694 					ipv6_ifa_notify(0, ifp);
4695 					in6_ifa_put(ifp);
4696 					goto restart;
4697 				}
4698 			} else {
4699 				/* ifp->prefered_lft <= ifp->valid_lft */
4700 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4701 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4702 				spin_unlock(&ifp->lock);
4703 			}
4704 		}
4705 	}
4706 
4707 	next_sec = round_jiffies_up(next);
4708 	next_sched = next;
4709 
4710 	/* If rounded timeout is accurate enough, accept it. */
4711 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4712 		next_sched = next_sec;
4713 
4714 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4715 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4716 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4717 
4718 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4719 		 now, next, next_sec, next_sched);
4720 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4721 	rcu_read_unlock_bh();
4722 }
4723 
4724 static void addrconf_verify_work(struct work_struct *w)
4725 {
4726 	struct net *net = container_of(to_delayed_work(w), struct net,
4727 				       ipv6.addr_chk_work);
4728 
4729 	rtnl_net_lock(net);
4730 	addrconf_verify_rtnl(net);
4731 	rtnl_net_unlock(net);
4732 }
4733 
4734 static void addrconf_verify(struct net *net)
4735 {
4736 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4737 }
4738 
4739 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4740 				     struct in6_addr **peer_pfx)
4741 {
4742 	struct in6_addr *pfx = NULL;
4743 
4744 	*peer_pfx = NULL;
4745 
4746 	if (addr)
4747 		pfx = nla_data(addr);
4748 
4749 	if (local) {
4750 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4751 			*peer_pfx = pfx;
4752 		pfx = nla_data(local);
4753 	}
4754 
4755 	return pfx;
4756 }
4757 
4758 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4759 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4760 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4761 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4762 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4763 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4764 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4765 	[IFA_PROTO]		= { .type = NLA_U8 },
4766 };
4767 
4768 static int
4769 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4770 		  struct netlink_ext_ack *extack)
4771 {
4772 	struct net *net = sock_net(skb->sk);
4773 	struct ifaddrmsg *ifm;
4774 	struct nlattr *tb[IFA_MAX+1];
4775 	struct in6_addr *pfx, *peer_pfx;
4776 	u32 ifa_flags;
4777 	int err;
4778 
4779 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4780 				     ifa_ipv6_policy, extack);
4781 	if (err < 0)
4782 		return err;
4783 
4784 	ifm = nlmsg_data(nlh);
4785 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4786 	if (!pfx)
4787 		return -EINVAL;
4788 
4789 	ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4790 
4791 	/* We ignore other flags so far. */
4792 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4793 
4794 	rtnl_net_lock(net);
4795 	err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4796 			     ifm->ifa_prefixlen, extack);
4797 	rtnl_net_unlock(net);
4798 
4799 	return err;
4800 }
4801 
4802 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp,
4803 			       unsigned long expires, u32 flags,
4804 			       bool modify_peer)
4805 {
4806 	struct fib6_table *table;
4807 	struct fib6_info *f6i;
4808 	u32 prio;
4809 
4810 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4811 					ifp->prefix_len,
4812 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4813 	if (!f6i)
4814 		return -ENOENT;
4815 
4816 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4817 	if (f6i->fib6_metric != prio) {
4818 		/* delete old one */
4819 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4820 
4821 		/* add new one */
4822 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4823 				      ifp->prefix_len,
4824 				      ifp->rt_priority, ifp->idev->dev,
4825 				      expires, flags, GFP_KERNEL);
4826 		return 0;
4827 	}
4828 	if (f6i != net->ipv6.fib6_null_entry) {
4829 		table = f6i->fib6_table;
4830 		spin_lock_bh(&table->tb6_lock);
4831 
4832 		if (!(flags & RTF_EXPIRES)) {
4833 			fib6_clean_expires(f6i);
4834 			fib6_remove_gc_list(f6i);
4835 		} else {
4836 			fib6_set_expires(f6i, expires);
4837 			fib6_add_gc_list(f6i);
4838 		}
4839 
4840 		spin_unlock_bh(&table->tb6_lock);
4841 	}
4842 	fib6_info_release(f6i);
4843 
4844 	return 0;
4845 }
4846 
4847 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4848 			     struct ifa6_config *cfg, clock_t expires,
4849 			     u32 flags)
4850 {
4851 	bool was_managetempaddr;
4852 	bool new_peer = false;
4853 	bool had_prefixroute;
4854 
4855 	ASSERT_RTNL_NET(net);
4856 
4857 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4858 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4859 		return -EINVAL;
4860 
4861 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4862 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4863 
4864 	if (cfg->peer_pfx &&
4865 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4866 		if (!ipv6_addr_any(&ifp->peer_addr))
4867 			cleanup_prefix_route(ifp, expires, true, true);
4868 		new_peer = true;
4869 	}
4870 
4871 	spin_lock_bh(&ifp->lock);
4872 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4873 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4874 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4875 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4876 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4877 			IFA_F_NOPREFIXROUTE);
4878 	ifp->flags |= cfg->ifa_flags;
4879 	WRITE_ONCE(ifp->tstamp, jiffies);
4880 	WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4881 	WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4882 	WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4883 
4884 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4885 		WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4886 
4887 	if (new_peer)
4888 		ifp->peer_addr = *cfg->peer_pfx;
4889 
4890 	spin_unlock_bh(&ifp->lock);
4891 	if (!(ifp->flags&IFA_F_TENTATIVE))
4892 		ipv6_ifa_notify(0, ifp);
4893 
4894 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4895 		int rc = -ENOENT;
4896 
4897 		if (had_prefixroute)
4898 			rc = modify_prefix_route(net, ifp, expires, flags, false);
4899 
4900 		/* prefix route could have been deleted; if so restore it */
4901 		if (rc == -ENOENT) {
4902 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4903 					      ifp->rt_priority, ifp->idev->dev,
4904 					      expires, flags, GFP_KERNEL);
4905 		}
4906 
4907 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4908 			rc = modify_prefix_route(net, ifp, expires, flags, true);
4909 
4910 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4911 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4912 					      ifp->rt_priority, ifp->idev->dev,
4913 					      expires, flags, GFP_KERNEL);
4914 		}
4915 	} else if (had_prefixroute) {
4916 		enum cleanup_prefix_rt_t action;
4917 		unsigned long rt_expires;
4918 
4919 		write_lock_bh(&ifp->idev->lock);
4920 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4921 		write_unlock_bh(&ifp->idev->lock);
4922 
4923 		if (action != CLEANUP_PREFIX_RT_NOP) {
4924 			cleanup_prefix_route(ifp, rt_expires,
4925 				action == CLEANUP_PREFIX_RT_DEL, false);
4926 		}
4927 	}
4928 
4929 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4930 		if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4931 			delete_tempaddrs(ifp->idev, ifp);
4932 		else
4933 			manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4934 					 cfg->preferred_lft, !was_managetempaddr,
4935 					 jiffies);
4936 	}
4937 
4938 	addrconf_verify_rtnl(net);
4939 
4940 	return 0;
4941 }
4942 
4943 static int
4944 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4945 		  struct netlink_ext_ack *extack)
4946 {
4947 	struct net *net = sock_net(skb->sk);
4948 	struct nlattr *tb[IFA_MAX+1];
4949 	struct in6_addr *peer_pfx;
4950 	struct inet6_ifaddr *ifa;
4951 	struct net_device *dev;
4952 	struct inet6_dev *idev;
4953 	struct ifa6_config cfg;
4954 	struct ifaddrmsg *ifm;
4955 	unsigned long timeout;
4956 	clock_t expires;
4957 	u32 flags;
4958 	int err;
4959 
4960 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4961 				     ifa_ipv6_policy, extack);
4962 	if (err < 0)
4963 		return err;
4964 
4965 	memset(&cfg, 0, sizeof(cfg));
4966 
4967 	ifm = nlmsg_data(nlh);
4968 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4969 	if (!cfg.pfx)
4970 		return -EINVAL;
4971 
4972 	cfg.peer_pfx = peer_pfx;
4973 	cfg.plen = ifm->ifa_prefixlen;
4974 	if (tb[IFA_RT_PRIORITY])
4975 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4976 
4977 	if (tb[IFA_PROTO])
4978 		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4979 
4980 	cfg.ifa_flags = nla_get_u32_default(tb[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 	cfg.ifa_flags |= IFA_F_PERMANENT;
4988 	cfg.valid_lft = INFINITY_LIFE_TIME;
4989 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4990 	expires = 0;
4991 	flags = 0;
4992 
4993 	if (tb[IFA_CACHEINFO]) {
4994 		struct ifa_cacheinfo *ci;
4995 
4996 		ci = nla_data(tb[IFA_CACHEINFO]);
4997 		cfg.valid_lft = ci->ifa_valid;
4998 		cfg.preferred_lft = ci->ifa_prefered;
4999 
5000 		if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) {
5001 			NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
5002 			return -EINVAL;
5003 		}
5004 
5005 		timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ);
5006 		if (addrconf_finite_timeout(timeout)) {
5007 			cfg.ifa_flags &= ~IFA_F_PERMANENT;
5008 			cfg.valid_lft = timeout;
5009 			expires = jiffies_to_clock_t(timeout * HZ);
5010 			flags = RTF_EXPIRES;
5011 		}
5012 
5013 		timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ);
5014 		if (addrconf_finite_timeout(timeout)) {
5015 			if (timeout == 0)
5016 				cfg.ifa_flags |= IFA_F_DEPRECATED;
5017 
5018 			cfg.preferred_lft = timeout;
5019 		}
5020 	}
5021 
5022 	rtnl_net_lock(net);
5023 
5024 	dev =  __dev_get_by_index(net, ifm->ifa_index);
5025 	if (!dev) {
5026 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5027 		err = -ENODEV;
5028 		goto unlock_rtnl;
5029 	}
5030 
5031 	netdev_lock_ops(dev);
5032 	idev = ipv6_find_idev(dev);
5033 	if (IS_ERR(idev)) {
5034 		err = PTR_ERR(idev);
5035 		goto unlock;
5036 	}
5037 
5038 	if (!ipv6_allow_optimistic_dad(net, idev))
5039 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5040 
5041 	if (cfg.ifa_flags & IFA_F_NODAD &&
5042 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5043 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5044 		err = -EINVAL;
5045 		goto unlock;
5046 	}
5047 
5048 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5049 	if (!ifa) {
5050 		/*
5051 		 * It would be best to check for !NLM_F_CREATE here but
5052 		 * userspace already relies on not having to provide this.
5053 		 */
5054 		err = inet6_addr_add(net, dev, &cfg, expires, flags, extack);
5055 		goto unlock;
5056 	}
5057 
5058 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
5059 	    !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5060 		NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5061 		err = -EEXIST;
5062 	} else {
5063 		err = inet6_addr_modify(net, ifa, &cfg, expires, flags);
5064 	}
5065 
5066 	in6_ifa_put(ifa);
5067 unlock:
5068 	netdev_unlock_ops(dev);
5069 unlock_rtnl:
5070 	rtnl_net_unlock(net);
5071 
5072 	return err;
5073 }
5074 
5075 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5076 			  u8 scope, int ifindex)
5077 {
5078 	struct ifaddrmsg *ifm;
5079 
5080 	ifm = nlmsg_data(nlh);
5081 	ifm->ifa_family = AF_INET6;
5082 	ifm->ifa_prefixlen = prefixlen;
5083 	ifm->ifa_flags = flags;
5084 	ifm->ifa_scope = scope;
5085 	ifm->ifa_index = ifindex;
5086 }
5087 
5088 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5089 			 unsigned long tstamp, u32 preferred, u32 valid)
5090 {
5091 	struct ifa_cacheinfo ci;
5092 
5093 	ci.cstamp = cstamp_delta(cstamp);
5094 	ci.tstamp = cstamp_delta(tstamp);
5095 	ci.ifa_prefered = preferred;
5096 	ci.ifa_valid = valid;
5097 
5098 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5099 }
5100 
5101 static inline int rt_scope(int ifa_scope)
5102 {
5103 	if (ifa_scope & IFA_HOST)
5104 		return RT_SCOPE_HOST;
5105 	else if (ifa_scope & IFA_LINK)
5106 		return RT_SCOPE_LINK;
5107 	else if (ifa_scope & IFA_SITE)
5108 		return RT_SCOPE_SITE;
5109 	else
5110 		return RT_SCOPE_UNIVERSE;
5111 }
5112 
5113 static inline int inet6_ifaddr_msgsize(void)
5114 {
5115 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5116 	       + nla_total_size(16) /* IFA_LOCAL */
5117 	       + nla_total_size(16) /* IFA_ADDRESS */
5118 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
5119 	       + nla_total_size(4)  /* IFA_FLAGS */
5120 	       + nla_total_size(1)  /* IFA_PROTO */
5121 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5122 }
5123 
5124 static int inet6_fill_ifaddr(struct sk_buff *skb,
5125 			     const struct inet6_ifaddr *ifa,
5126 			     struct inet6_fill_args *args)
5127 {
5128 	struct nlmsghdr *nlh;
5129 	u32 preferred, valid;
5130 	u32 flags, priority;
5131 	u8 proto;
5132 
5133 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5134 			sizeof(struct ifaddrmsg), args->flags);
5135 	if (!nlh)
5136 		return -EMSGSIZE;
5137 
5138 	flags = READ_ONCE(ifa->flags);
5139 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5140 		      ifa->idev->dev->ifindex);
5141 
5142 	if (args->netnsid >= 0 &&
5143 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5144 		goto error;
5145 
5146 	preferred = READ_ONCE(ifa->prefered_lft);
5147 	valid = READ_ONCE(ifa->valid_lft);
5148 
5149 	if (!((flags & IFA_F_PERMANENT) &&
5150 	      (preferred == INFINITY_LIFE_TIME))) {
5151 		if (preferred != INFINITY_LIFE_TIME) {
5152 			long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5153 
5154 			if (preferred > tval)
5155 				preferred -= tval;
5156 			else
5157 				preferred = 0;
5158 			if (valid != INFINITY_LIFE_TIME) {
5159 				if (valid > tval)
5160 					valid -= tval;
5161 				else
5162 					valid = 0;
5163 			}
5164 		}
5165 	} else {
5166 		preferred = INFINITY_LIFE_TIME;
5167 		valid = INFINITY_LIFE_TIME;
5168 	}
5169 
5170 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5171 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5172 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5173 			goto error;
5174 	} else {
5175 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5176 			goto error;
5177 	}
5178 
5179 	priority = READ_ONCE(ifa->rt_priority);
5180 	if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5181 		goto error;
5182 
5183 	if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5184 			  preferred, valid) < 0)
5185 		goto error;
5186 
5187 	if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5188 		goto error;
5189 
5190 	proto = READ_ONCE(ifa->ifa_proto);
5191 	if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5192 		goto error;
5193 
5194 	nlmsg_end(skb, nlh);
5195 	return 0;
5196 
5197 error:
5198 	nlmsg_cancel(skb, nlh);
5199 	return -EMSGSIZE;
5200 }
5201 
5202 int inet6_fill_ifmcaddr(struct sk_buff *skb,
5203 			const struct ifmcaddr6 *ifmca,
5204 			struct inet6_fill_args *args)
5205 {
5206 	int ifindex = ifmca->idev->dev->ifindex;
5207 	u8 scope = RT_SCOPE_UNIVERSE;
5208 	struct nlmsghdr *nlh;
5209 
5210 	if (!args->force_rt_scope_universe &&
5211 	    ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5212 		scope = RT_SCOPE_SITE;
5213 
5214 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5215 			sizeof(struct ifaddrmsg), args->flags);
5216 	if (!nlh)
5217 		return -EMSGSIZE;
5218 
5219 	if (args->netnsid >= 0 &&
5220 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5221 		nlmsg_cancel(skb, nlh);
5222 		return -EMSGSIZE;
5223 	}
5224 
5225 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5226 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5227 	    put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5228 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5229 		nlmsg_cancel(skb, nlh);
5230 		return -EMSGSIZE;
5231 	}
5232 
5233 	nlmsg_end(skb, nlh);
5234 	return 0;
5235 }
5236 
5237 int inet6_fill_ifacaddr(struct sk_buff *skb,
5238 			const struct ifacaddr6 *ifaca,
5239 			struct inet6_fill_args *args)
5240 {
5241 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5242 	int ifindex = dev ? dev->ifindex : 1;
5243 	u8 scope = RT_SCOPE_UNIVERSE;
5244 	struct nlmsghdr *nlh;
5245 
5246 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5247 		scope = RT_SCOPE_SITE;
5248 
5249 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5250 			sizeof(struct ifaddrmsg), args->flags);
5251 	if (!nlh)
5252 		return -EMSGSIZE;
5253 
5254 	if (args->netnsid >= 0 &&
5255 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5256 		nlmsg_cancel(skb, nlh);
5257 		return -EMSGSIZE;
5258 	}
5259 
5260 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5261 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5262 	    put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5263 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5264 		nlmsg_cancel(skb, nlh);
5265 		return -EMSGSIZE;
5266 	}
5267 
5268 	nlmsg_end(skb, nlh);
5269 	return 0;
5270 }
5271 
5272 /* called with rcu_read_lock() */
5273 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5274 			  struct netlink_callback *cb, int *s_ip_idx,
5275 			  struct inet6_fill_args *fillargs)
5276 {
5277 	const struct ifmcaddr6 *ifmca;
5278 	const struct ifacaddr6 *ifaca;
5279 	int ip_idx = 0;
5280 	int err = 0;
5281 
5282 	switch (fillargs->type) {
5283 	case UNICAST_ADDR: {
5284 		const struct inet6_ifaddr *ifa;
5285 		fillargs->event = RTM_NEWADDR;
5286 
5287 		/* unicast address incl. temp addr */
5288 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5289 			if (ip_idx < *s_ip_idx)
5290 				goto next;
5291 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5292 			if (err < 0)
5293 				break;
5294 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5295 next:
5296 			ip_idx++;
5297 		}
5298 		break;
5299 	}
5300 	case MULTICAST_ADDR:
5301 		fillargs->event = RTM_GETMULTICAST;
5302 
5303 		/* multicast address */
5304 		for (ifmca = rcu_dereference(idev->mc_list);
5305 		     ifmca;
5306 		     ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5307 			if (ip_idx < *s_ip_idx)
5308 				continue;
5309 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5310 			if (err < 0)
5311 				break;
5312 		}
5313 		break;
5314 	case ANYCAST_ADDR:
5315 		fillargs->event = RTM_GETANYCAST;
5316 		/* anycast address */
5317 		for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5318 		     ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5319 			if (ip_idx < *s_ip_idx)
5320 				continue;
5321 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5322 			if (err < 0)
5323 				break;
5324 		}
5325 		break;
5326 	default:
5327 		break;
5328 	}
5329 	*s_ip_idx = err ? ip_idx : 0;
5330 	return err;
5331 }
5332 
5333 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5334 				       struct inet6_fill_args *fillargs,
5335 				       struct net **tgt_net, struct sock *sk,
5336 				       struct netlink_callback *cb)
5337 {
5338 	struct netlink_ext_ack *extack = cb->extack;
5339 	struct nlattr *tb[IFA_MAX+1];
5340 	struct ifaddrmsg *ifm;
5341 	int err, i;
5342 
5343 	ifm = nlmsg_payload(nlh, sizeof(*ifm));
5344 	if (!ifm) {
5345 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5346 		return -EINVAL;
5347 	}
5348 
5349 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5350 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5351 		return -EINVAL;
5352 	}
5353 
5354 	fillargs->ifindex = ifm->ifa_index;
5355 	if (fillargs->ifindex) {
5356 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5357 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5358 	}
5359 
5360 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5361 					    ifa_ipv6_policy, extack);
5362 	if (err < 0)
5363 		return err;
5364 
5365 	for (i = 0; i <= IFA_MAX; ++i) {
5366 		if (!tb[i])
5367 			continue;
5368 
5369 		if (i == IFA_TARGET_NETNSID) {
5370 			struct net *net;
5371 
5372 			fillargs->netnsid = nla_get_s32(tb[i]);
5373 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5374 			if (IS_ERR(net)) {
5375 				fillargs->netnsid = -1;
5376 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5377 				return PTR_ERR(net);
5378 			}
5379 			*tgt_net = net;
5380 		} else {
5381 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5382 			return -EINVAL;
5383 		}
5384 	}
5385 
5386 	return 0;
5387 }
5388 
5389 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5390 			   enum addr_type_t type)
5391 {
5392 	struct net *tgt_net = sock_net(skb->sk);
5393 	const struct nlmsghdr *nlh = cb->nlh;
5394 	struct inet6_fill_args fillargs = {
5395 		.portid = NETLINK_CB(cb->skb).portid,
5396 		.seq = cb->nlh->nlmsg_seq,
5397 		.flags = NLM_F_MULTI,
5398 		.netnsid = -1,
5399 		.type = type,
5400 		.force_rt_scope_universe = false,
5401 	};
5402 	struct {
5403 		unsigned long ifindex;
5404 		int ip_idx;
5405 	} *ctx = (void *)cb->ctx;
5406 	struct net_device *dev;
5407 	struct inet6_dev *idev;
5408 	int err = 0;
5409 
5410 	rcu_read_lock();
5411 	if (cb->strict_check) {
5412 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5413 						  skb->sk, cb);
5414 		if (err < 0)
5415 			goto done;
5416 
5417 		err = 0;
5418 		if (fillargs.ifindex) {
5419 			dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5420 			if (!dev) {
5421 				err = -ENODEV;
5422 				goto done;
5423 			}
5424 			idev = __in6_dev_get(dev);
5425 			if (idev)
5426 				err = in6_dump_addrs(idev, skb, cb,
5427 						     &ctx->ip_idx,
5428 						     &fillargs);
5429 			goto done;
5430 		}
5431 	}
5432 
5433 	cb->seq = inet6_base_seq(tgt_net);
5434 	for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5435 		idev = __in6_dev_get(dev);
5436 		if (!idev)
5437 			continue;
5438 		err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5439 				     &fillargs);
5440 		if (err < 0)
5441 			goto done;
5442 	}
5443 done:
5444 	rcu_read_unlock();
5445 	if (fillargs.netnsid >= 0)
5446 		put_net(tgt_net);
5447 
5448 	return err;
5449 }
5450 
5451 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5452 {
5453 	enum addr_type_t type = UNICAST_ADDR;
5454 
5455 	return inet6_dump_addr(skb, cb, type);
5456 }
5457 
5458 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5459 {
5460 	enum addr_type_t type = MULTICAST_ADDR;
5461 
5462 	return inet6_dump_addr(skb, cb, type);
5463 }
5464 
5465 
5466 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5467 {
5468 	enum addr_type_t type = ANYCAST_ADDR;
5469 
5470 	return inet6_dump_addr(skb, cb, type);
5471 }
5472 
5473 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5474 				       const struct nlmsghdr *nlh,
5475 				       struct nlattr **tb,
5476 				       struct netlink_ext_ack *extack)
5477 {
5478 	struct ifaddrmsg *ifm;
5479 	int i, err;
5480 
5481 	ifm = nlmsg_payload(nlh, sizeof(*ifm));
5482 	if (!ifm) {
5483 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5484 		return -EINVAL;
5485 	}
5486 
5487 	if (!netlink_strict_get_check(skb))
5488 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5489 					      ifa_ipv6_policy, extack);
5490 
5491 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5492 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5493 		return -EINVAL;
5494 	}
5495 
5496 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5497 					    ifa_ipv6_policy, extack);
5498 	if (err)
5499 		return err;
5500 
5501 	for (i = 0; i <= IFA_MAX; i++) {
5502 		if (!tb[i])
5503 			continue;
5504 
5505 		switch (i) {
5506 		case IFA_TARGET_NETNSID:
5507 		case IFA_ADDRESS:
5508 		case IFA_LOCAL:
5509 			break;
5510 		default:
5511 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5512 			return -EINVAL;
5513 		}
5514 	}
5515 
5516 	return 0;
5517 }
5518 
5519 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5520 			     struct netlink_ext_ack *extack)
5521 {
5522 	struct net *tgt_net = sock_net(in_skb->sk);
5523 	struct inet6_fill_args fillargs = {
5524 		.portid = NETLINK_CB(in_skb).portid,
5525 		.seq = nlh->nlmsg_seq,
5526 		.event = RTM_NEWADDR,
5527 		.flags = 0,
5528 		.netnsid = -1,
5529 		.force_rt_scope_universe = false,
5530 	};
5531 	struct ifaddrmsg *ifm;
5532 	struct nlattr *tb[IFA_MAX+1];
5533 	struct in6_addr *addr = NULL, *peer;
5534 	struct net_device *dev = NULL;
5535 	struct inet6_ifaddr *ifa;
5536 	struct sk_buff *skb;
5537 	int err;
5538 
5539 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5540 	if (err < 0)
5541 		return err;
5542 
5543 	if (tb[IFA_TARGET_NETNSID]) {
5544 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5545 
5546 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5547 						  fillargs.netnsid);
5548 		if (IS_ERR(tgt_net))
5549 			return PTR_ERR(tgt_net);
5550 	}
5551 
5552 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5553 	if (!addr) {
5554 		err = -EINVAL;
5555 		goto errout;
5556 	}
5557 	ifm = nlmsg_data(nlh);
5558 	if (ifm->ifa_index)
5559 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5560 
5561 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5562 	if (!ifa) {
5563 		err = -EADDRNOTAVAIL;
5564 		goto errout;
5565 	}
5566 
5567 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5568 	if (!skb) {
5569 		err = -ENOBUFS;
5570 		goto errout_ifa;
5571 	}
5572 
5573 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5574 	if (err < 0) {
5575 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5576 		WARN_ON(err == -EMSGSIZE);
5577 		kfree_skb(skb);
5578 		goto errout_ifa;
5579 	}
5580 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5581 errout_ifa:
5582 	in6_ifa_put(ifa);
5583 errout:
5584 	dev_put(dev);
5585 	if (fillargs.netnsid >= 0)
5586 		put_net(tgt_net);
5587 
5588 	return err;
5589 }
5590 
5591 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5592 {
5593 	struct sk_buff *skb;
5594 	struct net *net = dev_net(ifa->idev->dev);
5595 	struct inet6_fill_args fillargs = {
5596 		.portid = 0,
5597 		.seq = 0,
5598 		.event = event,
5599 		.flags = 0,
5600 		.netnsid = -1,
5601 		.force_rt_scope_universe = false,
5602 	};
5603 	int err = -ENOBUFS;
5604 
5605 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5606 	if (!skb)
5607 		goto errout;
5608 
5609 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5610 	if (err < 0) {
5611 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5612 		WARN_ON(err == -EMSGSIZE);
5613 		kfree_skb(skb);
5614 		goto errout;
5615 	}
5616 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5617 	return;
5618 errout:
5619 	rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5620 }
5621 
5622 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5623 			       __s32 *array, int bytes)
5624 {
5625 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5626 
5627 	memset(array, 0, bytes);
5628 	array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5629 	array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5630 	array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5631 	array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5632 	array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5633 	array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5634 	array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5635 	array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5636 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5637 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5638 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5639 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5640 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5641 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5642 	array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5643 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5644 		jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5645 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5646 		jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5647 	array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5648 	array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5649 	array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5650 	array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5651 	array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5652 	array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5653 	array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5654 	array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5655 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5656 		READ_ONCE(cnf->accept_ra_min_hop_limit);
5657 	array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5658 #ifdef CONFIG_IPV6_ROUTER_PREF
5659 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5660 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5661 		jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5662 #ifdef CONFIG_IPV6_ROUTE_INFO
5663 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5664 		READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5665 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5666 		READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5667 #endif
5668 #endif
5669 	array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5670 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5671 		READ_ONCE(cnf->accept_source_route);
5672 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5673 	array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5674 	array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5675 #endif
5676 #ifdef CONFIG_IPV6_MROUTE
5677 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5678 #endif
5679 	array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5680 	array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5681 	array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5682 	array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5683 	array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5684 		READ_ONCE(cnf->suppress_frag_ndisc);
5685 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5686 		READ_ONCE(cnf->accept_ra_from_local);
5687 	array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5688 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5689 		READ_ONCE(cnf->ignore_routes_with_linkdown);
5690 	/* we omit DEVCONF_STABLE_SECRET for now */
5691 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5692 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5693 		READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5694 	array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5695 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5696 	array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5697 #ifdef CONFIG_IPV6_SEG6_HMAC
5698 	array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5699 #endif
5700 	array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5701 	array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5702 	array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5703 	array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5704 	array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5705 	array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5706 	array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5707 	array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5708 	array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5709 		READ_ONCE(cnf->ndisc_evict_nocarrier);
5710 	array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5711 		READ_ONCE(cnf->accept_untracked_na);
5712 	array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5713 }
5714 
5715 static inline size_t inet6_ifla6_size(void)
5716 {
5717 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5718 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5719 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5720 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5721 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5722 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5723 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5724 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5725 	     + 0;
5726 }
5727 
5728 static inline size_t inet6_if_nlmsg_size(void)
5729 {
5730 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5731 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5732 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5733 	       + nla_total_size(4) /* IFLA_MTU */
5734 	       + nla_total_size(4) /* IFLA_LINK */
5735 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5736 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5737 }
5738 
5739 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5740 					int bytes)
5741 {
5742 	int i;
5743 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5744 	BUG_ON(pad < 0);
5745 
5746 	/* Use put_unaligned() because stats may not be aligned for u64. */
5747 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5748 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5749 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5750 
5751 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5752 }
5753 
5754 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5755 					int bytes, size_t syncpoff)
5756 {
5757 	int i, c;
5758 	u64 buff[IPSTATS_MIB_MAX];
5759 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5760 
5761 	BUG_ON(pad < 0);
5762 
5763 	memset(buff, 0, sizeof(buff));
5764 	buff[0] = IPSTATS_MIB_MAX;
5765 
5766 	for_each_possible_cpu(c) {
5767 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5768 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5769 	}
5770 
5771 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5772 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5773 }
5774 
5775 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5776 			     int bytes)
5777 {
5778 	switch (attrtype) {
5779 	case IFLA_INET6_STATS:
5780 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5781 				     offsetof(struct ipstats_mib, syncp));
5782 		break;
5783 	case IFLA_INET6_ICMP6STATS:
5784 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5785 		break;
5786 	}
5787 }
5788 
5789 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb,
5790 					struct inet6_dev *idev)
5791 {
5792 	struct nlattr *nla;
5793 
5794 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5795 	if (!nla)
5796 		goto nla_put_failure;
5797 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5798 
5799 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5800 	if (!nla)
5801 		goto nla_put_failure;
5802 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5803 
5804 	return 0;
5805 
5806 nla_put_failure:
5807 	return -EMSGSIZE;
5808 }
5809 
5810 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5811 				  u32 ext_filter_mask)
5812 {
5813 	struct ifla_cacheinfo ci;
5814 	struct nlattr *nla;
5815 	u32 ra_mtu;
5816 
5817 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5818 		goto nla_put_failure;
5819 	ci.max_reasm_len = IPV6_MAXPLEN;
5820 	ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5821 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5822 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5823 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5824 		goto nla_put_failure;
5825 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5826 	if (!nla)
5827 		goto nla_put_failure;
5828 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5829 
5830 	/* XXX - MC not implemented */
5831 
5832 	if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) {
5833 		if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0)
5834 			goto nla_put_failure;
5835 	}
5836 
5837 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5838 	if (!nla)
5839 		goto nla_put_failure;
5840 	read_lock_bh(&idev->lock);
5841 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5842 	read_unlock_bh(&idev->lock);
5843 
5844 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5845 		       READ_ONCE(idev->cnf.addr_gen_mode)))
5846 		goto nla_put_failure;
5847 
5848 	ra_mtu = READ_ONCE(idev->ra_mtu);
5849 	if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5850 		goto nla_put_failure;
5851 
5852 	return 0;
5853 
5854 nla_put_failure:
5855 	return -EMSGSIZE;
5856 }
5857 
5858 static size_t inet6_get_link_af_size(const struct net_device *dev,
5859 				     u32 ext_filter_mask)
5860 {
5861 	if (!__in6_dev_get(dev))
5862 		return 0;
5863 
5864 	return inet6_ifla6_size();
5865 }
5866 
5867 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5868 			      u32 ext_filter_mask)
5869 {
5870 	struct inet6_dev *idev = __in6_dev_get(dev);
5871 
5872 	if (!idev)
5873 		return -ENODATA;
5874 
5875 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5876 		return -EMSGSIZE;
5877 
5878 	return 0;
5879 }
5880 
5881 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5882 			     struct netlink_ext_ack *extack)
5883 {
5884 	struct inet6_ifaddr *ifp;
5885 	struct net_device *dev = idev->dev;
5886 	bool clear_token, update_rs = false;
5887 	struct in6_addr ll_addr;
5888 
5889 	ASSERT_RTNL();
5890 
5891 	if (!token)
5892 		return -EINVAL;
5893 
5894 	if (dev->flags & IFF_LOOPBACK) {
5895 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5896 		return -EINVAL;
5897 	}
5898 
5899 	if (dev->flags & IFF_NOARP) {
5900 		NL_SET_ERR_MSG_MOD(extack,
5901 				   "Device does not do neighbour discovery");
5902 		return -EINVAL;
5903 	}
5904 
5905 	if (!ipv6_accept_ra(idev)) {
5906 		NL_SET_ERR_MSG_MOD(extack,
5907 				   "Router advertisement is disabled on device");
5908 		return -EINVAL;
5909 	}
5910 
5911 	if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5912 		NL_SET_ERR_MSG(extack,
5913 			       "Router solicitation is disabled on device");
5914 		return -EINVAL;
5915 	}
5916 
5917 	write_lock_bh(&idev->lock);
5918 
5919 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5920 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5921 
5922 	write_unlock_bh(&idev->lock);
5923 
5924 	clear_token = ipv6_addr_any(token);
5925 	if (clear_token)
5926 		goto update_lft;
5927 
5928 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5929 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5930 			     IFA_F_OPTIMISTIC)) {
5931 		/* If we're not ready, then normal ifup will take care
5932 		 * of this. Otherwise, we need to request our rs here.
5933 		 */
5934 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5935 		update_rs = true;
5936 	}
5937 
5938 update_lft:
5939 	write_lock_bh(&idev->lock);
5940 
5941 	if (update_rs) {
5942 		idev->if_flags |= IF_RS_SENT;
5943 		idev->rs_interval = rfc3315_s14_backoff_init(
5944 			READ_ONCE(idev->cnf.rtr_solicit_interval));
5945 		idev->rs_probes = 1;
5946 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5947 	}
5948 
5949 	/* Well, that's kinda nasty ... */
5950 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5951 		spin_lock(&ifp->lock);
5952 		if (ifp->tokenized) {
5953 			ifp->valid_lft = 0;
5954 			ifp->prefered_lft = 0;
5955 		}
5956 		spin_unlock(&ifp->lock);
5957 	}
5958 
5959 	write_unlock_bh(&idev->lock);
5960 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5961 	addrconf_verify_rtnl(dev_net(dev));
5962 	return 0;
5963 }
5964 
5965 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5966 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5967 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5968 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5969 					    .reject_message =
5970 						"IFLA_INET6_RA_MTU can not be set" },
5971 };
5972 
5973 static int check_addr_gen_mode(int mode)
5974 {
5975 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5976 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5977 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5978 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5979 		return -EINVAL;
5980 	return 1;
5981 }
5982 
5983 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5984 				int mode)
5985 {
5986 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5987 	    !idev->cnf.stable_secret.initialized &&
5988 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5989 		return -EINVAL;
5990 	return 1;
5991 }
5992 
5993 static int inet6_validate_link_af(const struct net_device *dev,
5994 				  const struct nlattr *nla,
5995 				  struct netlink_ext_ack *extack)
5996 {
5997 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5998 	struct inet6_dev *idev = NULL;
5999 	int err;
6000 
6001 	if (dev) {
6002 		idev = __in6_dev_get(dev);
6003 		if (!idev)
6004 			return -EAFNOSUPPORT;
6005 	}
6006 
6007 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
6008 					  inet6_af_policy, extack);
6009 	if (err)
6010 		return err;
6011 
6012 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6013 		return -EINVAL;
6014 
6015 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6016 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6017 
6018 		if (check_addr_gen_mode(mode) < 0)
6019 			return -EINVAL;
6020 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6021 			return -EINVAL;
6022 	}
6023 
6024 	return 0;
6025 }
6026 
6027 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6028 			     struct netlink_ext_ack *extack)
6029 {
6030 	struct inet6_dev *idev = __in6_dev_get(dev);
6031 	struct nlattr *tb[IFLA_INET6_MAX + 1];
6032 	int err;
6033 
6034 	if (!idev)
6035 		return -EAFNOSUPPORT;
6036 
6037 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6038 		return -EINVAL;
6039 
6040 	if (tb[IFLA_INET6_TOKEN]) {
6041 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6042 					extack);
6043 		if (err)
6044 			return err;
6045 	}
6046 
6047 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6048 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6049 
6050 		WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6051 	}
6052 
6053 	return 0;
6054 }
6055 
6056 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6057 			     u32 portid, u32 seq, int event, unsigned int flags)
6058 {
6059 	struct net_device *dev = idev->dev;
6060 	struct ifinfomsg *hdr;
6061 	struct nlmsghdr *nlh;
6062 	int ifindex, iflink;
6063 	void *protoinfo;
6064 
6065 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6066 	if (!nlh)
6067 		return -EMSGSIZE;
6068 
6069 	hdr = nlmsg_data(nlh);
6070 	hdr->ifi_family = AF_INET6;
6071 	hdr->__ifi_pad = 0;
6072 	hdr->ifi_type = dev->type;
6073 	ifindex = READ_ONCE(dev->ifindex);
6074 	hdr->ifi_index = ifindex;
6075 	hdr->ifi_flags = dev_get_flags(dev);
6076 	hdr->ifi_change = 0;
6077 
6078 	iflink = dev_get_iflink(dev);
6079 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6080 	    (dev->addr_len &&
6081 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6082 	    nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6083 	    (ifindex != iflink &&
6084 	     nla_put_u32(skb, IFLA_LINK, iflink)) ||
6085 	    nla_put_u8(skb, IFLA_OPERSTATE,
6086 		       netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6087 		goto nla_put_failure;
6088 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6089 	if (!protoinfo)
6090 		goto nla_put_failure;
6091 
6092 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6093 		goto nla_put_failure;
6094 
6095 	nla_nest_end(skb, protoinfo);
6096 	nlmsg_end(skb, nlh);
6097 	return 0;
6098 
6099 nla_put_failure:
6100 	nlmsg_cancel(skb, nlh);
6101 	return -EMSGSIZE;
6102 }
6103 
6104 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6105 				   struct netlink_ext_ack *extack)
6106 {
6107 	struct ifinfomsg *ifm;
6108 
6109 	ifm = nlmsg_payload(nlh, sizeof(*ifm));
6110 	if (!ifm) {
6111 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6112 		return -EINVAL;
6113 	}
6114 
6115 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6116 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6117 		return -EINVAL;
6118 	}
6119 
6120 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6121 	    ifm->ifi_change || ifm->ifi_index) {
6122 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6123 		return -EINVAL;
6124 	}
6125 
6126 	return 0;
6127 }
6128 
6129 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6130 {
6131 	struct net *net = sock_net(skb->sk);
6132 	struct {
6133 		unsigned long ifindex;
6134 	} *ctx = (void *)cb->ctx;
6135 	struct net_device *dev;
6136 	struct inet6_dev *idev;
6137 	int err;
6138 
6139 	/* only requests using strict checking can pass data to
6140 	 * influence the dump
6141 	 */
6142 	if (cb->strict_check) {
6143 		err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6144 
6145 		if (err < 0)
6146 			return err;
6147 	}
6148 
6149 	err = 0;
6150 	rcu_read_lock();
6151 	for_each_netdev_dump(net, dev, ctx->ifindex) {
6152 		idev = __in6_dev_get(dev);
6153 		if (!idev)
6154 			continue;
6155 		err = inet6_fill_ifinfo(skb, idev,
6156 					NETLINK_CB(cb->skb).portid,
6157 					cb->nlh->nlmsg_seq,
6158 					RTM_NEWLINK, NLM_F_MULTI);
6159 		if (err < 0)
6160 			break;
6161 	}
6162 	rcu_read_unlock();
6163 
6164 	return err;
6165 }
6166 
6167 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6168 {
6169 	struct sk_buff *skb;
6170 	struct net *net = dev_net(idev->dev);
6171 	int err = -ENOBUFS;
6172 
6173 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6174 	if (!skb)
6175 		goto errout;
6176 
6177 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6178 	if (err < 0) {
6179 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6180 		WARN_ON(err == -EMSGSIZE);
6181 		kfree_skb(skb);
6182 		goto errout;
6183 	}
6184 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6185 	return;
6186 errout:
6187 	rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6188 }
6189 
6190 static inline size_t inet6_prefix_nlmsg_size(void)
6191 {
6192 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6193 	       + nla_total_size(sizeof(struct in6_addr))
6194 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6195 }
6196 
6197 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6198 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6199 			     int event, unsigned int flags)
6200 {
6201 	struct prefixmsg *pmsg;
6202 	struct nlmsghdr *nlh;
6203 	struct prefix_cacheinfo	ci;
6204 
6205 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6206 	if (!nlh)
6207 		return -EMSGSIZE;
6208 
6209 	pmsg = nlmsg_data(nlh);
6210 	pmsg->prefix_family = AF_INET6;
6211 	pmsg->prefix_pad1 = 0;
6212 	pmsg->prefix_pad2 = 0;
6213 	pmsg->prefix_ifindex = idev->dev->ifindex;
6214 	pmsg->prefix_len = pinfo->prefix_len;
6215 	pmsg->prefix_type = pinfo->type;
6216 	pmsg->prefix_pad3 = 0;
6217 	pmsg->prefix_flags = pinfo->flags;
6218 
6219 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6220 		goto nla_put_failure;
6221 	ci.preferred_time = ntohl(pinfo->prefered);
6222 	ci.valid_time = ntohl(pinfo->valid);
6223 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6224 		goto nla_put_failure;
6225 	nlmsg_end(skb, nlh);
6226 	return 0;
6227 
6228 nla_put_failure:
6229 	nlmsg_cancel(skb, nlh);
6230 	return -EMSGSIZE;
6231 }
6232 
6233 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6234 			 struct prefix_info *pinfo)
6235 {
6236 	struct sk_buff *skb;
6237 	struct net *net = dev_net(idev->dev);
6238 	int err = -ENOBUFS;
6239 
6240 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6241 	if (!skb)
6242 		goto errout;
6243 
6244 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6245 	if (err < 0) {
6246 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6247 		WARN_ON(err == -EMSGSIZE);
6248 		kfree_skb(skb);
6249 		goto errout;
6250 	}
6251 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6252 	return;
6253 errout:
6254 	rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6255 }
6256 
6257 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6258 {
6259 	struct net *net = dev_net(ifp->idev->dev);
6260 
6261 	if (event)
6262 		ASSERT_RTNL();
6263 
6264 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6265 
6266 	switch (event) {
6267 	case RTM_NEWADDR:
6268 		/*
6269 		 * If the address was optimistic we inserted the route at the
6270 		 * start of our DAD process, so we don't need to do it again.
6271 		 * If the device was taken down in the middle of the DAD
6272 		 * cycle there is a race where we could get here without a
6273 		 * host route, so nothing to insert. That will be fixed when
6274 		 * the device is brought up.
6275 		 */
6276 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6277 			ip6_ins_rt(net, ifp->rt);
6278 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6279 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6280 				&ifp->addr, ifp->idev->dev->name);
6281 		}
6282 
6283 		if (ifp->idev->cnf.forwarding)
6284 			addrconf_join_anycast(ifp);
6285 		if (!ipv6_addr_any(&ifp->peer_addr))
6286 			addrconf_prefix_route(&ifp->peer_addr, 128,
6287 					      ifp->rt_priority, ifp->idev->dev,
6288 					      0, 0, GFP_ATOMIC);
6289 		break;
6290 	case RTM_DELADDR:
6291 		if (ifp->idev->cnf.forwarding)
6292 			addrconf_leave_anycast(ifp);
6293 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6294 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6295 			struct fib6_info *rt;
6296 
6297 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6298 						       ifp->idev->dev, 0, 0,
6299 						       false);
6300 			if (rt)
6301 				ip6_del_rt(net, rt, false);
6302 		}
6303 		if (ifp->rt) {
6304 			ip6_del_rt(net, ifp->rt, false);
6305 			ifp->rt = NULL;
6306 		}
6307 		rt_genid_bump_ipv6(net);
6308 		break;
6309 	}
6310 	atomic_inc(&net->ipv6.dev_addr_genid);
6311 }
6312 
6313 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6314 {
6315 	if (likely(ifp->idev->dead == 0))
6316 		__ipv6_ifa_notify(event, ifp);
6317 }
6318 
6319 #ifdef CONFIG_SYSCTL
6320 
6321 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6322 		void *buffer, size_t *lenp, loff_t *ppos)
6323 {
6324 	int *valp = ctl->data;
6325 	int val = *valp;
6326 	loff_t pos = *ppos;
6327 	struct ctl_table lctl;
6328 	int ret;
6329 
6330 	/*
6331 	 * ctl->data points to idev->cnf.forwarding, we should
6332 	 * not modify it until we get the rtnl lock.
6333 	 */
6334 	lctl = *ctl;
6335 	lctl.data = &val;
6336 
6337 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6338 
6339 	if (write)
6340 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6341 	if (ret)
6342 		*ppos = pos;
6343 	return ret;
6344 }
6345 
6346 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6347 		void *buffer, size_t *lenp, loff_t *ppos)
6348 {
6349 	struct inet6_dev *idev = ctl->extra1;
6350 	int min_mtu = IPV6_MIN_MTU;
6351 	struct ctl_table lctl;
6352 
6353 	lctl = *ctl;
6354 	lctl.extra1 = &min_mtu;
6355 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6356 
6357 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6358 }
6359 
6360 static void dev_disable_change(struct inet6_dev *idev)
6361 {
6362 	struct netdev_notifier_info info;
6363 
6364 	if (!idev || !idev->dev)
6365 		return;
6366 
6367 	netdev_notifier_info_init(&info, idev->dev);
6368 	if (idev->cnf.disable_ipv6)
6369 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6370 	else
6371 		addrconf_notify(NULL, NETDEV_UP, &info);
6372 }
6373 
6374 static void addrconf_disable_change(struct net *net, __s32 newf)
6375 {
6376 	struct net_device *dev;
6377 	struct inet6_dev *idev;
6378 
6379 	for_each_netdev(net, dev) {
6380 		idev = __in6_dev_get_rtnl_net(dev);
6381 		if (idev) {
6382 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6383 
6384 			WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6385 			if (changed)
6386 				dev_disable_change(idev);
6387 		}
6388 	}
6389 }
6390 
6391 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6392 {
6393 	struct net *net = (struct net *)table->extra2;
6394 	int old;
6395 
6396 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6397 		WRITE_ONCE(*p, newf);
6398 		return 0;
6399 	}
6400 
6401 	if (!rtnl_net_trylock(net))
6402 		return restart_syscall();
6403 
6404 	old = *p;
6405 	WRITE_ONCE(*p, newf);
6406 
6407 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6408 		WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6409 		addrconf_disable_change(net, newf);
6410 	} else if ((!newf) ^ (!old)) {
6411 		dev_disable_change((struct inet6_dev *)table->extra1);
6412 	}
6413 
6414 	rtnl_net_unlock(net);
6415 	return 0;
6416 }
6417 
6418 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6419 		void *buffer, size_t *lenp, loff_t *ppos)
6420 {
6421 	int *valp = ctl->data;
6422 	int val = *valp;
6423 	loff_t pos = *ppos;
6424 	struct ctl_table lctl;
6425 	int ret;
6426 
6427 	/*
6428 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6429 	 * not modify it until we get the rtnl lock.
6430 	 */
6431 	lctl = *ctl;
6432 	lctl.data = &val;
6433 
6434 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6435 
6436 	if (write)
6437 		ret = addrconf_disable_ipv6(ctl, valp, val);
6438 	if (ret)
6439 		*ppos = pos;
6440 	return ret;
6441 }
6442 
6443 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6444 		void *buffer, size_t *lenp, loff_t *ppos)
6445 {
6446 	int *valp = ctl->data;
6447 	int ret;
6448 	int old, new;
6449 
6450 	old = *valp;
6451 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6452 	new = *valp;
6453 
6454 	if (write && old != new) {
6455 		struct net *net = ctl->extra2;
6456 
6457 		if (!rtnl_net_trylock(net))
6458 			return restart_syscall();
6459 
6460 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp) {
6461 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6462 						     NETCONFA_PROXY_NEIGH,
6463 						     NETCONFA_IFINDEX_DEFAULT,
6464 						     net->ipv6.devconf_dflt);
6465 		} else if (valp == &net->ipv6.devconf_all->proxy_ndp) {
6466 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6467 						     NETCONFA_PROXY_NEIGH,
6468 						     NETCONFA_IFINDEX_ALL,
6469 						     net->ipv6.devconf_all);
6470 		} else {
6471 			struct inet6_dev *idev = ctl->extra1;
6472 
6473 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6474 						     NETCONFA_PROXY_NEIGH,
6475 						     idev->dev->ifindex,
6476 						     &idev->cnf);
6477 		}
6478 		rtnl_net_unlock(net);
6479 	}
6480 
6481 	return ret;
6482 }
6483 
6484 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6485 					 void *buffer, size_t *lenp,
6486 					 loff_t *ppos)
6487 {
6488 	int ret = 0;
6489 	u32 new_val;
6490 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6491 	struct net *net = (struct net *)ctl->extra2;
6492 	struct ctl_table tmp = {
6493 		.data = &new_val,
6494 		.maxlen = sizeof(new_val),
6495 		.mode = ctl->mode,
6496 	};
6497 
6498 	if (!rtnl_net_trylock(net))
6499 		return restart_syscall();
6500 
6501 	new_val = *((u32 *)ctl->data);
6502 
6503 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6504 	if (ret != 0)
6505 		goto out;
6506 
6507 	if (write) {
6508 		if (check_addr_gen_mode(new_val) < 0) {
6509 			ret = -EINVAL;
6510 			goto out;
6511 		}
6512 
6513 		if (idev) {
6514 			if (check_stable_privacy(idev, net, new_val) < 0) {
6515 				ret = -EINVAL;
6516 				goto out;
6517 			}
6518 
6519 			if (idev->cnf.addr_gen_mode != new_val) {
6520 				WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6521 				netdev_lock_ops(idev->dev);
6522 				addrconf_init_auto_addrs(idev->dev);
6523 				netdev_unlock_ops(idev->dev);
6524 			}
6525 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6526 			struct net_device *dev;
6527 
6528 			WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6529 			for_each_netdev(net, dev) {
6530 				idev = __in6_dev_get_rtnl_net(dev);
6531 				if (idev &&
6532 				    idev->cnf.addr_gen_mode != new_val) {
6533 					WRITE_ONCE(idev->cnf.addr_gen_mode,
6534 						  new_val);
6535 					netdev_lock_ops(idev->dev);
6536 					addrconf_init_auto_addrs(idev->dev);
6537 					netdev_unlock_ops(idev->dev);
6538 				}
6539 			}
6540 		}
6541 
6542 		WRITE_ONCE(*((u32 *)ctl->data), new_val);
6543 	}
6544 
6545 out:
6546 	rtnl_net_unlock(net);
6547 
6548 	return ret;
6549 }
6550 
6551 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6552 					 void *buffer, size_t *lenp,
6553 					 loff_t *ppos)
6554 {
6555 	int err;
6556 	struct in6_addr addr;
6557 	char str[IPV6_MAX_STRLEN];
6558 	struct ctl_table lctl = *ctl;
6559 	struct net *net = ctl->extra2;
6560 	struct ipv6_stable_secret *secret = ctl->data;
6561 
6562 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6563 		return -EIO;
6564 
6565 	lctl.maxlen = IPV6_MAX_STRLEN;
6566 	lctl.data = str;
6567 
6568 	if (!rtnl_net_trylock(net))
6569 		return restart_syscall();
6570 
6571 	if (!write && !secret->initialized) {
6572 		err = -EIO;
6573 		goto out;
6574 	}
6575 
6576 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6577 	if (err >= sizeof(str)) {
6578 		err = -EIO;
6579 		goto out;
6580 	}
6581 
6582 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6583 	if (err || !write)
6584 		goto out;
6585 
6586 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6587 		err = -EIO;
6588 		goto out;
6589 	}
6590 
6591 	secret->initialized = true;
6592 	secret->secret = addr;
6593 
6594 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6595 		struct net_device *dev;
6596 
6597 		for_each_netdev(net, dev) {
6598 			struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev);
6599 
6600 			if (idev) {
6601 				WRITE_ONCE(idev->cnf.addr_gen_mode,
6602 					   IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6603 			}
6604 		}
6605 	} else {
6606 		struct inet6_dev *idev = ctl->extra1;
6607 
6608 		WRITE_ONCE(idev->cnf.addr_gen_mode,
6609 			   IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6610 	}
6611 
6612 out:
6613 	rtnl_net_unlock(net);
6614 
6615 	return err;
6616 }
6617 
6618 static
6619 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6620 						int write, void *buffer,
6621 						size_t *lenp,
6622 						loff_t *ppos)
6623 {
6624 	int *valp = ctl->data;
6625 	int val = *valp;
6626 	loff_t pos = *ppos;
6627 	struct ctl_table lctl;
6628 	int ret;
6629 
6630 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6631 	 * we should not modify it until we get the rtnl lock.
6632 	 */
6633 	lctl = *ctl;
6634 	lctl.data = &val;
6635 
6636 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6637 
6638 	if (write)
6639 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6640 	if (ret)
6641 		*ppos = pos;
6642 	return ret;
6643 }
6644 
6645 static
6646 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6647 {
6648 	if (rt) {
6649 		if (action)
6650 			rt->dst.flags |= DST_NOPOLICY;
6651 		else
6652 			rt->dst.flags &= ~DST_NOPOLICY;
6653 	}
6654 }
6655 
6656 static
6657 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6658 {
6659 	struct inet6_ifaddr *ifa;
6660 
6661 	read_lock_bh(&idev->lock);
6662 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6663 		spin_lock(&ifa->lock);
6664 		if (ifa->rt) {
6665 			/* host routes only use builtin fib6_nh */
6666 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6667 			int cpu;
6668 
6669 			rcu_read_lock();
6670 			ifa->rt->dst_nopolicy = val ? true : false;
6671 			if (nh->rt6i_pcpu) {
6672 				for_each_possible_cpu(cpu) {
6673 					struct rt6_info **rtp;
6674 
6675 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6676 					addrconf_set_nopolicy(*rtp, val);
6677 				}
6678 			}
6679 			rcu_read_unlock();
6680 		}
6681 		spin_unlock(&ifa->lock);
6682 	}
6683 	read_unlock_bh(&idev->lock);
6684 }
6685 
6686 static
6687 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6688 {
6689 	struct net *net = (struct net *)ctl->extra2;
6690 	struct inet6_dev *idev;
6691 
6692 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6693 		WRITE_ONCE(*valp, val);
6694 		return 0;
6695 	}
6696 
6697 	if (!rtnl_net_trylock(net))
6698 		return restart_syscall();
6699 
6700 	WRITE_ONCE(*valp, val);
6701 
6702 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6703 		struct net_device *dev;
6704 
6705 		for_each_netdev(net, dev) {
6706 			idev = __in6_dev_get_rtnl_net(dev);
6707 			if (idev)
6708 				addrconf_disable_policy_idev(idev, val);
6709 		}
6710 	} else {
6711 		idev = (struct inet6_dev *)ctl->extra1;
6712 		addrconf_disable_policy_idev(idev, val);
6713 	}
6714 
6715 	rtnl_net_unlock(net);
6716 	return 0;
6717 }
6718 
6719 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6720 				   void *buffer, size_t *lenp, loff_t *ppos)
6721 {
6722 	int *valp = ctl->data;
6723 	int val = *valp;
6724 	loff_t pos = *ppos;
6725 	struct ctl_table lctl;
6726 	int ret;
6727 
6728 	lctl = *ctl;
6729 	lctl.data = &val;
6730 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6731 
6732 	if (write && (*valp != val))
6733 		ret = addrconf_disable_policy(ctl, valp, val);
6734 
6735 	if (ret)
6736 		*ppos = pos;
6737 
6738 	return ret;
6739 }
6740 
6741 static int minus_one = -1;
6742 static const int two_five_five = 255;
6743 static u32 ioam6_if_id_max = U16_MAX;
6744 
6745 static const struct ctl_table addrconf_sysctl[] = {
6746 	{
6747 		.procname	= "forwarding",
6748 		.data		= &ipv6_devconf.forwarding,
6749 		.maxlen		= sizeof(int),
6750 		.mode		= 0644,
6751 		.proc_handler	= addrconf_sysctl_forward,
6752 	},
6753 	{
6754 		.procname	= "hop_limit",
6755 		.data		= &ipv6_devconf.hop_limit,
6756 		.maxlen		= sizeof(int),
6757 		.mode		= 0644,
6758 		.proc_handler	= proc_dointvec_minmax,
6759 		.extra1		= (void *)SYSCTL_ONE,
6760 		.extra2		= (void *)&two_five_five,
6761 	},
6762 	{
6763 		.procname	= "mtu",
6764 		.data		= &ipv6_devconf.mtu6,
6765 		.maxlen		= sizeof(int),
6766 		.mode		= 0644,
6767 		.proc_handler	= addrconf_sysctl_mtu,
6768 	},
6769 	{
6770 		.procname	= "accept_ra",
6771 		.data		= &ipv6_devconf.accept_ra,
6772 		.maxlen		= sizeof(int),
6773 		.mode		= 0644,
6774 		.proc_handler	= proc_dointvec,
6775 	},
6776 	{
6777 		.procname	= "accept_redirects",
6778 		.data		= &ipv6_devconf.accept_redirects,
6779 		.maxlen		= sizeof(int),
6780 		.mode		= 0644,
6781 		.proc_handler	= proc_dointvec,
6782 	},
6783 	{
6784 		.procname	= "autoconf",
6785 		.data		= &ipv6_devconf.autoconf,
6786 		.maxlen		= sizeof(int),
6787 		.mode		= 0644,
6788 		.proc_handler	= proc_dointvec,
6789 	},
6790 	{
6791 		.procname	= "dad_transmits",
6792 		.data		= &ipv6_devconf.dad_transmits,
6793 		.maxlen		= sizeof(int),
6794 		.mode		= 0644,
6795 		.proc_handler	= proc_dointvec,
6796 	},
6797 	{
6798 		.procname	= "router_solicitations",
6799 		.data		= &ipv6_devconf.rtr_solicits,
6800 		.maxlen		= sizeof(int),
6801 		.mode		= 0644,
6802 		.proc_handler	= proc_dointvec_minmax,
6803 		.extra1		= &minus_one,
6804 	},
6805 	{
6806 		.procname	= "router_solicitation_interval",
6807 		.data		= &ipv6_devconf.rtr_solicit_interval,
6808 		.maxlen		= sizeof(int),
6809 		.mode		= 0644,
6810 		.proc_handler	= proc_dointvec_jiffies,
6811 	},
6812 	{
6813 		.procname	= "router_solicitation_max_interval",
6814 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6815 		.maxlen		= sizeof(int),
6816 		.mode		= 0644,
6817 		.proc_handler	= proc_dointvec_jiffies,
6818 	},
6819 	{
6820 		.procname	= "router_solicitation_delay",
6821 		.data		= &ipv6_devconf.rtr_solicit_delay,
6822 		.maxlen		= sizeof(int),
6823 		.mode		= 0644,
6824 		.proc_handler	= proc_dointvec_jiffies,
6825 	},
6826 	{
6827 		.procname	= "force_mld_version",
6828 		.data		= &ipv6_devconf.force_mld_version,
6829 		.maxlen		= sizeof(int),
6830 		.mode		= 0644,
6831 		.proc_handler	= proc_dointvec,
6832 	},
6833 	{
6834 		.procname	= "mldv1_unsolicited_report_interval",
6835 		.data		=
6836 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6837 		.maxlen		= sizeof(int),
6838 		.mode		= 0644,
6839 		.proc_handler	= proc_dointvec_ms_jiffies,
6840 	},
6841 	{
6842 		.procname	= "mldv2_unsolicited_report_interval",
6843 		.data		=
6844 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6845 		.maxlen		= sizeof(int),
6846 		.mode		= 0644,
6847 		.proc_handler	= proc_dointvec_ms_jiffies,
6848 	},
6849 	{
6850 		.procname	= "use_tempaddr",
6851 		.data		= &ipv6_devconf.use_tempaddr,
6852 		.maxlen		= sizeof(int),
6853 		.mode		= 0644,
6854 		.proc_handler	= proc_dointvec,
6855 	},
6856 	{
6857 		.procname	= "temp_valid_lft",
6858 		.data		= &ipv6_devconf.temp_valid_lft,
6859 		.maxlen		= sizeof(int),
6860 		.mode		= 0644,
6861 		.proc_handler	= proc_dointvec,
6862 	},
6863 	{
6864 		.procname	= "temp_prefered_lft",
6865 		.data		= &ipv6_devconf.temp_prefered_lft,
6866 		.maxlen		= sizeof(int),
6867 		.mode		= 0644,
6868 		.proc_handler	= proc_dointvec,
6869 	},
6870 	{
6871 		.procname       = "regen_min_advance",
6872 		.data           = &ipv6_devconf.regen_min_advance,
6873 		.maxlen         = sizeof(int),
6874 		.mode           = 0644,
6875 		.proc_handler   = proc_dointvec,
6876 	},
6877 	{
6878 		.procname	= "regen_max_retry",
6879 		.data		= &ipv6_devconf.regen_max_retry,
6880 		.maxlen		= sizeof(int),
6881 		.mode		= 0644,
6882 		.proc_handler	= proc_dointvec,
6883 	},
6884 	{
6885 		.procname	= "max_desync_factor",
6886 		.data		= &ipv6_devconf.max_desync_factor,
6887 		.maxlen		= sizeof(int),
6888 		.mode		= 0644,
6889 		.proc_handler	= proc_dointvec,
6890 	},
6891 	{
6892 		.procname	= "max_addresses",
6893 		.data		= &ipv6_devconf.max_addresses,
6894 		.maxlen		= sizeof(int),
6895 		.mode		= 0644,
6896 		.proc_handler	= proc_dointvec,
6897 	},
6898 	{
6899 		.procname	= "accept_ra_defrtr",
6900 		.data		= &ipv6_devconf.accept_ra_defrtr,
6901 		.maxlen		= sizeof(int),
6902 		.mode		= 0644,
6903 		.proc_handler	= proc_dointvec,
6904 	},
6905 	{
6906 		.procname	= "ra_defrtr_metric",
6907 		.data		= &ipv6_devconf.ra_defrtr_metric,
6908 		.maxlen		= sizeof(u32),
6909 		.mode		= 0644,
6910 		.proc_handler	= proc_douintvec_minmax,
6911 		.extra1		= (void *)SYSCTL_ONE,
6912 	},
6913 	{
6914 		.procname	= "accept_ra_min_hop_limit",
6915 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6916 		.maxlen		= sizeof(int),
6917 		.mode		= 0644,
6918 		.proc_handler	= proc_dointvec,
6919 	},
6920 	{
6921 		.procname	= "accept_ra_min_lft",
6922 		.data		= &ipv6_devconf.accept_ra_min_lft,
6923 		.maxlen		= sizeof(int),
6924 		.mode		= 0644,
6925 		.proc_handler	= proc_dointvec,
6926 	},
6927 	{
6928 		.procname	= "accept_ra_pinfo",
6929 		.data		= &ipv6_devconf.accept_ra_pinfo,
6930 		.maxlen		= sizeof(int),
6931 		.mode		= 0644,
6932 		.proc_handler	= proc_dointvec,
6933 	},
6934 	{
6935 		.procname	= "ra_honor_pio_life",
6936 		.data		= &ipv6_devconf.ra_honor_pio_life,
6937 		.maxlen		= sizeof(u8),
6938 		.mode		= 0644,
6939 		.proc_handler	= proc_dou8vec_minmax,
6940 		.extra1		= SYSCTL_ZERO,
6941 		.extra2		= SYSCTL_ONE,
6942 	},
6943 	{
6944 		.procname	= "ra_honor_pio_pflag",
6945 		.data		= &ipv6_devconf.ra_honor_pio_pflag,
6946 		.maxlen		= sizeof(u8),
6947 		.mode		= 0644,
6948 		.proc_handler	= proc_dou8vec_minmax,
6949 		.extra1		= SYSCTL_ZERO,
6950 		.extra2		= SYSCTL_ONE,
6951 	},
6952 #ifdef CONFIG_IPV6_ROUTER_PREF
6953 	{
6954 		.procname	= "accept_ra_rtr_pref",
6955 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6956 		.maxlen		= sizeof(int),
6957 		.mode		= 0644,
6958 		.proc_handler	= proc_dointvec,
6959 	},
6960 	{
6961 		.procname	= "router_probe_interval",
6962 		.data		= &ipv6_devconf.rtr_probe_interval,
6963 		.maxlen		= sizeof(int),
6964 		.mode		= 0644,
6965 		.proc_handler	= proc_dointvec_jiffies,
6966 	},
6967 #ifdef CONFIG_IPV6_ROUTE_INFO
6968 	{
6969 		.procname	= "accept_ra_rt_info_min_plen",
6970 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6971 		.maxlen		= sizeof(int),
6972 		.mode		= 0644,
6973 		.proc_handler	= proc_dointvec,
6974 	},
6975 	{
6976 		.procname	= "accept_ra_rt_info_max_plen",
6977 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6978 		.maxlen		= sizeof(int),
6979 		.mode		= 0644,
6980 		.proc_handler	= proc_dointvec,
6981 	},
6982 #endif
6983 #endif
6984 	{
6985 		.procname	= "proxy_ndp",
6986 		.data		= &ipv6_devconf.proxy_ndp,
6987 		.maxlen		= sizeof(int),
6988 		.mode		= 0644,
6989 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6990 	},
6991 	{
6992 		.procname	= "accept_source_route",
6993 		.data		= &ipv6_devconf.accept_source_route,
6994 		.maxlen		= sizeof(int),
6995 		.mode		= 0644,
6996 		.proc_handler	= proc_dointvec,
6997 	},
6998 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6999 	{
7000 		.procname	= "optimistic_dad",
7001 		.data		= &ipv6_devconf.optimistic_dad,
7002 		.maxlen		= sizeof(int),
7003 		.mode		= 0644,
7004 		.proc_handler   = proc_dointvec,
7005 	},
7006 	{
7007 		.procname	= "use_optimistic",
7008 		.data		= &ipv6_devconf.use_optimistic,
7009 		.maxlen		= sizeof(int),
7010 		.mode		= 0644,
7011 		.proc_handler	= proc_dointvec,
7012 	},
7013 #endif
7014 #ifdef CONFIG_IPV6_MROUTE
7015 	{
7016 		.procname	= "mc_forwarding",
7017 		.data		= &ipv6_devconf.mc_forwarding,
7018 		.maxlen		= sizeof(int),
7019 		.mode		= 0444,
7020 		.proc_handler	= proc_dointvec,
7021 	},
7022 #endif
7023 	{
7024 		.procname	= "disable_ipv6",
7025 		.data		= &ipv6_devconf.disable_ipv6,
7026 		.maxlen		= sizeof(int),
7027 		.mode		= 0644,
7028 		.proc_handler	= addrconf_sysctl_disable,
7029 	},
7030 	{
7031 		.procname	= "accept_dad",
7032 		.data		= &ipv6_devconf.accept_dad,
7033 		.maxlen		= sizeof(int),
7034 		.mode		= 0644,
7035 		.proc_handler	= proc_dointvec,
7036 	},
7037 	{
7038 		.procname	= "force_tllao",
7039 		.data		= &ipv6_devconf.force_tllao,
7040 		.maxlen		= sizeof(int),
7041 		.mode		= 0644,
7042 		.proc_handler	= proc_dointvec
7043 	},
7044 	{
7045 		.procname	= "ndisc_notify",
7046 		.data		= &ipv6_devconf.ndisc_notify,
7047 		.maxlen		= sizeof(int),
7048 		.mode		= 0644,
7049 		.proc_handler	= proc_dointvec
7050 	},
7051 	{
7052 		.procname	= "suppress_frag_ndisc",
7053 		.data		= &ipv6_devconf.suppress_frag_ndisc,
7054 		.maxlen		= sizeof(int),
7055 		.mode		= 0644,
7056 		.proc_handler	= proc_dointvec
7057 	},
7058 	{
7059 		.procname	= "accept_ra_from_local",
7060 		.data		= &ipv6_devconf.accept_ra_from_local,
7061 		.maxlen		= sizeof(int),
7062 		.mode		= 0644,
7063 		.proc_handler	= proc_dointvec,
7064 	},
7065 	{
7066 		.procname	= "accept_ra_mtu",
7067 		.data		= &ipv6_devconf.accept_ra_mtu,
7068 		.maxlen		= sizeof(int),
7069 		.mode		= 0644,
7070 		.proc_handler	= proc_dointvec,
7071 	},
7072 	{
7073 		.procname	= "stable_secret",
7074 		.data		= &ipv6_devconf.stable_secret,
7075 		.maxlen		= IPV6_MAX_STRLEN,
7076 		.mode		= 0600,
7077 		.proc_handler	= addrconf_sysctl_stable_secret,
7078 	},
7079 	{
7080 		.procname	= "use_oif_addrs_only",
7081 		.data		= &ipv6_devconf.use_oif_addrs_only,
7082 		.maxlen		= sizeof(int),
7083 		.mode		= 0644,
7084 		.proc_handler	= proc_dointvec,
7085 	},
7086 	{
7087 		.procname	= "ignore_routes_with_linkdown",
7088 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
7089 		.maxlen		= sizeof(int),
7090 		.mode		= 0644,
7091 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
7092 	},
7093 	{
7094 		.procname	= "drop_unicast_in_l2_multicast",
7095 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
7096 		.maxlen		= sizeof(int),
7097 		.mode		= 0644,
7098 		.proc_handler	= proc_dointvec,
7099 	},
7100 	{
7101 		.procname	= "drop_unsolicited_na",
7102 		.data		= &ipv6_devconf.drop_unsolicited_na,
7103 		.maxlen		= sizeof(int),
7104 		.mode		= 0644,
7105 		.proc_handler	= proc_dointvec,
7106 	},
7107 	{
7108 		.procname	= "keep_addr_on_down",
7109 		.data		= &ipv6_devconf.keep_addr_on_down,
7110 		.maxlen		= sizeof(int),
7111 		.mode		= 0644,
7112 		.proc_handler	= proc_dointvec,
7113 
7114 	},
7115 	{
7116 		.procname	= "seg6_enabled",
7117 		.data		= &ipv6_devconf.seg6_enabled,
7118 		.maxlen		= sizeof(int),
7119 		.mode		= 0644,
7120 		.proc_handler	= proc_dointvec,
7121 	},
7122 #ifdef CONFIG_IPV6_SEG6_HMAC
7123 	{
7124 		.procname	= "seg6_require_hmac",
7125 		.data		= &ipv6_devconf.seg6_require_hmac,
7126 		.maxlen		= sizeof(int),
7127 		.mode		= 0644,
7128 		.proc_handler	= proc_dointvec,
7129 	},
7130 #endif
7131 	{
7132 		.procname       = "enhanced_dad",
7133 		.data           = &ipv6_devconf.enhanced_dad,
7134 		.maxlen         = sizeof(int),
7135 		.mode           = 0644,
7136 		.proc_handler   = proc_dointvec,
7137 	},
7138 	{
7139 		.procname	= "addr_gen_mode",
7140 		.data		= &ipv6_devconf.addr_gen_mode,
7141 		.maxlen		= sizeof(int),
7142 		.mode		= 0644,
7143 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
7144 	},
7145 	{
7146 		.procname       = "disable_policy",
7147 		.data           = &ipv6_devconf.disable_policy,
7148 		.maxlen         = sizeof(int),
7149 		.mode           = 0644,
7150 		.proc_handler   = addrconf_sysctl_disable_policy,
7151 	},
7152 	{
7153 		.procname	= "ndisc_tclass",
7154 		.data		= &ipv6_devconf.ndisc_tclass,
7155 		.maxlen		= sizeof(int),
7156 		.mode		= 0644,
7157 		.proc_handler	= proc_dointvec_minmax,
7158 		.extra1		= (void *)SYSCTL_ZERO,
7159 		.extra2		= (void *)&two_five_five,
7160 	},
7161 	{
7162 		.procname	= "rpl_seg_enabled",
7163 		.data		= &ipv6_devconf.rpl_seg_enabled,
7164 		.maxlen		= sizeof(int),
7165 		.mode		= 0644,
7166 		.proc_handler	= proc_dointvec,
7167 	},
7168 	{
7169 		.procname	= "ioam6_enabled",
7170 		.data		= &ipv6_devconf.ioam6_enabled,
7171 		.maxlen		= sizeof(u8),
7172 		.mode		= 0644,
7173 		.proc_handler	= proc_dou8vec_minmax,
7174 		.extra1		= (void *)SYSCTL_ZERO,
7175 		.extra2		= (void *)SYSCTL_ONE,
7176 	},
7177 	{
7178 		.procname	= "ioam6_id",
7179 		.data		= &ipv6_devconf.ioam6_id,
7180 		.maxlen		= sizeof(u32),
7181 		.mode		= 0644,
7182 		.proc_handler	= proc_douintvec_minmax,
7183 		.extra1		= (void *)SYSCTL_ZERO,
7184 		.extra2		= (void *)&ioam6_if_id_max,
7185 	},
7186 	{
7187 		.procname	= "ioam6_id_wide",
7188 		.data		= &ipv6_devconf.ioam6_id_wide,
7189 		.maxlen		= sizeof(u32),
7190 		.mode		= 0644,
7191 		.proc_handler	= proc_douintvec,
7192 	},
7193 	{
7194 		.procname	= "ndisc_evict_nocarrier",
7195 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
7196 		.maxlen		= sizeof(u8),
7197 		.mode		= 0644,
7198 		.proc_handler	= proc_dou8vec_minmax,
7199 		.extra1		= (void *)SYSCTL_ZERO,
7200 		.extra2		= (void *)SYSCTL_ONE,
7201 	},
7202 	{
7203 		.procname	= "accept_untracked_na",
7204 		.data		= &ipv6_devconf.accept_untracked_na,
7205 		.maxlen		= sizeof(int),
7206 		.mode		= 0644,
7207 		.proc_handler	= proc_dointvec_minmax,
7208 		.extra1		= SYSCTL_ZERO,
7209 		.extra2		= SYSCTL_TWO,
7210 	},
7211 };
7212 
7213 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7214 		struct inet6_dev *idev, struct ipv6_devconf *p)
7215 {
7216 	size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7217 	int i, ifindex;
7218 	struct ctl_table *table;
7219 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7220 
7221 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7222 	if (!table)
7223 		goto out;
7224 
7225 	for (i = 0; i < table_size; i++) {
7226 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7227 		/* If one of these is already set, then it is not safe to
7228 		 * overwrite either of them: this makes proc_dointvec_minmax
7229 		 * usable.
7230 		 */
7231 		if (!table[i].extra1 && !table[i].extra2) {
7232 			table[i].extra1 = idev; /* embedded; no ref */
7233 			table[i].extra2 = net;
7234 		}
7235 	}
7236 
7237 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7238 
7239 	p->sysctl_header = register_net_sysctl_sz(net, path, table,
7240 						  table_size);
7241 	if (!p->sysctl_header)
7242 		goto free;
7243 
7244 	if (!strcmp(dev_name, "all"))
7245 		ifindex = NETCONFA_IFINDEX_ALL;
7246 	else if (!strcmp(dev_name, "default"))
7247 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7248 	else
7249 		ifindex = idev->dev->ifindex;
7250 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7251 				     ifindex, p);
7252 	return 0;
7253 
7254 free:
7255 	kfree(table);
7256 out:
7257 	return -ENOBUFS;
7258 }
7259 
7260 static void __addrconf_sysctl_unregister(struct net *net,
7261 					 struct ipv6_devconf *p, int ifindex)
7262 {
7263 	const struct ctl_table *table;
7264 
7265 	if (!p->sysctl_header)
7266 		return;
7267 
7268 	table = p->sysctl_header->ctl_table_arg;
7269 	unregister_net_sysctl_table(p->sysctl_header);
7270 	p->sysctl_header = NULL;
7271 	kfree(table);
7272 
7273 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7274 }
7275 
7276 static int addrconf_sysctl_register(struct inet6_dev *idev)
7277 {
7278 	int err;
7279 
7280 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7281 		return -EINVAL;
7282 
7283 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7284 				    &ndisc_ifinfo_sysctl_change);
7285 	if (err)
7286 		return err;
7287 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7288 					 idev, &idev->cnf);
7289 	if (err)
7290 		neigh_sysctl_unregister(idev->nd_parms);
7291 
7292 	return err;
7293 }
7294 
7295 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7296 {
7297 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7298 				     idev->dev->ifindex);
7299 	neigh_sysctl_unregister(idev->nd_parms);
7300 }
7301 
7302 
7303 #endif
7304 
7305 static int __net_init addrconf_init_net(struct net *net)
7306 {
7307 	int err = -ENOMEM;
7308 	struct ipv6_devconf *all, *dflt;
7309 
7310 	spin_lock_init(&net->ipv6.addrconf_hash_lock);
7311 	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7312 	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7313 					   sizeof(struct hlist_head),
7314 					   GFP_KERNEL);
7315 	if (!net->ipv6.inet6_addr_lst)
7316 		goto err_alloc_addr;
7317 
7318 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7319 	if (!all)
7320 		goto err_alloc_all;
7321 
7322 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7323 	if (!dflt)
7324 		goto err_alloc_dflt;
7325 
7326 	if (!net_eq(net, &init_net)) {
7327 		switch (net_inherit_devconf()) {
7328 		case 1:  /* copy from init_net */
7329 			memcpy(all, init_net.ipv6.devconf_all,
7330 			       sizeof(ipv6_devconf));
7331 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7332 			       sizeof(ipv6_devconf_dflt));
7333 			break;
7334 		case 3: /* copy from the current netns */
7335 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7336 			       sizeof(ipv6_devconf));
7337 			memcpy(dflt,
7338 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7339 			       sizeof(ipv6_devconf_dflt));
7340 			break;
7341 		case 0:
7342 		case 2:
7343 			/* use compiled values */
7344 			break;
7345 		}
7346 	}
7347 
7348 	/* these will be inherited by all namespaces */
7349 	dflt->autoconf = ipv6_defaults.autoconf;
7350 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7351 
7352 	dflt->stable_secret.initialized = false;
7353 	all->stable_secret.initialized = false;
7354 
7355 	net->ipv6.devconf_all = all;
7356 	net->ipv6.devconf_dflt = dflt;
7357 
7358 #ifdef CONFIG_SYSCTL
7359 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7360 	if (err < 0)
7361 		goto err_reg_all;
7362 
7363 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7364 	if (err < 0)
7365 		goto err_reg_dflt;
7366 #endif
7367 	return 0;
7368 
7369 #ifdef CONFIG_SYSCTL
7370 err_reg_dflt:
7371 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7372 err_reg_all:
7373 	kfree(dflt);
7374 	net->ipv6.devconf_dflt = NULL;
7375 #endif
7376 err_alloc_dflt:
7377 	kfree(all);
7378 	net->ipv6.devconf_all = NULL;
7379 err_alloc_all:
7380 	kfree(net->ipv6.inet6_addr_lst);
7381 err_alloc_addr:
7382 	return err;
7383 }
7384 
7385 static void __net_exit addrconf_exit_net(struct net *net)
7386 {
7387 	int i;
7388 
7389 #ifdef CONFIG_SYSCTL
7390 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7391 				     NETCONFA_IFINDEX_DEFAULT);
7392 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7393 				     NETCONFA_IFINDEX_ALL);
7394 #endif
7395 	kfree(net->ipv6.devconf_dflt);
7396 	net->ipv6.devconf_dflt = NULL;
7397 	kfree(net->ipv6.devconf_all);
7398 	net->ipv6.devconf_all = NULL;
7399 
7400 	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7401 	/*
7402 	 *	Check hash table, then free it.
7403 	 */
7404 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7405 		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7406 
7407 	kfree(net->ipv6.inet6_addr_lst);
7408 	net->ipv6.inet6_addr_lst = NULL;
7409 }
7410 
7411 static struct pernet_operations addrconf_ops = {
7412 	.init = addrconf_init_net,
7413 	.exit = addrconf_exit_net,
7414 };
7415 
7416 static struct rtnl_af_ops inet6_ops __read_mostly = {
7417 	.family		  = AF_INET6,
7418 	.fill_link_af	  = inet6_fill_link_af,
7419 	.get_link_af_size = inet6_get_link_af_size,
7420 	.validate_link_af = inet6_validate_link_af,
7421 	.set_link_af	  = inet6_set_link_af,
7422 };
7423 
7424 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = {
7425 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK,
7426 	 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED},
7427 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR,
7428 	 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET},
7429 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR,
7430 	 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET},
7431 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR,
7432 	 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr,
7433 	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7434 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST,
7435 	 .dumpit = inet6_dump_ifmcaddr,
7436 	 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7437 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST,
7438 	 .dumpit = inet6_dump_ifacaddr,
7439 	 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7440 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF,
7441 	 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf,
7442 	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7443 };
7444 
7445 /*
7446  *	Init / cleanup code
7447  */
7448 
7449 int __init addrconf_init(void)
7450 {
7451 	struct inet6_dev *idev;
7452 	int err;
7453 
7454 	err = ipv6_addr_label_init();
7455 	if (err < 0) {
7456 		pr_crit("%s: cannot initialize default policy table: %d\n",
7457 			__func__, err);
7458 		goto out;
7459 	}
7460 
7461 	err = register_pernet_subsys(&addrconf_ops);
7462 	if (err < 0)
7463 		goto out_addrlabel;
7464 
7465 	/* All works using addrconf_wq need to lock rtnl. */
7466 	addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7467 	if (!addrconf_wq) {
7468 		err = -ENOMEM;
7469 		goto out_nowq;
7470 	}
7471 
7472 	rtnl_net_lock(&init_net);
7473 	idev = ipv6_add_dev(blackhole_netdev);
7474 	rtnl_net_unlock(&init_net);
7475 	if (IS_ERR(idev)) {
7476 		err = PTR_ERR(idev);
7477 		goto errlo;
7478 	}
7479 
7480 	ip6_route_init_special_entries();
7481 
7482 	register_netdevice_notifier(&ipv6_dev_notf);
7483 
7484 	addrconf_verify(&init_net);
7485 
7486 	err = rtnl_af_register(&inet6_ops);
7487 	if (err)
7488 		goto erraf;
7489 
7490 	err = rtnl_register_many(addrconf_rtnl_msg_handlers);
7491 	if (err)
7492 		goto errout;
7493 
7494 	err = ipv6_addr_label_rtnl_register();
7495 	if (err < 0)
7496 		goto errout;
7497 
7498 	return 0;
7499 errout:
7500 	rtnl_unregister_all(PF_INET6);
7501 	rtnl_af_unregister(&inet6_ops);
7502 erraf:
7503 	unregister_netdevice_notifier(&ipv6_dev_notf);
7504 errlo:
7505 	destroy_workqueue(addrconf_wq);
7506 out_nowq:
7507 	unregister_pernet_subsys(&addrconf_ops);
7508 out_addrlabel:
7509 	ipv6_addr_label_cleanup();
7510 out:
7511 	return err;
7512 }
7513 
7514 void addrconf_cleanup(void)
7515 {
7516 	struct net_device *dev;
7517 
7518 	unregister_netdevice_notifier(&ipv6_dev_notf);
7519 	unregister_pernet_subsys(&addrconf_ops);
7520 	ipv6_addr_label_cleanup();
7521 
7522 	rtnl_af_unregister(&inet6_ops);
7523 
7524 	rtnl_net_lock(&init_net);
7525 
7526 	/* clean dev list */
7527 	for_each_netdev(&init_net, dev) {
7528 		if (!__in6_dev_get_rtnl_net(dev))
7529 			continue;
7530 		addrconf_ifdown(dev, true);
7531 	}
7532 	addrconf_ifdown(init_net.loopback_dev, true);
7533 
7534 	rtnl_net_unlock(&init_net);
7535 
7536 	destroy_workqueue(addrconf_wq);
7537 }
7538