xref: /linux/net/ipv6/addrconf.c (revision 6e7fd890f1d6ac83805409e9c346240de2705584)
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 <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
89 
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
93 #include <linux/ioam6.h>
94 
95 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
96 
97 #define IPV6_MAX_STRLEN \
98 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
99 
100 static inline u32 cstamp_delta(unsigned long cstamp)
101 {
102 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
103 }
104 
105 static inline s32 rfc3315_s14_backoff_init(s32 irt)
106 {
107 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
108 	u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
109 	do_div(tmp, 1000000);
110 	return (s32)tmp;
111 }
112 
113 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
114 {
115 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
116 	u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
117 	do_div(tmp, 1000000);
118 	if ((s32)tmp > mrt) {
119 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
120 		tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
121 		do_div(tmp, 1000000);
122 	}
123 	return (s32)tmp;
124 }
125 
126 #ifdef CONFIG_SYSCTL
127 static int addrconf_sysctl_register(struct inet6_dev *idev);
128 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
129 #else
130 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
131 {
132 	return 0;
133 }
134 
135 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
136 {
137 }
138 #endif
139 
140 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
141 
142 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
143 static int ipv6_count_addresses(const struct inet6_dev *idev);
144 static int ipv6_generate_stable_address(struct in6_addr *addr,
145 					u8 dad_count,
146 					const struct inet6_dev *idev);
147 
148 #define IN6_ADDR_HSIZE_SHIFT	8
149 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
150 
151 static void addrconf_verify(struct net *net);
152 static void addrconf_verify_rtnl(struct net *net);
153 
154 static struct workqueue_struct *addrconf_wq;
155 
156 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
157 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
158 
159 static void addrconf_type_change(struct net_device *dev,
160 				 unsigned long event);
161 static int addrconf_ifdown(struct net_device *dev, bool unregister);
162 
163 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
164 						  int plen,
165 						  const struct net_device *dev,
166 						  u32 flags, u32 noflags,
167 						  bool no_gw);
168 
169 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
170 static void addrconf_dad_work(struct work_struct *w);
171 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
172 				   bool send_na);
173 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
174 static void addrconf_rs_timer(struct timer_list *t);
175 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
177 
178 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
179 				struct prefix_info *pinfo);
180 
181 static struct ipv6_devconf ipv6_devconf __read_mostly = {
182 	.forwarding		= 0,
183 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
184 	.mtu6			= IPV6_MIN_MTU,
185 	.accept_ra		= 1,
186 	.accept_redirects	= 1,
187 	.autoconf		= 1,
188 	.force_mld_version	= 0,
189 	.mldv1_unsolicited_report_interval = 10 * HZ,
190 	.mldv2_unsolicited_report_interval = HZ,
191 	.dad_transmits		= 1,
192 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
193 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
194 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
195 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
196 	.use_tempaddr		= 0,
197 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
198 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
199 	.regen_min_advance	= REGEN_MIN_ADVANCE,
200 	.regen_max_retry	= REGEN_MAX_RETRY,
201 	.max_desync_factor	= MAX_DESYNC_FACTOR,
202 	.max_addresses		= IPV6_MAX_ADDRESSES,
203 	.accept_ra_defrtr	= 1,
204 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
205 	.accept_ra_from_local	= 0,
206 	.accept_ra_min_hop_limit= 1,
207 	.accept_ra_min_lft	= 0,
208 	.accept_ra_pinfo	= 1,
209 #ifdef CONFIG_IPV6_ROUTER_PREF
210 	.accept_ra_rtr_pref	= 1,
211 	.rtr_probe_interval	= 60 * HZ,
212 #ifdef CONFIG_IPV6_ROUTE_INFO
213 	.accept_ra_rt_info_min_plen = 0,
214 	.accept_ra_rt_info_max_plen = 0,
215 #endif
216 #endif
217 	.proxy_ndp		= 0,
218 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
219 	.disable_ipv6		= 0,
220 	.accept_dad		= 0,
221 	.suppress_frag_ndisc	= 1,
222 	.accept_ra_mtu		= 1,
223 	.stable_secret		= {
224 		.initialized = false,
225 	},
226 	.use_oif_addrs_only	= 0,
227 	.ignore_routes_with_linkdown = 0,
228 	.keep_addr_on_down	= 0,
229 	.seg6_enabled		= 0,
230 #ifdef CONFIG_IPV6_SEG6_HMAC
231 	.seg6_require_hmac	= 0,
232 #endif
233 	.enhanced_dad           = 1,
234 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
235 	.disable_policy		= 0,
236 	.rpl_seg_enabled	= 0,
237 	.ioam6_enabled		= 0,
238 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
239 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
240 	.ndisc_evict_nocarrier	= 1,
241 	.ra_honor_pio_life	= 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 };
306 
307 /* Check if link is ready: is it up and is a valid qdisc available */
308 static inline bool addrconf_link_ready(const struct net_device *dev)
309 {
310 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
311 }
312 
313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
314 {
315 	if (del_timer(&idev->rs_timer))
316 		__in6_dev_put(idev);
317 }
318 
319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
320 {
321 	if (cancel_delayed_work(&ifp->dad_work))
322 		__in6_ifa_put(ifp);
323 }
324 
325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
326 				  unsigned long when)
327 {
328 	if (!mod_timer(&idev->rs_timer, jiffies + when))
329 		in6_dev_hold(idev);
330 }
331 
332 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
333 				   unsigned long delay)
334 {
335 	in6_ifa_hold(ifp);
336 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
337 		in6_ifa_put(ifp);
338 }
339 
340 static int snmp6_alloc_dev(struct inet6_dev *idev)
341 {
342 	int i;
343 
344 	idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
345 	if (!idev->stats.ipv6)
346 		goto err_ip;
347 
348 	for_each_possible_cpu(i) {
349 		struct ipstats_mib *addrconf_stats;
350 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
351 		u64_stats_init(&addrconf_stats->syncp);
352 	}
353 
354 
355 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
356 					GFP_KERNEL);
357 	if (!idev->stats.icmpv6dev)
358 		goto err_icmp;
359 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
360 					   GFP_KERNEL_ACCOUNT);
361 	if (!idev->stats.icmpv6msgdev)
362 		goto err_icmpmsg;
363 
364 	return 0;
365 
366 err_icmpmsg:
367 	kfree(idev->stats.icmpv6dev);
368 err_icmp:
369 	free_percpu(idev->stats.ipv6);
370 err_ip:
371 	return -ENOMEM;
372 }
373 
374 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
375 {
376 	struct inet6_dev *ndev;
377 	int err = -ENOMEM;
378 
379 	ASSERT_RTNL();
380 
381 	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
382 		return ERR_PTR(-EINVAL);
383 
384 	ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
385 	if (!ndev)
386 		return ERR_PTR(err);
387 
388 	rwlock_init(&ndev->lock);
389 	ndev->dev = dev;
390 	INIT_LIST_HEAD(&ndev->addr_list);
391 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
392 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
393 
394 	if (ndev->cnf.stable_secret.initialized)
395 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
396 
397 	ndev->cnf.mtu6 = dev->mtu;
398 	ndev->ra_mtu = 0;
399 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
400 	if (!ndev->nd_parms) {
401 		kfree(ndev);
402 		return ERR_PTR(err);
403 	}
404 	if (ndev->cnf.forwarding)
405 		dev_disable_lro(dev);
406 	/* We refer to the device */
407 	netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
408 
409 	if (snmp6_alloc_dev(ndev) < 0) {
410 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
411 			   __func__);
412 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
413 		netdev_put(dev, &ndev->dev_tracker);
414 		kfree(ndev);
415 		return ERR_PTR(err);
416 	}
417 
418 	if (dev != blackhole_netdev) {
419 		if (snmp6_register_dev(ndev) < 0) {
420 			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
421 				   __func__, dev->name);
422 			goto err_release;
423 		}
424 	}
425 	/* One reference from device. */
426 	refcount_set(&ndev->refcnt, 1);
427 
428 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
429 		ndev->cnf.accept_dad = -1;
430 
431 #if IS_ENABLED(CONFIG_IPV6_SIT)
432 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
433 		pr_info("%s: Disabled Multicast RS\n", dev->name);
434 		ndev->cnf.rtr_solicits = 0;
435 	}
436 #endif
437 
438 	INIT_LIST_HEAD(&ndev->tempaddr_list);
439 	ndev->desync_factor = U32_MAX;
440 	if ((dev->flags&IFF_LOOPBACK) ||
441 	    dev->type == ARPHRD_TUNNEL ||
442 	    dev->type == ARPHRD_TUNNEL6 ||
443 	    dev->type == ARPHRD_SIT ||
444 	    dev->type == ARPHRD_NONE) {
445 		ndev->cnf.use_tempaddr = -1;
446 	}
447 
448 	ndev->token = in6addr_any;
449 
450 	if (netif_running(dev) && addrconf_link_ready(dev))
451 		ndev->if_flags |= IF_READY;
452 
453 	ipv6_mc_init_dev(ndev);
454 	ndev->tstamp = jiffies;
455 	if (dev != blackhole_netdev) {
456 		err = addrconf_sysctl_register(ndev);
457 		if (err) {
458 			ipv6_mc_destroy_dev(ndev);
459 			snmp6_unregister_dev(ndev);
460 			goto err_release;
461 		}
462 	}
463 	/* protected by rtnl_lock */
464 	rcu_assign_pointer(dev->ip6_ptr, ndev);
465 
466 	if (dev != blackhole_netdev) {
467 		/* Join interface-local all-node multicast group */
468 		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
469 
470 		/* Join all-node multicast group */
471 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
472 
473 		/* Join all-router multicast group if forwarding is set */
474 		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
475 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
476 	}
477 	return ndev;
478 
479 err_release:
480 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
481 	ndev->dead = 1;
482 	in6_dev_finish_destroy(ndev);
483 	return ERR_PTR(err);
484 }
485 
486 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
487 {
488 	struct inet6_dev *idev;
489 
490 	ASSERT_RTNL();
491 
492 	idev = __in6_dev_get(dev);
493 	if (!idev) {
494 		idev = ipv6_add_dev(dev);
495 		if (IS_ERR(idev))
496 			return idev;
497 	}
498 
499 	if (dev->flags&IFF_UP)
500 		ipv6_mc_up(idev);
501 	return idev;
502 }
503 
504 static int inet6_netconf_msgsize_devconf(int type)
505 {
506 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
507 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
508 	bool all = false;
509 
510 	if (type == NETCONFA_ALL)
511 		all = true;
512 
513 	if (all || type == NETCONFA_FORWARDING)
514 		size += nla_total_size(4);
515 #ifdef CONFIG_IPV6_MROUTE
516 	if (all || type == NETCONFA_MC_FORWARDING)
517 		size += nla_total_size(4);
518 #endif
519 	if (all || type == NETCONFA_PROXY_NEIGH)
520 		size += nla_total_size(4);
521 
522 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
523 		size += nla_total_size(4);
524 
525 	return size;
526 }
527 
528 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
529 				      struct ipv6_devconf *devconf, u32 portid,
530 				      u32 seq, int event, unsigned int flags,
531 				      int type)
532 {
533 	struct nlmsghdr  *nlh;
534 	struct netconfmsg *ncm;
535 	bool all = false;
536 
537 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
538 			flags);
539 	if (!nlh)
540 		return -EMSGSIZE;
541 
542 	if (type == NETCONFA_ALL)
543 		all = true;
544 
545 	ncm = nlmsg_data(nlh);
546 	ncm->ncm_family = AF_INET6;
547 
548 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
549 		goto nla_put_failure;
550 
551 	if (!devconf)
552 		goto out;
553 
554 	if ((all || type == NETCONFA_FORWARDING) &&
555 	    nla_put_s32(skb, NETCONFA_FORWARDING,
556 			READ_ONCE(devconf->forwarding)) < 0)
557 		goto nla_put_failure;
558 #ifdef CONFIG_IPV6_MROUTE
559 	if ((all || type == NETCONFA_MC_FORWARDING) &&
560 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
561 			atomic_read(&devconf->mc_forwarding)) < 0)
562 		goto nla_put_failure;
563 #endif
564 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
565 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
566 			READ_ONCE(devconf->proxy_ndp)) < 0)
567 		goto nla_put_failure;
568 
569 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
570 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
571 			READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
572 		goto nla_put_failure;
573 
574 out:
575 	nlmsg_end(skb, nlh);
576 	return 0;
577 
578 nla_put_failure:
579 	nlmsg_cancel(skb, nlh);
580 	return -EMSGSIZE;
581 }
582 
583 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
584 				  int ifindex, struct ipv6_devconf *devconf)
585 {
586 	struct sk_buff *skb;
587 	int err = -ENOBUFS;
588 
589 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
590 	if (!skb)
591 		goto errout;
592 
593 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
594 					 event, 0, type);
595 	if (err < 0) {
596 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
597 		WARN_ON(err == -EMSGSIZE);
598 		kfree_skb(skb);
599 		goto errout;
600 	}
601 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
602 	return;
603 errout:
604 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
605 }
606 
607 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
608 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
609 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
610 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
611 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
612 };
613 
614 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
615 				       const struct nlmsghdr *nlh,
616 				       struct nlattr **tb,
617 				       struct netlink_ext_ack *extack)
618 {
619 	int i, err;
620 
621 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
622 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
623 		return -EINVAL;
624 	}
625 
626 	if (!netlink_strict_get_check(skb))
627 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
628 					      tb, NETCONFA_MAX,
629 					      devconf_ipv6_policy, extack);
630 
631 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
632 					    tb, NETCONFA_MAX,
633 					    devconf_ipv6_policy, extack);
634 	if (err)
635 		return err;
636 
637 	for (i = 0; i <= NETCONFA_MAX; i++) {
638 		if (!tb[i])
639 			continue;
640 
641 		switch (i) {
642 		case NETCONFA_IFINDEX:
643 			break;
644 		default:
645 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
646 			return -EINVAL;
647 		}
648 	}
649 
650 	return 0;
651 }
652 
653 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
654 				     struct nlmsghdr *nlh,
655 				     struct netlink_ext_ack *extack)
656 {
657 	struct net *net = sock_net(in_skb->sk);
658 	struct nlattr *tb[NETCONFA_MAX+1];
659 	struct inet6_dev *in6_dev = NULL;
660 	struct net_device *dev = NULL;
661 	struct sk_buff *skb;
662 	struct ipv6_devconf *devconf;
663 	int ifindex;
664 	int err;
665 
666 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
667 	if (err < 0)
668 		return err;
669 
670 	if (!tb[NETCONFA_IFINDEX])
671 		return -EINVAL;
672 
673 	err = -EINVAL;
674 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
675 	switch (ifindex) {
676 	case NETCONFA_IFINDEX_ALL:
677 		devconf = net->ipv6.devconf_all;
678 		break;
679 	case NETCONFA_IFINDEX_DEFAULT:
680 		devconf = net->ipv6.devconf_dflt;
681 		break;
682 	default:
683 		dev = dev_get_by_index(net, ifindex);
684 		if (!dev)
685 			return -EINVAL;
686 		in6_dev = in6_dev_get(dev);
687 		if (!in6_dev)
688 			goto errout;
689 		devconf = &in6_dev->cnf;
690 		break;
691 	}
692 
693 	err = -ENOBUFS;
694 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
695 	if (!skb)
696 		goto errout;
697 
698 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
699 					 NETLINK_CB(in_skb).portid,
700 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
701 					 NETCONFA_ALL);
702 	if (err < 0) {
703 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
704 		WARN_ON(err == -EMSGSIZE);
705 		kfree_skb(skb);
706 		goto errout;
707 	}
708 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
709 errout:
710 	if (in6_dev)
711 		in6_dev_put(in6_dev);
712 	dev_put(dev);
713 	return err;
714 }
715 
716 /* Combine dev_addr_genid and dev_base_seq to detect changes.
717  */
718 static u32 inet6_base_seq(const struct net *net)
719 {
720 	u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
721 		  READ_ONCE(net->dev_base_seq);
722 
723 	/* Must not return 0 (see nl_dump_check_consistent()).
724 	 * Chose a value far away from 0.
725 	 */
726 	if (!res)
727 		res = 0x80000000;
728 	return res;
729 }
730 
731 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
732 				      struct netlink_callback *cb)
733 {
734 	const struct nlmsghdr *nlh = cb->nlh;
735 	struct net *net = sock_net(skb->sk);
736 	struct {
737 		unsigned long ifindex;
738 		unsigned int all_default;
739 	} *ctx = (void *)cb->ctx;
740 	struct net_device *dev;
741 	struct inet6_dev *idev;
742 	int err = 0;
743 
744 	if (cb->strict_check) {
745 		struct netlink_ext_ack *extack = cb->extack;
746 		struct netconfmsg *ncm;
747 
748 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
749 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
750 			return -EINVAL;
751 		}
752 
753 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
754 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
755 			return -EINVAL;
756 		}
757 	}
758 
759 	rcu_read_lock();
760 	for_each_netdev_dump(net, dev, ctx->ifindex) {
761 		idev = __in6_dev_get(dev);
762 		if (!idev)
763 			continue;
764 		err = inet6_netconf_fill_devconf(skb, dev->ifindex,
765 					         &idev->cnf,
766 						 NETLINK_CB(cb->skb).portid,
767 						 nlh->nlmsg_seq,
768 						 RTM_NEWNETCONF,
769 						 NLM_F_MULTI,
770 						 NETCONFA_ALL);
771 		if (err < 0)
772 			goto done;
773 	}
774 	if (ctx->all_default == 0) {
775 		err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
776 						 net->ipv6.devconf_all,
777 						 NETLINK_CB(cb->skb).portid,
778 						 nlh->nlmsg_seq,
779 						 RTM_NEWNETCONF, NLM_F_MULTI,
780 						 NETCONFA_ALL);
781 		if (err < 0)
782 			goto done;
783 		ctx->all_default++;
784 	}
785 	if (ctx->all_default == 1) {
786 		err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
787 						 net->ipv6.devconf_dflt,
788 						 NETLINK_CB(cb->skb).portid,
789 						 nlh->nlmsg_seq,
790 						 RTM_NEWNETCONF, NLM_F_MULTI,
791 						 NETCONFA_ALL);
792 		if (err < 0)
793 			goto done;
794 		ctx->all_default++;
795 	}
796 done:
797 	rcu_read_unlock();
798 	return err;
799 }
800 
801 #ifdef CONFIG_SYSCTL
802 static void dev_forward_change(struct inet6_dev *idev)
803 {
804 	struct net_device *dev;
805 	struct inet6_ifaddr *ifa;
806 	LIST_HEAD(tmp_addr_list);
807 
808 	if (!idev)
809 		return;
810 	dev = idev->dev;
811 	if (idev->cnf.forwarding)
812 		dev_disable_lro(dev);
813 	if (dev->flags & IFF_MULTICAST) {
814 		if (idev->cnf.forwarding) {
815 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
816 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
817 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
818 		} else {
819 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
820 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
821 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
822 		}
823 	}
824 
825 	read_lock_bh(&idev->lock);
826 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
827 		if (ifa->flags&IFA_F_TENTATIVE)
828 			continue;
829 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
830 	}
831 	read_unlock_bh(&idev->lock);
832 
833 	while (!list_empty(&tmp_addr_list)) {
834 		ifa = list_first_entry(&tmp_addr_list,
835 				       struct inet6_ifaddr, if_list_aux);
836 		list_del(&ifa->if_list_aux);
837 		if (idev->cnf.forwarding)
838 			addrconf_join_anycast(ifa);
839 		else
840 			addrconf_leave_anycast(ifa);
841 	}
842 
843 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
844 				     NETCONFA_FORWARDING,
845 				     dev->ifindex, &idev->cnf);
846 }
847 
848 
849 static void addrconf_forward_change(struct net *net, __s32 newf)
850 {
851 	struct net_device *dev;
852 	struct inet6_dev *idev;
853 
854 	for_each_netdev(net, dev) {
855 		idev = __in6_dev_get(dev);
856 		if (idev) {
857 			int changed = (!idev->cnf.forwarding) ^ (!newf);
858 
859 			WRITE_ONCE(idev->cnf.forwarding, newf);
860 			if (changed)
861 				dev_forward_change(idev);
862 		}
863 	}
864 }
865 
866 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf)
867 {
868 	struct net *net;
869 	int old;
870 
871 	if (!rtnl_trylock())
872 		return restart_syscall();
873 
874 	net = (struct net *)table->extra2;
875 	old = *p;
876 	WRITE_ONCE(*p, newf);
877 
878 	if (p == &net->ipv6.devconf_dflt->forwarding) {
879 		if ((!newf) ^ (!old))
880 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
881 						     NETCONFA_FORWARDING,
882 						     NETCONFA_IFINDEX_DEFAULT,
883 						     net->ipv6.devconf_dflt);
884 		rtnl_unlock();
885 		return 0;
886 	}
887 
888 	if (p == &net->ipv6.devconf_all->forwarding) {
889 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
890 
891 		WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
892 		if ((!newf) ^ (!old_dflt))
893 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
894 						     NETCONFA_FORWARDING,
895 						     NETCONFA_IFINDEX_DEFAULT,
896 						     net->ipv6.devconf_dflt);
897 
898 		addrconf_forward_change(net, newf);
899 		if ((!newf) ^ (!old))
900 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
901 						     NETCONFA_FORWARDING,
902 						     NETCONFA_IFINDEX_ALL,
903 						     net->ipv6.devconf_all);
904 	} else if ((!newf) ^ (!old))
905 		dev_forward_change((struct inet6_dev *)table->extra1);
906 	rtnl_unlock();
907 
908 	if (newf)
909 		rt6_purge_dflt_routers(net);
910 	return 1;
911 }
912 
913 static void addrconf_linkdown_change(struct net *net, __s32 newf)
914 {
915 	struct net_device *dev;
916 	struct inet6_dev *idev;
917 
918 	for_each_netdev(net, dev) {
919 		idev = __in6_dev_get(dev);
920 		if (idev) {
921 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
922 
923 			WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
924 			if (changed)
925 				inet6_netconf_notify_devconf(dev_net(dev),
926 							     RTM_NEWNETCONF,
927 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
928 							     dev->ifindex,
929 							     &idev->cnf);
930 		}
931 	}
932 }
933 
934 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf)
935 {
936 	struct net *net;
937 	int old;
938 
939 	if (!rtnl_trylock())
940 		return restart_syscall();
941 
942 	net = (struct net *)table->extra2;
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_unlock();
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 	rtnl_unlock();
968 
969 	return 1;
970 }
971 
972 #endif
973 
974 /* Nobody refers to this ifaddr, destroy it */
975 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
976 {
977 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
978 
979 #ifdef NET_REFCNT_DEBUG
980 	pr_debug("%s\n", __func__);
981 #endif
982 
983 	in6_dev_put(ifp->idev);
984 
985 	if (cancel_delayed_work(&ifp->dad_work))
986 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
987 			  ifp);
988 
989 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
990 		pr_warn("Freeing alive inet6 address %p\n", ifp);
991 		return;
992 	}
993 
994 	kfree_rcu(ifp, rcu);
995 }
996 
997 static void
998 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
999 {
1000 	struct list_head *p;
1001 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1002 
1003 	/*
1004 	 * Each device address list is sorted in order of scope -
1005 	 * global before linklocal.
1006 	 */
1007 	list_for_each(p, &idev->addr_list) {
1008 		struct inet6_ifaddr *ifa
1009 			= list_entry(p, struct inet6_ifaddr, if_list);
1010 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1011 			break;
1012 	}
1013 
1014 	list_add_tail_rcu(&ifp->if_list, p);
1015 }
1016 
1017 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1018 {
1019 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1020 
1021 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1022 }
1023 
1024 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1025 			       struct net_device *dev, unsigned int hash)
1026 {
1027 	struct inet6_ifaddr *ifp;
1028 
1029 	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1030 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1031 			if (!dev || ifp->idev->dev == dev)
1032 				return true;
1033 		}
1034 	}
1035 	return false;
1036 }
1037 
1038 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1039 {
1040 	struct net *net = dev_net(dev);
1041 	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1042 	int err = 0;
1043 
1044 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1045 
1046 	/* Ignore adding duplicate addresses on an interface */
1047 	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1048 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1049 		err = -EEXIST;
1050 	} else {
1051 		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1052 	}
1053 
1054 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1055 
1056 	return err;
1057 }
1058 
1059 /* On success it returns ifp with increased reference count */
1060 
1061 static struct inet6_ifaddr *
1062 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1063 	      bool can_block, struct netlink_ext_ack *extack)
1064 {
1065 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1066 	int addr_type = ipv6_addr_type(cfg->pfx);
1067 	struct net *net = dev_net(idev->dev);
1068 	struct inet6_ifaddr *ifa = NULL;
1069 	struct fib6_info *f6i = NULL;
1070 	int err = 0;
1071 
1072 	if (addr_type == IPV6_ADDR_ANY) {
1073 		NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1074 		return ERR_PTR(-EADDRNOTAVAIL);
1075 	} else if (addr_type & IPV6_ADDR_MULTICAST &&
1076 		   !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1077 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1078 		return ERR_PTR(-EADDRNOTAVAIL);
1079 	} else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1080 		   !netif_is_l3_master(idev->dev) &&
1081 		   addr_type & IPV6_ADDR_LOOPBACK) {
1082 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1083 		return ERR_PTR(-EADDRNOTAVAIL);
1084 	}
1085 
1086 	if (idev->dead) {
1087 		NL_SET_ERR_MSG_MOD(extack, "device is going away");
1088 		err = -ENODEV;
1089 		goto out;
1090 	}
1091 
1092 	if (idev->cnf.disable_ipv6) {
1093 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1094 		err = -EACCES;
1095 		goto out;
1096 	}
1097 
1098 	/* validator notifier needs to be blocking;
1099 	 * do not call in atomic context
1100 	 */
1101 	if (can_block) {
1102 		struct in6_validator_info i6vi = {
1103 			.i6vi_addr = *cfg->pfx,
1104 			.i6vi_dev = idev,
1105 			.extack = extack,
1106 		};
1107 
1108 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1109 		err = notifier_to_errno(err);
1110 		if (err < 0)
1111 			goto out;
1112 	}
1113 
1114 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1115 	if (!ifa) {
1116 		err = -ENOBUFS;
1117 		goto out;
1118 	}
1119 
1120 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1121 	if (IS_ERR(f6i)) {
1122 		err = PTR_ERR(f6i);
1123 		f6i = NULL;
1124 		goto out;
1125 	}
1126 
1127 	neigh_parms_data_state_setall(idev->nd_parms);
1128 
1129 	ifa->addr = *cfg->pfx;
1130 	if (cfg->peer_pfx)
1131 		ifa->peer_addr = *cfg->peer_pfx;
1132 
1133 	spin_lock_init(&ifa->lock);
1134 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1135 	INIT_HLIST_NODE(&ifa->addr_lst);
1136 	ifa->scope = cfg->scope;
1137 	ifa->prefix_len = cfg->plen;
1138 	ifa->rt_priority = cfg->rt_priority;
1139 	ifa->flags = cfg->ifa_flags;
1140 	ifa->ifa_proto = cfg->ifa_proto;
1141 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1142 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1143 		ifa->flags |= IFA_F_TENTATIVE;
1144 	ifa->valid_lft = cfg->valid_lft;
1145 	ifa->prefered_lft = cfg->preferred_lft;
1146 	ifa->cstamp = ifa->tstamp = jiffies;
1147 	ifa->tokenized = false;
1148 
1149 	ifa->rt = f6i;
1150 
1151 	ifa->idev = idev;
1152 	in6_dev_hold(idev);
1153 
1154 	/* For caller */
1155 	refcount_set(&ifa->refcnt, 1);
1156 
1157 	rcu_read_lock();
1158 
1159 	err = ipv6_add_addr_hash(idev->dev, ifa);
1160 	if (err < 0) {
1161 		rcu_read_unlock();
1162 		goto out;
1163 	}
1164 
1165 	write_lock_bh(&idev->lock);
1166 
1167 	/* Add to inet6_dev unicast addr list. */
1168 	ipv6_link_dev_addr(idev, ifa);
1169 
1170 	if (ifa->flags&IFA_F_TEMPORARY) {
1171 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1172 		in6_ifa_hold(ifa);
1173 	}
1174 
1175 	in6_ifa_hold(ifa);
1176 	write_unlock_bh(&idev->lock);
1177 
1178 	rcu_read_unlock();
1179 
1180 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1181 out:
1182 	if (unlikely(err < 0)) {
1183 		fib6_info_release(f6i);
1184 
1185 		if (ifa) {
1186 			if (ifa->idev)
1187 				in6_dev_put(ifa->idev);
1188 			kfree(ifa);
1189 		}
1190 		ifa = ERR_PTR(err);
1191 	}
1192 
1193 	return ifa;
1194 }
1195 
1196 enum cleanup_prefix_rt_t {
1197 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1198 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1199 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1200 };
1201 
1202 /*
1203  * Check, whether the prefix for ifp would still need a prefix route
1204  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1205  * constants.
1206  *
1207  * 1) we don't purge prefix if address was not permanent.
1208  *    prefix is managed by its own lifetime.
1209  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1210  * 3) if there are no addresses, delete prefix.
1211  * 4) if there are still other permanent address(es),
1212  *    corresponding prefix is still permanent.
1213  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1214  *    don't purge the prefix, assume user space is managing it.
1215  * 6) otherwise, update prefix lifetime to the
1216  *    longest valid lifetime among the corresponding
1217  *    addresses on the device.
1218  *    Note: subsequent RA will update lifetime.
1219  **/
1220 static enum cleanup_prefix_rt_t
1221 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1222 {
1223 	struct inet6_ifaddr *ifa;
1224 	struct inet6_dev *idev = ifp->idev;
1225 	unsigned long lifetime;
1226 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1227 
1228 	*expires = jiffies;
1229 
1230 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1231 		if (ifa == ifp)
1232 			continue;
1233 		if (ifa->prefix_len != ifp->prefix_len ||
1234 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1235 				       ifp->prefix_len))
1236 			continue;
1237 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1238 			return CLEANUP_PREFIX_RT_NOP;
1239 
1240 		action = CLEANUP_PREFIX_RT_EXPIRE;
1241 
1242 		spin_lock(&ifa->lock);
1243 
1244 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1245 		/*
1246 		 * Note: Because this address is
1247 		 * not permanent, lifetime <
1248 		 * LONG_MAX / HZ here.
1249 		 */
1250 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1251 			*expires = ifa->tstamp + lifetime * HZ;
1252 		spin_unlock(&ifa->lock);
1253 	}
1254 
1255 	return action;
1256 }
1257 
1258 static void
1259 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1260 		     bool del_rt, bool del_peer)
1261 {
1262 	struct fib6_table *table;
1263 	struct fib6_info *f6i;
1264 
1265 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1266 					ifp->prefix_len,
1267 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1268 	if (f6i) {
1269 		if (del_rt)
1270 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1271 		else {
1272 			if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1273 				table = f6i->fib6_table;
1274 				spin_lock_bh(&table->tb6_lock);
1275 
1276 				fib6_set_expires(f6i, expires);
1277 				fib6_add_gc_list(f6i);
1278 
1279 				spin_unlock_bh(&table->tb6_lock);
1280 			}
1281 			fib6_info_release(f6i);
1282 		}
1283 	}
1284 }
1285 
1286 
1287 /* This function wants to get referenced ifp and releases it before return */
1288 
1289 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1290 {
1291 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1292 	struct net *net = dev_net(ifp->idev->dev);
1293 	unsigned long expires;
1294 	int state;
1295 
1296 	ASSERT_RTNL();
1297 
1298 	spin_lock_bh(&ifp->lock);
1299 	state = ifp->state;
1300 	ifp->state = INET6_IFADDR_STATE_DEAD;
1301 	spin_unlock_bh(&ifp->lock);
1302 
1303 	if (state == INET6_IFADDR_STATE_DEAD)
1304 		goto out;
1305 
1306 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1307 	hlist_del_init_rcu(&ifp->addr_lst);
1308 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1309 
1310 	write_lock_bh(&ifp->idev->lock);
1311 
1312 	if (ifp->flags&IFA_F_TEMPORARY) {
1313 		list_del(&ifp->tmp_list);
1314 		if (ifp->ifpub) {
1315 			in6_ifa_put(ifp->ifpub);
1316 			ifp->ifpub = NULL;
1317 		}
1318 		__in6_ifa_put(ifp);
1319 	}
1320 
1321 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1322 		action = check_cleanup_prefix_route(ifp, &expires);
1323 
1324 	list_del_rcu(&ifp->if_list);
1325 	__in6_ifa_put(ifp);
1326 
1327 	write_unlock_bh(&ifp->idev->lock);
1328 
1329 	addrconf_del_dad_work(ifp);
1330 
1331 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1332 
1333 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1334 
1335 	if (action != CLEANUP_PREFIX_RT_NOP) {
1336 		cleanup_prefix_route(ifp, expires,
1337 			action == CLEANUP_PREFIX_RT_DEL, false);
1338 	}
1339 
1340 	/* clean up prefsrc entries */
1341 	rt6_remove_prefsrc(ifp);
1342 out:
1343 	in6_ifa_put(ifp);
1344 }
1345 
1346 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1347 {
1348 	return READ_ONCE(idev->cnf.regen_min_advance) +
1349 		READ_ONCE(idev->cnf.regen_max_retry) *
1350 		READ_ONCE(idev->cnf.dad_transmits) *
1351 		max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1352 }
1353 
1354 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1355 {
1356 	struct inet6_dev *idev = ifp->idev;
1357 	unsigned long tmp_tstamp, age;
1358 	unsigned long regen_advance;
1359 	unsigned long now = jiffies;
1360 	u32 if_public_preferred_lft;
1361 	s32 cnf_temp_preferred_lft;
1362 	struct inet6_ifaddr *ift;
1363 	struct ifa6_config cfg;
1364 	long max_desync_factor;
1365 	struct in6_addr addr;
1366 	int ret = 0;
1367 
1368 	write_lock_bh(&idev->lock);
1369 
1370 retry:
1371 	in6_dev_hold(idev);
1372 	if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1373 		write_unlock_bh(&idev->lock);
1374 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1375 		in6_dev_put(idev);
1376 		ret = -1;
1377 		goto out;
1378 	}
1379 	spin_lock_bh(&ifp->lock);
1380 	if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1381 		WRITE_ONCE(idev->cnf.use_tempaddr, -1);	/*XXX*/
1382 		spin_unlock_bh(&ifp->lock);
1383 		write_unlock_bh(&idev->lock);
1384 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1385 			__func__);
1386 		in6_dev_put(idev);
1387 		ret = -1;
1388 		goto out;
1389 	}
1390 	in6_ifa_hold(ifp);
1391 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1392 	ipv6_gen_rnd_iid(&addr);
1393 
1394 	age = (now - ifp->tstamp) / HZ;
1395 
1396 	regen_advance = ipv6_get_regen_advance(idev);
1397 
1398 	/* recalculate max_desync_factor each time and update
1399 	 * idev->desync_factor if it's larger
1400 	 */
1401 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1402 	max_desync_factor = min_t(long,
1403 				  READ_ONCE(idev->cnf.max_desync_factor),
1404 				  cnf_temp_preferred_lft - regen_advance);
1405 
1406 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1407 		if (max_desync_factor > 0) {
1408 			get_random_bytes(&idev->desync_factor,
1409 					 sizeof(idev->desync_factor));
1410 			idev->desync_factor %= max_desync_factor;
1411 		} else {
1412 			idev->desync_factor = 0;
1413 		}
1414 	}
1415 
1416 	if_public_preferred_lft = ifp->prefered_lft;
1417 
1418 	memset(&cfg, 0, sizeof(cfg));
1419 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1420 			      READ_ONCE(idev->cnf.temp_valid_lft) + age);
1421 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1422 	cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1423 	cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1424 
1425 	cfg.plen = ifp->prefix_len;
1426 	tmp_tstamp = ifp->tstamp;
1427 	spin_unlock_bh(&ifp->lock);
1428 
1429 	write_unlock_bh(&idev->lock);
1430 
1431 	/* From RFC 4941:
1432 	 *
1433 	 *     A temporary address is created only if this calculated Preferred
1434 	 *     Lifetime is greater than REGEN_ADVANCE time units.  In
1435 	 *     particular, an implementation must not create a temporary address
1436 	 *     with a zero Preferred Lifetime.
1437 	 *
1438 	 *     ...
1439 	 *
1440 	 *     When creating a temporary address, the lifetime values MUST be
1441 	 *     derived from the corresponding prefix as follows:
1442 	 *
1443 	 *     ...
1444 	 *
1445 	 *     *  Its Preferred Lifetime is the lower of the Preferred Lifetime
1446 	 *        of the public address or TEMP_PREFERRED_LIFETIME -
1447 	 *        DESYNC_FACTOR.
1448 	 *
1449 	 * To comply with the RFC's requirements, clamp the preferred lifetime
1450 	 * to a minimum of regen_advance, unless that would exceed valid_lft or
1451 	 * ifp->prefered_lft.
1452 	 *
1453 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1454 	 * temporary addresses being generated.
1455 	 */
1456 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1457 	if (cfg.preferred_lft <= regen_advance + age) {
1458 		cfg.preferred_lft = regen_advance + age + 1;
1459 		if (cfg.preferred_lft > cfg.valid_lft ||
1460 		    cfg.preferred_lft > if_public_preferred_lft) {
1461 			in6_ifa_put(ifp);
1462 			in6_dev_put(idev);
1463 			ret = -1;
1464 			goto out;
1465 		}
1466 	}
1467 
1468 	cfg.ifa_flags = IFA_F_TEMPORARY;
1469 	/* set in addrconf_prefix_rcv() */
1470 	if (ifp->flags & IFA_F_OPTIMISTIC)
1471 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1472 
1473 	cfg.pfx = &addr;
1474 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1475 
1476 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1477 	if (IS_ERR(ift)) {
1478 		in6_ifa_put(ifp);
1479 		in6_dev_put(idev);
1480 		pr_info("%s: retry temporary address regeneration\n", __func__);
1481 		write_lock_bh(&idev->lock);
1482 		goto retry;
1483 	}
1484 
1485 	spin_lock_bh(&ift->lock);
1486 	ift->ifpub = ifp;
1487 	ift->cstamp = now;
1488 	ift->tstamp = tmp_tstamp;
1489 	spin_unlock_bh(&ift->lock);
1490 
1491 	addrconf_dad_start(ift);
1492 	in6_ifa_put(ift);
1493 	in6_dev_put(idev);
1494 out:
1495 	return ret;
1496 }
1497 
1498 /*
1499  *	Choose an appropriate source address (RFC3484)
1500  */
1501 enum {
1502 	IPV6_SADDR_RULE_INIT = 0,
1503 	IPV6_SADDR_RULE_LOCAL,
1504 	IPV6_SADDR_RULE_SCOPE,
1505 	IPV6_SADDR_RULE_PREFERRED,
1506 #ifdef CONFIG_IPV6_MIP6
1507 	IPV6_SADDR_RULE_HOA,
1508 #endif
1509 	IPV6_SADDR_RULE_OIF,
1510 	IPV6_SADDR_RULE_LABEL,
1511 	IPV6_SADDR_RULE_PRIVACY,
1512 	IPV6_SADDR_RULE_ORCHID,
1513 	IPV6_SADDR_RULE_PREFIX,
1514 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1515 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1516 #endif
1517 	IPV6_SADDR_RULE_MAX
1518 };
1519 
1520 struct ipv6_saddr_score {
1521 	int			rule;
1522 	int			addr_type;
1523 	struct inet6_ifaddr	*ifa;
1524 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1525 	int			scopedist;
1526 	int			matchlen;
1527 };
1528 
1529 struct ipv6_saddr_dst {
1530 	const struct in6_addr *addr;
1531 	int ifindex;
1532 	int scope;
1533 	int label;
1534 	unsigned int prefs;
1535 };
1536 
1537 static inline int ipv6_saddr_preferred(int type)
1538 {
1539 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1540 		return 1;
1541 	return 0;
1542 }
1543 
1544 static bool ipv6_use_optimistic_addr(const struct net *net,
1545 				     const struct inet6_dev *idev)
1546 {
1547 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1548 	if (!idev)
1549 		return false;
1550 	if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1551 	    !READ_ONCE(idev->cnf.optimistic_dad))
1552 		return false;
1553 	if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1554 	    !READ_ONCE(idev->cnf.use_optimistic))
1555 		return false;
1556 
1557 	return true;
1558 #else
1559 	return false;
1560 #endif
1561 }
1562 
1563 static bool ipv6_allow_optimistic_dad(const struct net *net,
1564 				      const struct inet6_dev *idev)
1565 {
1566 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1567 	if (!idev)
1568 		return false;
1569 	if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1570 	    !READ_ONCE(idev->cnf.optimistic_dad))
1571 		return false;
1572 
1573 	return true;
1574 #else
1575 	return false;
1576 #endif
1577 }
1578 
1579 static int ipv6_get_saddr_eval(struct net *net,
1580 			       struct ipv6_saddr_score *score,
1581 			       struct ipv6_saddr_dst *dst,
1582 			       int i)
1583 {
1584 	int ret;
1585 
1586 	if (i <= score->rule) {
1587 		switch (i) {
1588 		case IPV6_SADDR_RULE_SCOPE:
1589 			ret = score->scopedist;
1590 			break;
1591 		case IPV6_SADDR_RULE_PREFIX:
1592 			ret = score->matchlen;
1593 			break;
1594 		default:
1595 			ret = !!test_bit(i, score->scorebits);
1596 		}
1597 		goto out;
1598 	}
1599 
1600 	switch (i) {
1601 	case IPV6_SADDR_RULE_INIT:
1602 		/* Rule 0: remember if hiscore is not ready yet */
1603 		ret = !!score->ifa;
1604 		break;
1605 	case IPV6_SADDR_RULE_LOCAL:
1606 		/* Rule 1: Prefer same address */
1607 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1608 		break;
1609 	case IPV6_SADDR_RULE_SCOPE:
1610 		/* Rule 2: Prefer appropriate scope
1611 		 *
1612 		 *      ret
1613 		 *       ^
1614 		 *    -1 |  d 15
1615 		 *    ---+--+-+---> scope
1616 		 *       |
1617 		 *       |             d is scope of the destination.
1618 		 *  B-d  |  \
1619 		 *       |   \      <- smaller scope is better if
1620 		 *  B-15 |    \        if scope is enough for destination.
1621 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1622 		 * d-C-1 | /
1623 		 *       |/         <- greater is better
1624 		 *   -C  /             if scope is not enough for destination.
1625 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1626 		 *
1627 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1628 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1629 		 * Assume B = 0 and we get C > 29.
1630 		 */
1631 		ret = __ipv6_addr_src_scope(score->addr_type);
1632 		if (ret >= dst->scope)
1633 			ret = -ret;
1634 		else
1635 			ret -= 128;	/* 30 is enough */
1636 		score->scopedist = ret;
1637 		break;
1638 	case IPV6_SADDR_RULE_PREFERRED:
1639 	    {
1640 		/* Rule 3: Avoid deprecated and optimistic addresses */
1641 		u8 avoid = IFA_F_DEPRECATED;
1642 
1643 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1644 			avoid |= IFA_F_OPTIMISTIC;
1645 		ret = ipv6_saddr_preferred(score->addr_type) ||
1646 		      !(score->ifa->flags & avoid);
1647 		break;
1648 	    }
1649 #ifdef CONFIG_IPV6_MIP6
1650 	case IPV6_SADDR_RULE_HOA:
1651 	    {
1652 		/* Rule 4: Prefer home address */
1653 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1654 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1655 		break;
1656 	    }
1657 #endif
1658 	case IPV6_SADDR_RULE_OIF:
1659 		/* Rule 5: Prefer outgoing interface */
1660 		ret = (!dst->ifindex ||
1661 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1662 		break;
1663 	case IPV6_SADDR_RULE_LABEL:
1664 		/* Rule 6: Prefer matching label */
1665 		ret = ipv6_addr_label(net,
1666 				      &score->ifa->addr, score->addr_type,
1667 				      score->ifa->idev->dev->ifindex) == dst->label;
1668 		break;
1669 	case IPV6_SADDR_RULE_PRIVACY:
1670 	    {
1671 		/* Rule 7: Prefer public address
1672 		 * Note: prefer temporary address if use_tempaddr >= 2
1673 		 */
1674 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1675 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1676 				READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1677 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1678 		break;
1679 	    }
1680 	case IPV6_SADDR_RULE_ORCHID:
1681 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1682 		 *	    non-ORCHID vs non-ORCHID
1683 		 */
1684 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1685 			ipv6_addr_orchid(dst->addr));
1686 		break;
1687 	case IPV6_SADDR_RULE_PREFIX:
1688 		/* Rule 8: Use longest matching prefix */
1689 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1690 		if (ret > score->ifa->prefix_len)
1691 			ret = score->ifa->prefix_len;
1692 		score->matchlen = ret;
1693 		break;
1694 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1695 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1696 		/* Optimistic addresses still have lower precedence than other
1697 		 * preferred addresses.
1698 		 */
1699 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1700 		break;
1701 #endif
1702 	default:
1703 		ret = 0;
1704 	}
1705 
1706 	if (ret)
1707 		__set_bit(i, score->scorebits);
1708 	score->rule = i;
1709 out:
1710 	return ret;
1711 }
1712 
1713 static int __ipv6_dev_get_saddr(struct net *net,
1714 				struct ipv6_saddr_dst *dst,
1715 				struct inet6_dev *idev,
1716 				struct ipv6_saddr_score *scores,
1717 				int hiscore_idx)
1718 {
1719 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1720 
1721 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1722 		int i;
1723 
1724 		/*
1725 		 * - Tentative Address (RFC2462 section 5.4)
1726 		 *  - A tentative address is not considered
1727 		 *    "assigned to an interface" in the traditional
1728 		 *    sense, unless it is also flagged as optimistic.
1729 		 * - Candidate Source Address (section 4)
1730 		 *  - In any case, anycast addresses, multicast
1731 		 *    addresses, and the unspecified address MUST
1732 		 *    NOT be included in a candidate set.
1733 		 */
1734 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1735 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1736 			continue;
1737 
1738 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1739 
1740 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1741 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1742 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1743 					    idev->dev->name);
1744 			continue;
1745 		}
1746 
1747 		score->rule = -1;
1748 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1749 
1750 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1751 			int minihiscore, miniscore;
1752 
1753 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1754 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1755 
1756 			if (minihiscore > miniscore) {
1757 				if (i == IPV6_SADDR_RULE_SCOPE &&
1758 				    score->scopedist > 0) {
1759 					/*
1760 					 * special case:
1761 					 * each remaining entry
1762 					 * has too small (not enough)
1763 					 * scope, because ifa entries
1764 					 * are sorted by their scope
1765 					 * values.
1766 					 */
1767 					goto out;
1768 				}
1769 				break;
1770 			} else if (minihiscore < miniscore) {
1771 				swap(hiscore, score);
1772 				hiscore_idx = 1 - hiscore_idx;
1773 
1774 				/* restore our iterator */
1775 				score->ifa = hiscore->ifa;
1776 
1777 				break;
1778 			}
1779 		}
1780 	}
1781 out:
1782 	return hiscore_idx;
1783 }
1784 
1785 static int ipv6_get_saddr_master(struct net *net,
1786 				 const struct net_device *dst_dev,
1787 				 const struct net_device *master,
1788 				 struct ipv6_saddr_dst *dst,
1789 				 struct ipv6_saddr_score *scores,
1790 				 int hiscore_idx)
1791 {
1792 	struct inet6_dev *idev;
1793 
1794 	idev = __in6_dev_get(dst_dev);
1795 	if (idev)
1796 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1797 						   scores, hiscore_idx);
1798 
1799 	idev = __in6_dev_get(master);
1800 	if (idev)
1801 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1802 						   scores, hiscore_idx);
1803 
1804 	return hiscore_idx;
1805 }
1806 
1807 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1808 		       const struct in6_addr *daddr, unsigned int prefs,
1809 		       struct in6_addr *saddr)
1810 {
1811 	struct ipv6_saddr_score scores[2], *hiscore;
1812 	struct ipv6_saddr_dst dst;
1813 	struct inet6_dev *idev;
1814 	struct net_device *dev;
1815 	int dst_type;
1816 	bool use_oif_addr = false;
1817 	int hiscore_idx = 0;
1818 	int ret = 0;
1819 
1820 	dst_type = __ipv6_addr_type(daddr);
1821 	dst.addr = daddr;
1822 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1823 	dst.scope = __ipv6_addr_src_scope(dst_type);
1824 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1825 	dst.prefs = prefs;
1826 
1827 	scores[hiscore_idx].rule = -1;
1828 	scores[hiscore_idx].ifa = NULL;
1829 
1830 	rcu_read_lock();
1831 
1832 	/* Candidate Source Address (section 4)
1833 	 *  - multicast and link-local destination address,
1834 	 *    the set of candidate source address MUST only
1835 	 *    include addresses assigned to interfaces
1836 	 *    belonging to the same link as the outgoing
1837 	 *    interface.
1838 	 * (- For site-local destination addresses, the
1839 	 *    set of candidate source addresses MUST only
1840 	 *    include addresses assigned to interfaces
1841 	 *    belonging to the same site as the outgoing
1842 	 *    interface.)
1843 	 *  - "It is RECOMMENDED that the candidate source addresses
1844 	 *    be the set of unicast addresses assigned to the
1845 	 *    interface that will be used to send to the destination
1846 	 *    (the 'outgoing' interface)." (RFC 6724)
1847 	 */
1848 	if (dst_dev) {
1849 		idev = __in6_dev_get(dst_dev);
1850 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1851 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1852 		    (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1853 			use_oif_addr = true;
1854 		}
1855 	}
1856 
1857 	if (use_oif_addr) {
1858 		if (idev)
1859 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1860 	} else {
1861 		const struct net_device *master;
1862 		int master_idx = 0;
1863 
1864 		/* if dst_dev exists and is enslaved to an L3 device, then
1865 		 * prefer addresses from dst_dev and then the master over
1866 		 * any other enslaved devices in the L3 domain.
1867 		 */
1868 		master = l3mdev_master_dev_rcu(dst_dev);
1869 		if (master) {
1870 			master_idx = master->ifindex;
1871 
1872 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1873 							    master, &dst,
1874 							    scores, hiscore_idx);
1875 
1876 			if (scores[hiscore_idx].ifa &&
1877 			    scores[hiscore_idx].scopedist >= 0)
1878 				goto out;
1879 		}
1880 
1881 		for_each_netdev_rcu(net, dev) {
1882 			/* only consider addresses on devices in the
1883 			 * same L3 domain
1884 			 */
1885 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1886 				continue;
1887 			idev = __in6_dev_get(dev);
1888 			if (!idev)
1889 				continue;
1890 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1891 		}
1892 	}
1893 
1894 out:
1895 	hiscore = &scores[hiscore_idx];
1896 	if (!hiscore->ifa)
1897 		ret = -EADDRNOTAVAIL;
1898 	else
1899 		*saddr = hiscore->ifa->addr;
1900 
1901 	rcu_read_unlock();
1902 	return ret;
1903 }
1904 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1905 
1906 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1907 			      u32 banned_flags)
1908 {
1909 	struct inet6_ifaddr *ifp;
1910 	int err = -EADDRNOTAVAIL;
1911 
1912 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1913 		if (ifp->scope > IFA_LINK)
1914 			break;
1915 		if (ifp->scope == IFA_LINK &&
1916 		    !(ifp->flags & banned_flags)) {
1917 			*addr = ifp->addr;
1918 			err = 0;
1919 			break;
1920 		}
1921 	}
1922 	return err;
1923 }
1924 
1925 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1926 		    u32 banned_flags)
1927 {
1928 	struct inet6_dev *idev;
1929 	int err = -EADDRNOTAVAIL;
1930 
1931 	rcu_read_lock();
1932 	idev = __in6_dev_get(dev);
1933 	if (idev) {
1934 		read_lock_bh(&idev->lock);
1935 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1936 		read_unlock_bh(&idev->lock);
1937 	}
1938 	rcu_read_unlock();
1939 	return err;
1940 }
1941 
1942 static int ipv6_count_addresses(const struct inet6_dev *idev)
1943 {
1944 	const struct inet6_ifaddr *ifp;
1945 	int cnt = 0;
1946 
1947 	rcu_read_lock();
1948 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1949 		cnt++;
1950 	rcu_read_unlock();
1951 	return cnt;
1952 }
1953 
1954 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1955 		  const struct net_device *dev, int strict)
1956 {
1957 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1958 				       strict, IFA_F_TENTATIVE);
1959 }
1960 EXPORT_SYMBOL(ipv6_chk_addr);
1961 
1962 /* device argument is used to find the L3 domain of interest. If
1963  * skip_dev_check is set, then the ifp device is not checked against
1964  * the passed in dev argument. So the 2 cases for addresses checks are:
1965  *   1. does the address exist in the L3 domain that dev is part of
1966  *      (skip_dev_check = true), or
1967  *
1968  *   2. does the address exist on the specific device
1969  *      (skip_dev_check = false)
1970  */
1971 static struct net_device *
1972 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1973 			  const struct net_device *dev, bool skip_dev_check,
1974 			  int strict, u32 banned_flags)
1975 {
1976 	unsigned int hash = inet6_addr_hash(net, addr);
1977 	struct net_device *l3mdev, *ndev;
1978 	struct inet6_ifaddr *ifp;
1979 	u32 ifp_flags;
1980 
1981 	rcu_read_lock();
1982 
1983 	l3mdev = l3mdev_master_dev_rcu(dev);
1984 	if (skip_dev_check)
1985 		dev = NULL;
1986 
1987 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1988 		ndev = ifp->idev->dev;
1989 
1990 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1991 			continue;
1992 
1993 		/* Decouple optimistic from tentative for evaluation here.
1994 		 * Ban optimistic addresses explicitly, when required.
1995 		 */
1996 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1997 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1998 			    : ifp->flags;
1999 		if (ipv6_addr_equal(&ifp->addr, addr) &&
2000 		    !(ifp_flags&banned_flags) &&
2001 		    (!dev || ndev == dev ||
2002 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2003 			rcu_read_unlock();
2004 			return ndev;
2005 		}
2006 	}
2007 
2008 	rcu_read_unlock();
2009 	return NULL;
2010 }
2011 
2012 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2013 			    const struct net_device *dev, bool skip_dev_check,
2014 			    int strict, u32 banned_flags)
2015 {
2016 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2017 					 strict, banned_flags) ? 1 : 0;
2018 }
2019 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2020 
2021 
2022 /* Compares an address/prefix_len with addresses on device @dev.
2023  * If one is found it returns true.
2024  */
2025 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2026 	const unsigned int prefix_len, struct net_device *dev)
2027 {
2028 	const struct inet6_ifaddr *ifa;
2029 	const struct inet6_dev *idev;
2030 	bool ret = false;
2031 
2032 	rcu_read_lock();
2033 	idev = __in6_dev_get(dev);
2034 	if (idev) {
2035 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2036 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2037 			if (ret)
2038 				break;
2039 		}
2040 	}
2041 	rcu_read_unlock();
2042 
2043 	return ret;
2044 }
2045 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2046 
2047 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2048 {
2049 	const struct inet6_ifaddr *ifa;
2050 	const struct inet6_dev *idev;
2051 	int	onlink;
2052 
2053 	onlink = 0;
2054 	rcu_read_lock();
2055 	idev = __in6_dev_get(dev);
2056 	if (idev) {
2057 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2058 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
2059 						   ifa->prefix_len);
2060 			if (onlink)
2061 				break;
2062 		}
2063 	}
2064 	rcu_read_unlock();
2065 	return onlink;
2066 }
2067 EXPORT_SYMBOL(ipv6_chk_prefix);
2068 
2069 /**
2070  * ipv6_dev_find - find the first device with a given source address.
2071  * @net: the net namespace
2072  * @addr: the source address
2073  * @dev: used to find the L3 domain of interest
2074  *
2075  * The caller should be protected by RCU, or RTNL.
2076  */
2077 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2078 				 struct net_device *dev)
2079 {
2080 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2081 					 IFA_F_TENTATIVE);
2082 }
2083 EXPORT_SYMBOL(ipv6_dev_find);
2084 
2085 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2086 				     struct net_device *dev, int strict)
2087 {
2088 	unsigned int hash = inet6_addr_hash(net, addr);
2089 	struct inet6_ifaddr *ifp, *result = NULL;
2090 
2091 	rcu_read_lock();
2092 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2093 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2094 			if (!dev || ifp->idev->dev == dev ||
2095 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2096 				if (in6_ifa_hold_safe(ifp)) {
2097 					result = ifp;
2098 					break;
2099 				}
2100 			}
2101 		}
2102 	}
2103 	rcu_read_unlock();
2104 
2105 	return result;
2106 }
2107 
2108 /* Gets referenced address, destroys ifaddr */
2109 
2110 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2111 {
2112 	if (dad_failed)
2113 		ifp->flags |= IFA_F_DADFAILED;
2114 
2115 	if (ifp->flags&IFA_F_TEMPORARY) {
2116 		struct inet6_ifaddr *ifpub;
2117 		spin_lock_bh(&ifp->lock);
2118 		ifpub = ifp->ifpub;
2119 		if (ifpub) {
2120 			in6_ifa_hold(ifpub);
2121 			spin_unlock_bh(&ifp->lock);
2122 			ipv6_create_tempaddr(ifpub, true);
2123 			in6_ifa_put(ifpub);
2124 		} else {
2125 			spin_unlock_bh(&ifp->lock);
2126 		}
2127 		ipv6_del_addr(ifp);
2128 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2129 		spin_lock_bh(&ifp->lock);
2130 		addrconf_del_dad_work(ifp);
2131 		ifp->flags |= IFA_F_TENTATIVE;
2132 		if (dad_failed)
2133 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2134 		spin_unlock_bh(&ifp->lock);
2135 		if (dad_failed)
2136 			ipv6_ifa_notify(0, ifp);
2137 		in6_ifa_put(ifp);
2138 	} else {
2139 		ipv6_del_addr(ifp);
2140 	}
2141 }
2142 
2143 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2144 {
2145 	int err = -ENOENT;
2146 
2147 	spin_lock_bh(&ifp->lock);
2148 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2149 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2150 		err = 0;
2151 	}
2152 	spin_unlock_bh(&ifp->lock);
2153 
2154 	return err;
2155 }
2156 
2157 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2158 {
2159 	struct inet6_dev *idev = ifp->idev;
2160 	struct net *net = dev_net(idev->dev);
2161 	int max_addresses;
2162 
2163 	if (addrconf_dad_end(ifp)) {
2164 		in6_ifa_put(ifp);
2165 		return;
2166 	}
2167 
2168 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2169 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2170 
2171 	spin_lock_bh(&ifp->lock);
2172 
2173 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2174 		struct in6_addr new_addr;
2175 		struct inet6_ifaddr *ifp2;
2176 		int retries = ifp->stable_privacy_retry + 1;
2177 		struct ifa6_config cfg = {
2178 			.pfx = &new_addr,
2179 			.plen = ifp->prefix_len,
2180 			.ifa_flags = ifp->flags,
2181 			.valid_lft = ifp->valid_lft,
2182 			.preferred_lft = ifp->prefered_lft,
2183 			.scope = ifp->scope,
2184 		};
2185 
2186 		if (retries > net->ipv6.sysctl.idgen_retries) {
2187 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2188 					     ifp->idev->dev->name);
2189 			goto errdad;
2190 		}
2191 
2192 		new_addr = ifp->addr;
2193 		if (ipv6_generate_stable_address(&new_addr, retries,
2194 						 idev))
2195 			goto errdad;
2196 
2197 		spin_unlock_bh(&ifp->lock);
2198 
2199 		max_addresses = READ_ONCE(idev->cnf.max_addresses);
2200 		if (max_addresses &&
2201 		    ipv6_count_addresses(idev) >= max_addresses)
2202 			goto lock_errdad;
2203 
2204 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2205 				     ifp->idev->dev->name);
2206 
2207 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2208 		if (IS_ERR(ifp2))
2209 			goto lock_errdad;
2210 
2211 		spin_lock_bh(&ifp2->lock);
2212 		ifp2->stable_privacy_retry = retries;
2213 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2214 		spin_unlock_bh(&ifp2->lock);
2215 
2216 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2217 		in6_ifa_put(ifp2);
2218 lock_errdad:
2219 		spin_lock_bh(&ifp->lock);
2220 	}
2221 
2222 errdad:
2223 	/* transition from _POSTDAD to _ERRDAD */
2224 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2225 	spin_unlock_bh(&ifp->lock);
2226 
2227 	addrconf_mod_dad_work(ifp, 0);
2228 	in6_ifa_put(ifp);
2229 }
2230 
2231 /* Join to solicited addr multicast group.
2232  * caller must hold RTNL */
2233 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2234 {
2235 	struct in6_addr maddr;
2236 
2237 	if (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 /* caller must hold RTNL */
2245 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2246 {
2247 	struct in6_addr maddr;
2248 
2249 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2250 		return;
2251 
2252 	addrconf_addr_solict_mult(addr, &maddr);
2253 	__ipv6_dev_mc_dec(idev, &maddr);
2254 }
2255 
2256 /* caller must hold RTNL */
2257 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2258 {
2259 	struct in6_addr addr;
2260 
2261 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2262 		return;
2263 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2264 	if (ipv6_addr_any(&addr))
2265 		return;
2266 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2267 }
2268 
2269 /* caller must hold RTNL */
2270 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2271 {
2272 	struct in6_addr addr;
2273 
2274 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2275 		return;
2276 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2277 	if (ipv6_addr_any(&addr))
2278 		return;
2279 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2280 }
2281 
2282 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2283 {
2284 	switch (dev->addr_len) {
2285 	case ETH_ALEN:
2286 		memcpy(eui, dev->dev_addr, 3);
2287 		eui[3] = 0xFF;
2288 		eui[4] = 0xFE;
2289 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2290 		break;
2291 	case EUI64_ADDR_LEN:
2292 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2293 		eui[0] ^= 2;
2294 		break;
2295 	default:
2296 		return -1;
2297 	}
2298 
2299 	return 0;
2300 }
2301 
2302 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2303 {
2304 	const union fwnet_hwaddr *ha;
2305 
2306 	if (dev->addr_len != FWNET_ALEN)
2307 		return -1;
2308 
2309 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2310 
2311 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2312 	eui[0] ^= 2;
2313 	return 0;
2314 }
2315 
2316 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2317 {
2318 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2319 	if (dev->addr_len != ARCNET_ALEN)
2320 		return -1;
2321 	memset(eui, 0, 7);
2322 	eui[7] = *(u8 *)dev->dev_addr;
2323 	return 0;
2324 }
2325 
2326 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2327 {
2328 	if (dev->addr_len != INFINIBAND_ALEN)
2329 		return -1;
2330 	memcpy(eui, dev->dev_addr + 12, 8);
2331 	eui[0] |= 2;
2332 	return 0;
2333 }
2334 
2335 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2336 {
2337 	if (addr == 0)
2338 		return -1;
2339 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2340 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2341 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2342 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2343 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2344 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2345 	eui[1] = 0;
2346 	eui[2] = 0x5E;
2347 	eui[3] = 0xFE;
2348 	memcpy(eui + 4, &addr, 4);
2349 	return 0;
2350 }
2351 
2352 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2353 {
2354 	if (dev->priv_flags & IFF_ISATAP)
2355 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2356 	return -1;
2357 }
2358 
2359 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2360 {
2361 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2362 }
2363 
2364 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2365 {
2366 	memcpy(eui, dev->perm_addr, 3);
2367 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2368 	eui[3] = 0xFF;
2369 	eui[4] = 0xFE;
2370 	eui[0] ^= 2;
2371 	return 0;
2372 }
2373 
2374 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2375 {
2376 	switch (dev->type) {
2377 	case ARPHRD_ETHER:
2378 	case ARPHRD_FDDI:
2379 		return addrconf_ifid_eui48(eui, dev);
2380 	case ARPHRD_ARCNET:
2381 		return addrconf_ifid_arcnet(eui, dev);
2382 	case ARPHRD_INFINIBAND:
2383 		return addrconf_ifid_infiniband(eui, dev);
2384 	case ARPHRD_SIT:
2385 		return addrconf_ifid_sit(eui, dev);
2386 	case ARPHRD_IPGRE:
2387 	case ARPHRD_TUNNEL:
2388 		return addrconf_ifid_gre(eui, dev);
2389 	case ARPHRD_6LOWPAN:
2390 		return addrconf_ifid_6lowpan(eui, dev);
2391 	case ARPHRD_IEEE1394:
2392 		return addrconf_ifid_ieee1394(eui, dev);
2393 	case ARPHRD_TUNNEL6:
2394 	case ARPHRD_IP6GRE:
2395 	case ARPHRD_RAWIP:
2396 		return addrconf_ifid_ip6tnl(eui, dev);
2397 	}
2398 	return -1;
2399 }
2400 
2401 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2402 {
2403 	int err = -1;
2404 	struct inet6_ifaddr *ifp;
2405 
2406 	read_lock_bh(&idev->lock);
2407 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2408 		if (ifp->scope > IFA_LINK)
2409 			break;
2410 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2411 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2412 			err = 0;
2413 			break;
2414 		}
2415 	}
2416 	read_unlock_bh(&idev->lock);
2417 	return err;
2418 }
2419 
2420 /* Generation of a randomized Interface Identifier
2421  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2422  */
2423 
2424 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2425 {
2426 regen:
2427 	get_random_bytes(&addr->s6_addr[8], 8);
2428 
2429 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2430 	 * check if generated address is not inappropriate:
2431 	 *
2432 	 * - Reserved IPv6 Interface Identifiers
2433 	 * - XXX: already assigned to an address on the device
2434 	 */
2435 
2436 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2437 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2438 		goto regen;
2439 
2440 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2441 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2442 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2443 	 */
2444 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2445 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2446 		goto regen;
2447 
2448 	/* Reserved subnet anycast addresses */
2449 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2450 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2451 		goto regen;
2452 }
2453 
2454 /*
2455  *	Add prefix route.
2456  */
2457 
2458 static void
2459 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2460 		      struct net_device *dev, unsigned long expires,
2461 		      u32 flags, gfp_t gfp_flags)
2462 {
2463 	struct fib6_config cfg = {
2464 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2465 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2466 		.fc_ifindex = dev->ifindex,
2467 		.fc_expires = expires,
2468 		.fc_dst_len = plen,
2469 		.fc_flags = RTF_UP | flags,
2470 		.fc_nlinfo.nl_net = dev_net(dev),
2471 		.fc_protocol = RTPROT_KERNEL,
2472 		.fc_type = RTN_UNICAST,
2473 	};
2474 
2475 	cfg.fc_dst = *pfx;
2476 
2477 	/* Prevent useless cloning on PtP SIT.
2478 	   This thing is done here expecting that the whole
2479 	   class of non-broadcast devices need not cloning.
2480 	 */
2481 #if IS_ENABLED(CONFIG_IPV6_SIT)
2482 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2483 		cfg.fc_flags |= RTF_NONEXTHOP;
2484 #endif
2485 
2486 	ip6_route_add(&cfg, gfp_flags, NULL);
2487 }
2488 
2489 
2490 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2491 						  int plen,
2492 						  const struct net_device *dev,
2493 						  u32 flags, u32 noflags,
2494 						  bool no_gw)
2495 {
2496 	struct fib6_node *fn;
2497 	struct fib6_info *rt = NULL;
2498 	struct fib6_table *table;
2499 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2500 
2501 	table = fib6_get_table(dev_net(dev), tb_id);
2502 	if (!table)
2503 		return NULL;
2504 
2505 	rcu_read_lock();
2506 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2507 	if (!fn)
2508 		goto out;
2509 
2510 	for_each_fib6_node_rt_rcu(fn) {
2511 		/* prefix routes only use builtin fib6_nh */
2512 		if (rt->nh)
2513 			continue;
2514 
2515 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2516 			continue;
2517 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2518 			continue;
2519 		if ((rt->fib6_flags & flags) != flags)
2520 			continue;
2521 		if ((rt->fib6_flags & noflags) != 0)
2522 			continue;
2523 		if (!fib6_info_hold_safe(rt))
2524 			continue;
2525 		break;
2526 	}
2527 out:
2528 	rcu_read_unlock();
2529 	return rt;
2530 }
2531 
2532 
2533 /* Create "default" multicast route to the interface */
2534 
2535 static void addrconf_add_mroute(struct net_device *dev)
2536 {
2537 	struct fib6_config cfg = {
2538 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2539 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2540 		.fc_ifindex = dev->ifindex,
2541 		.fc_dst_len = 8,
2542 		.fc_flags = RTF_UP,
2543 		.fc_type = RTN_MULTICAST,
2544 		.fc_nlinfo.nl_net = dev_net(dev),
2545 		.fc_protocol = RTPROT_KERNEL,
2546 	};
2547 
2548 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2549 
2550 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2551 }
2552 
2553 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2554 {
2555 	struct inet6_dev *idev;
2556 
2557 	ASSERT_RTNL();
2558 
2559 	idev = ipv6_find_idev(dev);
2560 	if (IS_ERR(idev))
2561 		return idev;
2562 
2563 	if (idev->cnf.disable_ipv6)
2564 		return ERR_PTR(-EACCES);
2565 
2566 	/* Add default multicast route */
2567 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2568 		addrconf_add_mroute(dev);
2569 
2570 	return idev;
2571 }
2572 
2573 static void manage_tempaddrs(struct inet6_dev *idev,
2574 			     struct inet6_ifaddr *ifp,
2575 			     __u32 valid_lft, __u32 prefered_lft,
2576 			     bool create, unsigned long now)
2577 {
2578 	u32 flags;
2579 	struct inet6_ifaddr *ift;
2580 
2581 	read_lock_bh(&idev->lock);
2582 	/* update all temporary addresses in the list */
2583 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2584 		int age, max_valid, max_prefered;
2585 
2586 		if (ifp != ift->ifpub)
2587 			continue;
2588 
2589 		/* RFC 4941 section 3.3:
2590 		 * If a received option will extend the lifetime of a public
2591 		 * address, the lifetimes of temporary addresses should
2592 		 * be extended, subject to the overall constraint that no
2593 		 * temporary addresses should ever remain "valid" or "preferred"
2594 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2595 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2596 		 */
2597 		age = (now - ift->cstamp) / HZ;
2598 		max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2599 		if (max_valid < 0)
2600 			max_valid = 0;
2601 
2602 		max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2603 			       idev->desync_factor - age;
2604 		if (max_prefered < 0)
2605 			max_prefered = 0;
2606 
2607 		if (valid_lft > max_valid)
2608 			valid_lft = max_valid;
2609 
2610 		if (prefered_lft > max_prefered)
2611 			prefered_lft = max_prefered;
2612 
2613 		spin_lock(&ift->lock);
2614 		flags = ift->flags;
2615 		ift->valid_lft = valid_lft;
2616 		ift->prefered_lft = prefered_lft;
2617 		ift->tstamp = now;
2618 		if (prefered_lft > 0)
2619 			ift->flags &= ~IFA_F_DEPRECATED;
2620 
2621 		spin_unlock(&ift->lock);
2622 		if (!(flags&IFA_F_TENTATIVE))
2623 			ipv6_ifa_notify(0, ift);
2624 	}
2625 
2626 	/* Also create a temporary address if it's enabled but no temporary
2627 	 * address currently exists.
2628 	 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2629 	 * as part of cleanup (ie. deleting the mngtmpaddr).
2630 	 * We don't want that to result in creating a new temporary ip address.
2631 	 */
2632 	if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2633 		create = true;
2634 
2635 	if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2636 		/* When a new public address is created as described
2637 		 * in [ADDRCONF], also create a new temporary address.
2638 		 */
2639 		read_unlock_bh(&idev->lock);
2640 		ipv6_create_tempaddr(ifp, false);
2641 	} else {
2642 		read_unlock_bh(&idev->lock);
2643 	}
2644 }
2645 
2646 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2647 {
2648 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2649 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2650 }
2651 
2652 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2653 				 const struct prefix_info *pinfo,
2654 				 struct inet6_dev *in6_dev,
2655 				 const struct in6_addr *addr, int addr_type,
2656 				 u32 addr_flags, bool sllao, bool tokenized,
2657 				 __u32 valid_lft, u32 prefered_lft)
2658 {
2659 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2660 	int create = 0, update_lft = 0;
2661 
2662 	if (!ifp && valid_lft) {
2663 		int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2664 		struct ifa6_config cfg = {
2665 			.pfx = addr,
2666 			.plen = pinfo->prefix_len,
2667 			.ifa_flags = addr_flags,
2668 			.valid_lft = valid_lft,
2669 			.preferred_lft = prefered_lft,
2670 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2671 			.ifa_proto = IFAPROT_KERNEL_RA
2672 		};
2673 
2674 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2675 		if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2676 		     READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2677 		    !net->ipv6.devconf_all->forwarding && sllao)
2678 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2679 #endif
2680 
2681 		/* Do not allow to create too much of autoconfigured
2682 		 * addresses; this would be too easy way to crash kernel.
2683 		 */
2684 		if (!max_addresses ||
2685 		    ipv6_count_addresses(in6_dev) < max_addresses)
2686 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2687 
2688 		if (IS_ERR_OR_NULL(ifp))
2689 			return -1;
2690 
2691 		create = 1;
2692 		spin_lock_bh(&ifp->lock);
2693 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2694 		ifp->cstamp = jiffies;
2695 		ifp->tokenized = tokenized;
2696 		spin_unlock_bh(&ifp->lock);
2697 		addrconf_dad_start(ifp);
2698 	}
2699 
2700 	if (ifp) {
2701 		u32 flags;
2702 		unsigned long now;
2703 		u32 stored_lft;
2704 
2705 		/* update lifetime (RFC2462 5.5.3 e) */
2706 		spin_lock_bh(&ifp->lock);
2707 		now = jiffies;
2708 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2709 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2710 		else
2711 			stored_lft = 0;
2712 
2713 		/* RFC4862 Section 5.5.3e:
2714 		 * "Note that the preferred lifetime of the
2715 		 *  corresponding address is always reset to
2716 		 *  the Preferred Lifetime in the received
2717 		 *  Prefix Information option, regardless of
2718 		 *  whether the valid lifetime is also reset or
2719 		 *  ignored."
2720 		 *
2721 		 * So we should always update prefered_lft here.
2722 		 */
2723 		update_lft = !create && stored_lft;
2724 
2725 		if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2726 			const u32 minimum_lft = min_t(u32,
2727 				stored_lft, MIN_VALID_LIFETIME);
2728 			valid_lft = max(valid_lft, minimum_lft);
2729 		}
2730 
2731 		if (update_lft) {
2732 			ifp->valid_lft = valid_lft;
2733 			ifp->prefered_lft = prefered_lft;
2734 			WRITE_ONCE(ifp->tstamp, now);
2735 			flags = ifp->flags;
2736 			ifp->flags &= ~IFA_F_DEPRECATED;
2737 			spin_unlock_bh(&ifp->lock);
2738 
2739 			if (!(flags&IFA_F_TENTATIVE))
2740 				ipv6_ifa_notify(0, ifp);
2741 		} else
2742 			spin_unlock_bh(&ifp->lock);
2743 
2744 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2745 				 create, now);
2746 
2747 		in6_ifa_put(ifp);
2748 		addrconf_verify(net);
2749 	}
2750 
2751 	return 0;
2752 }
2753 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2754 
2755 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2756 {
2757 	struct prefix_info *pinfo;
2758 	struct fib6_table *table;
2759 	__u32 valid_lft;
2760 	__u32 prefered_lft;
2761 	int addr_type, err;
2762 	u32 addr_flags = 0;
2763 	struct inet6_dev *in6_dev;
2764 	struct net *net = dev_net(dev);
2765 
2766 	pinfo = (struct prefix_info *) opt;
2767 
2768 	if (len < sizeof(struct prefix_info)) {
2769 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2770 		return;
2771 	}
2772 
2773 	/*
2774 	 *	Validation checks ([ADDRCONF], page 19)
2775 	 */
2776 
2777 	addr_type = ipv6_addr_type(&pinfo->prefix);
2778 
2779 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2780 		return;
2781 
2782 	valid_lft = ntohl(pinfo->valid);
2783 	prefered_lft = ntohl(pinfo->prefered);
2784 
2785 	if (prefered_lft > valid_lft) {
2786 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2787 		return;
2788 	}
2789 
2790 	in6_dev = in6_dev_get(dev);
2791 
2792 	if (!in6_dev) {
2793 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2794 				    dev->name);
2795 		return;
2796 	}
2797 
2798 	if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2799 		goto put;
2800 
2801 	/*
2802 	 *	Two things going on here:
2803 	 *	1) Add routes for on-link prefixes
2804 	 *	2) Configure prefixes with the auto flag set
2805 	 */
2806 
2807 	if (pinfo->onlink) {
2808 		struct fib6_info *rt;
2809 		unsigned long rt_expires;
2810 
2811 		/* Avoid arithmetic overflow. Really, we could
2812 		 * save rt_expires in seconds, likely valid_lft,
2813 		 * but it would require division in fib gc, that it
2814 		 * not good.
2815 		 */
2816 		if (HZ > USER_HZ)
2817 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2818 		else
2819 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2820 
2821 		if (addrconf_finite_timeout(rt_expires))
2822 			rt_expires *= HZ;
2823 
2824 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2825 					       pinfo->prefix_len,
2826 					       dev,
2827 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2828 					       RTF_DEFAULT, true);
2829 
2830 		if (rt) {
2831 			/* Autoconf prefix route */
2832 			if (valid_lft == 0) {
2833 				ip6_del_rt(net, rt, false);
2834 				rt = NULL;
2835 			} else {
2836 				table = rt->fib6_table;
2837 				spin_lock_bh(&table->tb6_lock);
2838 
2839 				if (addrconf_finite_timeout(rt_expires)) {
2840 					/* not infinity */
2841 					fib6_set_expires(rt, jiffies + rt_expires);
2842 					fib6_add_gc_list(rt);
2843 				} else {
2844 					fib6_clean_expires(rt);
2845 					fib6_remove_gc_list(rt);
2846 				}
2847 
2848 				spin_unlock_bh(&table->tb6_lock);
2849 			}
2850 		} else if (valid_lft) {
2851 			clock_t expires = 0;
2852 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2853 			if (addrconf_finite_timeout(rt_expires)) {
2854 				/* not infinity */
2855 				flags |= RTF_EXPIRES;
2856 				expires = jiffies_to_clock_t(rt_expires);
2857 			}
2858 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2859 					      0, dev, expires, flags,
2860 					      GFP_ATOMIC);
2861 		}
2862 		fib6_info_release(rt);
2863 	}
2864 
2865 	/* Try to figure out our local address for this prefix */
2866 
2867 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2868 		struct in6_addr addr;
2869 		bool tokenized = false, dev_addr_generated = false;
2870 
2871 		if (pinfo->prefix_len == 64) {
2872 			memcpy(&addr, &pinfo->prefix, 8);
2873 
2874 			if (!ipv6_addr_any(&in6_dev->token)) {
2875 				read_lock_bh(&in6_dev->lock);
2876 				memcpy(addr.s6_addr + 8,
2877 				       in6_dev->token.s6_addr + 8, 8);
2878 				read_unlock_bh(&in6_dev->lock);
2879 				tokenized = true;
2880 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2881 				   !ipv6_generate_stable_address(&addr, 0,
2882 								 in6_dev)) {
2883 				addr_flags |= IFA_F_STABLE_PRIVACY;
2884 				goto ok;
2885 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2886 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2887 				goto put;
2888 			} else {
2889 				dev_addr_generated = true;
2890 			}
2891 			goto ok;
2892 		}
2893 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2894 				    pinfo->prefix_len);
2895 		goto put;
2896 
2897 ok:
2898 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2899 						   &addr, addr_type,
2900 						   addr_flags, sllao,
2901 						   tokenized, valid_lft,
2902 						   prefered_lft);
2903 		if (err)
2904 			goto put;
2905 
2906 		/* Ignore error case here because previous prefix add addr was
2907 		 * successful which will be notified.
2908 		 */
2909 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2910 					      addr_type, addr_flags, sllao,
2911 					      tokenized, valid_lft,
2912 					      prefered_lft,
2913 					      dev_addr_generated);
2914 	}
2915 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2916 put:
2917 	in6_dev_put(in6_dev);
2918 }
2919 
2920 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2921 		struct in6_ifreq *ireq)
2922 {
2923 	struct ip_tunnel_parm_kern p = { };
2924 	int err;
2925 
2926 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2927 		return -EADDRNOTAVAIL;
2928 
2929 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2930 	p.iph.version = 4;
2931 	p.iph.ihl = 5;
2932 	p.iph.protocol = IPPROTO_IPV6;
2933 	p.iph.ttl = 64;
2934 
2935 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2936 		return -EOPNOTSUPP;
2937 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2938 	if (err)
2939 		return err;
2940 
2941 	dev = __dev_get_by_name(net, p.name);
2942 	if (!dev)
2943 		return -ENOBUFS;
2944 	return dev_open(dev, NULL);
2945 }
2946 
2947 /*
2948  *	Set destination address.
2949  *	Special case for SIT interfaces where we create a new "virtual"
2950  *	device.
2951  */
2952 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2953 {
2954 	struct net_device *dev;
2955 	struct in6_ifreq ireq;
2956 	int err = -ENODEV;
2957 
2958 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2959 		return -ENODEV;
2960 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2961 		return -EFAULT;
2962 
2963 	rtnl_lock();
2964 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2965 	if (dev && dev->type == ARPHRD_SIT)
2966 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2967 	rtnl_unlock();
2968 	return err;
2969 }
2970 
2971 static int ipv6_mc_config(struct sock *sk, bool join,
2972 			  const struct in6_addr *addr, int ifindex)
2973 {
2974 	int ret;
2975 
2976 	ASSERT_RTNL();
2977 
2978 	lock_sock(sk);
2979 	if (join)
2980 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2981 	else
2982 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2983 	release_sock(sk);
2984 
2985 	return ret;
2986 }
2987 
2988 /*
2989  *	Manual configuration of address on an interface
2990  */
2991 static int inet6_addr_add(struct net *net, int ifindex,
2992 			  struct ifa6_config *cfg,
2993 			  struct netlink_ext_ack *extack)
2994 {
2995 	struct inet6_ifaddr *ifp;
2996 	struct inet6_dev *idev;
2997 	struct net_device *dev;
2998 	unsigned long timeout;
2999 	clock_t expires;
3000 	u32 flags;
3001 
3002 	ASSERT_RTNL();
3003 
3004 	if (cfg->plen > 128) {
3005 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3006 		return -EINVAL;
3007 	}
3008 
3009 	/* check the lifetime */
3010 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
3011 		NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
3012 		return -EINVAL;
3013 	}
3014 
3015 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3016 		NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3017 		return -EINVAL;
3018 	}
3019 
3020 	dev = __dev_get_by_index(net, ifindex);
3021 	if (!dev)
3022 		return -ENODEV;
3023 
3024 	idev = addrconf_add_dev(dev);
3025 	if (IS_ERR(idev)) {
3026 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3027 		return PTR_ERR(idev);
3028 	}
3029 
3030 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3031 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3032 					 true, cfg->pfx, ifindex);
3033 
3034 		if (ret < 0) {
3035 			NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3036 			return ret;
3037 		}
3038 	}
3039 
3040 	cfg->scope = ipv6_addr_scope(cfg->pfx);
3041 
3042 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
3043 	if (addrconf_finite_timeout(timeout)) {
3044 		expires = jiffies_to_clock_t(timeout * HZ);
3045 		cfg->valid_lft = timeout;
3046 		flags = RTF_EXPIRES;
3047 	} else {
3048 		expires = 0;
3049 		flags = 0;
3050 		cfg->ifa_flags |= IFA_F_PERMANENT;
3051 	}
3052 
3053 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3054 	if (addrconf_finite_timeout(timeout)) {
3055 		if (timeout == 0)
3056 			cfg->ifa_flags |= IFA_F_DEPRECATED;
3057 		cfg->preferred_lft = timeout;
3058 	}
3059 
3060 	ifp = ipv6_add_addr(idev, cfg, true, extack);
3061 	if (!IS_ERR(ifp)) {
3062 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3063 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3064 					      ifp->rt_priority, dev, expires,
3065 					      flags, GFP_KERNEL);
3066 		}
3067 
3068 		/* Send a netlink notification if DAD is enabled and
3069 		 * optimistic flag is not set
3070 		 */
3071 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3072 			ipv6_ifa_notify(0, ifp);
3073 		/*
3074 		 * Note that section 3.1 of RFC 4429 indicates
3075 		 * that the Optimistic flag should not be set for
3076 		 * manually configured addresses
3077 		 */
3078 		addrconf_dad_start(ifp);
3079 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3080 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
3081 					 cfg->preferred_lft, true, jiffies);
3082 		in6_ifa_put(ifp);
3083 		addrconf_verify_rtnl(net);
3084 		return 0;
3085 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3086 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3087 			       cfg->pfx, ifindex);
3088 	}
3089 
3090 	return PTR_ERR(ifp);
3091 }
3092 
3093 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3094 			  const struct in6_addr *pfx, unsigned int plen,
3095 			  struct netlink_ext_ack *extack)
3096 {
3097 	struct inet6_ifaddr *ifp;
3098 	struct inet6_dev *idev;
3099 	struct net_device *dev;
3100 
3101 	if (plen > 128) {
3102 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3103 		return -EINVAL;
3104 	}
3105 
3106 	dev = __dev_get_by_index(net, ifindex);
3107 	if (!dev) {
3108 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3109 		return -ENODEV;
3110 	}
3111 
3112 	idev = __in6_dev_get(dev);
3113 	if (!idev) {
3114 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3115 		return -ENXIO;
3116 	}
3117 
3118 	read_lock_bh(&idev->lock);
3119 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3120 		if (ifp->prefix_len == plen &&
3121 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3122 			in6_ifa_hold(ifp);
3123 			read_unlock_bh(&idev->lock);
3124 
3125 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3126 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
3127 				manage_tempaddrs(idev, ifp, 0, 0, false,
3128 						 jiffies);
3129 			ipv6_del_addr(ifp);
3130 			addrconf_verify_rtnl(net);
3131 			if (ipv6_addr_is_multicast(pfx)) {
3132 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3133 					       false, pfx, dev->ifindex);
3134 			}
3135 			return 0;
3136 		}
3137 	}
3138 	read_unlock_bh(&idev->lock);
3139 
3140 	NL_SET_ERR_MSG_MOD(extack, "address not found");
3141 	return -EADDRNOTAVAIL;
3142 }
3143 
3144 
3145 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3146 {
3147 	struct ifa6_config cfg = {
3148 		.ifa_flags = IFA_F_PERMANENT,
3149 		.preferred_lft = INFINITY_LIFE_TIME,
3150 		.valid_lft = INFINITY_LIFE_TIME,
3151 	};
3152 	struct in6_ifreq ireq;
3153 	int err;
3154 
3155 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3156 		return -EPERM;
3157 
3158 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3159 		return -EFAULT;
3160 
3161 	cfg.pfx = &ireq.ifr6_addr;
3162 	cfg.plen = ireq.ifr6_prefixlen;
3163 
3164 	rtnl_lock();
3165 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3166 	rtnl_unlock();
3167 	return err;
3168 }
3169 
3170 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3171 {
3172 	struct in6_ifreq ireq;
3173 	int err;
3174 
3175 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3176 		return -EPERM;
3177 
3178 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3179 		return -EFAULT;
3180 
3181 	rtnl_lock();
3182 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3183 			     ireq.ifr6_prefixlen, NULL);
3184 	rtnl_unlock();
3185 	return err;
3186 }
3187 
3188 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3189 		     int plen, int scope, u8 proto)
3190 {
3191 	struct inet6_ifaddr *ifp;
3192 	struct ifa6_config cfg = {
3193 		.pfx = addr,
3194 		.plen = plen,
3195 		.ifa_flags = IFA_F_PERMANENT,
3196 		.valid_lft = INFINITY_LIFE_TIME,
3197 		.preferred_lft = INFINITY_LIFE_TIME,
3198 		.scope = scope,
3199 		.ifa_proto = proto
3200 	};
3201 
3202 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3203 	if (!IS_ERR(ifp)) {
3204 		spin_lock_bh(&ifp->lock);
3205 		ifp->flags &= ~IFA_F_TENTATIVE;
3206 		spin_unlock_bh(&ifp->lock);
3207 		rt_genid_bump_ipv6(dev_net(idev->dev));
3208 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3209 		in6_ifa_put(ifp);
3210 	}
3211 }
3212 
3213 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3214 static void add_v4_addrs(struct inet6_dev *idev)
3215 {
3216 	struct in6_addr addr;
3217 	struct net_device *dev;
3218 	struct net *net = dev_net(idev->dev);
3219 	int scope, plen, offset = 0;
3220 	u32 pflags = 0;
3221 
3222 	ASSERT_RTNL();
3223 
3224 	memset(&addr, 0, sizeof(struct in6_addr));
3225 	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3226 	if (idev->dev->addr_len == sizeof(struct in6_addr))
3227 		offset = sizeof(struct in6_addr) - 4;
3228 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3229 
3230 	if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3231 		scope = IPV6_ADDR_COMPATv4;
3232 		plen = 96;
3233 		pflags |= RTF_NONEXTHOP;
3234 	} else {
3235 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3236 			return;
3237 
3238 		addr.s6_addr32[0] = htonl(0xfe800000);
3239 		scope = IFA_LINK;
3240 		plen = 64;
3241 	}
3242 
3243 	if (addr.s6_addr32[3]) {
3244 		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3245 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3246 				      GFP_KERNEL);
3247 		return;
3248 	}
3249 
3250 	for_each_netdev(net, dev) {
3251 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3252 		if (in_dev && (dev->flags & IFF_UP)) {
3253 			struct in_ifaddr *ifa;
3254 			int flag = scope;
3255 
3256 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3257 				addr.s6_addr32[3] = ifa->ifa_local;
3258 
3259 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3260 					continue;
3261 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3262 					if (idev->dev->flags&IFF_POINTOPOINT)
3263 						continue;
3264 					flag |= IFA_HOST;
3265 				}
3266 
3267 				add_addr(idev, &addr, plen, flag,
3268 					 IFAPROT_UNSPEC);
3269 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3270 						      0, pflags, GFP_KERNEL);
3271 			}
3272 		}
3273 	}
3274 }
3275 #endif
3276 
3277 static void init_loopback(struct net_device *dev)
3278 {
3279 	struct inet6_dev  *idev;
3280 
3281 	/* ::1 */
3282 
3283 	ASSERT_RTNL();
3284 
3285 	idev = ipv6_find_idev(dev);
3286 	if (IS_ERR(idev)) {
3287 		pr_debug("%s: add_dev failed\n", __func__);
3288 		return;
3289 	}
3290 
3291 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3292 }
3293 
3294 void addrconf_add_linklocal(struct inet6_dev *idev,
3295 			    const struct in6_addr *addr, u32 flags)
3296 {
3297 	struct ifa6_config cfg = {
3298 		.pfx = addr,
3299 		.plen = 64,
3300 		.ifa_flags = flags | IFA_F_PERMANENT,
3301 		.valid_lft = INFINITY_LIFE_TIME,
3302 		.preferred_lft = INFINITY_LIFE_TIME,
3303 		.scope = IFA_LINK,
3304 		.ifa_proto = IFAPROT_KERNEL_LL
3305 	};
3306 	struct inet6_ifaddr *ifp;
3307 
3308 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3309 	if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3310 	     READ_ONCE(idev->cnf.optimistic_dad)) &&
3311 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3312 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3313 #endif
3314 
3315 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3316 	if (!IS_ERR(ifp)) {
3317 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3318 				      0, 0, GFP_ATOMIC);
3319 		addrconf_dad_start(ifp);
3320 		in6_ifa_put(ifp);
3321 	}
3322 }
3323 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3324 
3325 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3326 {
3327 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3328 		return true;
3329 
3330 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3331 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3332 		return true;
3333 
3334 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3335 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3336 		return true;
3337 
3338 	return false;
3339 }
3340 
3341 static int ipv6_generate_stable_address(struct in6_addr *address,
3342 					u8 dad_count,
3343 					const struct inet6_dev *idev)
3344 {
3345 	static DEFINE_SPINLOCK(lock);
3346 	static __u32 digest[SHA1_DIGEST_WORDS];
3347 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3348 
3349 	static union {
3350 		char __data[SHA1_BLOCK_SIZE];
3351 		struct {
3352 			struct in6_addr secret;
3353 			__be32 prefix[2];
3354 			unsigned char hwaddr[MAX_ADDR_LEN];
3355 			u8 dad_count;
3356 		} __packed;
3357 	} data;
3358 
3359 	struct in6_addr secret;
3360 	struct in6_addr temp;
3361 	struct net *net = dev_net(idev->dev);
3362 
3363 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3364 
3365 	if (idev->cnf.stable_secret.initialized)
3366 		secret = idev->cnf.stable_secret.secret;
3367 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3368 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3369 	else
3370 		return -1;
3371 
3372 retry:
3373 	spin_lock_bh(&lock);
3374 
3375 	sha1_init(digest);
3376 	memset(&data, 0, sizeof(data));
3377 	memset(workspace, 0, sizeof(workspace));
3378 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3379 	data.prefix[0] = address->s6_addr32[0];
3380 	data.prefix[1] = address->s6_addr32[1];
3381 	data.secret = secret;
3382 	data.dad_count = dad_count;
3383 
3384 	sha1_transform(digest, data.__data, workspace);
3385 
3386 	temp = *address;
3387 	temp.s6_addr32[2] = (__force __be32)digest[0];
3388 	temp.s6_addr32[3] = (__force __be32)digest[1];
3389 
3390 	spin_unlock_bh(&lock);
3391 
3392 	if (ipv6_reserved_interfaceid(temp)) {
3393 		dad_count++;
3394 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3395 			return -1;
3396 		goto retry;
3397 	}
3398 
3399 	*address = temp;
3400 	return 0;
3401 }
3402 
3403 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3404 {
3405 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3406 
3407 	if (s->initialized)
3408 		return;
3409 	s = &idev->cnf.stable_secret;
3410 	get_random_bytes(&s->secret, sizeof(s->secret));
3411 	s->initialized = true;
3412 }
3413 
3414 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3415 {
3416 	struct in6_addr addr;
3417 
3418 	/* no link local addresses on L3 master devices */
3419 	if (netif_is_l3_master(idev->dev))
3420 		return;
3421 
3422 	/* no link local addresses on devices flagged as slaves */
3423 	if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3424 		return;
3425 
3426 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3427 
3428 	switch (idev->cnf.addr_gen_mode) {
3429 	case IN6_ADDR_GEN_MODE_RANDOM:
3430 		ipv6_gen_mode_random_init(idev);
3431 		fallthrough;
3432 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3433 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3434 			addrconf_add_linklocal(idev, &addr,
3435 					       IFA_F_STABLE_PRIVACY);
3436 		else if (prefix_route)
3437 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3438 					      0, 0, GFP_KERNEL);
3439 		break;
3440 	case IN6_ADDR_GEN_MODE_EUI64:
3441 		/* addrconf_add_linklocal also adds a prefix_route and we
3442 		 * only need to care about prefix routes if ipv6_generate_eui64
3443 		 * couldn't generate one.
3444 		 */
3445 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3446 			addrconf_add_linklocal(idev, &addr, 0);
3447 		else if (prefix_route)
3448 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3449 					      0, 0, GFP_KERNEL);
3450 		break;
3451 	case IN6_ADDR_GEN_MODE_NONE:
3452 	default:
3453 		/* will not add any link local address */
3454 		break;
3455 	}
3456 }
3457 
3458 static void addrconf_dev_config(struct net_device *dev)
3459 {
3460 	struct inet6_dev *idev;
3461 
3462 	ASSERT_RTNL();
3463 
3464 	if ((dev->type != ARPHRD_ETHER) &&
3465 	    (dev->type != ARPHRD_FDDI) &&
3466 	    (dev->type != ARPHRD_ARCNET) &&
3467 	    (dev->type != ARPHRD_INFINIBAND) &&
3468 	    (dev->type != ARPHRD_IEEE1394) &&
3469 	    (dev->type != ARPHRD_TUNNEL6) &&
3470 	    (dev->type != ARPHRD_6LOWPAN) &&
3471 	    (dev->type != ARPHRD_TUNNEL) &&
3472 	    (dev->type != ARPHRD_NONE) &&
3473 	    (dev->type != ARPHRD_RAWIP)) {
3474 		/* Alas, we support only Ethernet autoconfiguration. */
3475 		idev = __in6_dev_get(dev);
3476 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3477 		    dev->flags & IFF_MULTICAST)
3478 			ipv6_mc_up(idev);
3479 		return;
3480 	}
3481 
3482 	idev = addrconf_add_dev(dev);
3483 	if (IS_ERR(idev))
3484 		return;
3485 
3486 	/* this device type has no EUI support */
3487 	if (dev->type == ARPHRD_NONE &&
3488 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3489 		WRITE_ONCE(idev->cnf.addr_gen_mode,
3490 			   IN6_ADDR_GEN_MODE_RANDOM);
3491 
3492 	addrconf_addr_gen(idev, false);
3493 }
3494 
3495 #if IS_ENABLED(CONFIG_IPV6_SIT)
3496 static void addrconf_sit_config(struct net_device *dev)
3497 {
3498 	struct inet6_dev *idev;
3499 
3500 	ASSERT_RTNL();
3501 
3502 	/*
3503 	 * Configure the tunnel with one of our IPv4
3504 	 * addresses... we should configure all of
3505 	 * our v4 addrs in the tunnel
3506 	 */
3507 
3508 	idev = ipv6_find_idev(dev);
3509 	if (IS_ERR(idev)) {
3510 		pr_debug("%s: add_dev failed\n", __func__);
3511 		return;
3512 	}
3513 
3514 	if (dev->priv_flags & IFF_ISATAP) {
3515 		addrconf_addr_gen(idev, false);
3516 		return;
3517 	}
3518 
3519 	add_v4_addrs(idev);
3520 
3521 	if (dev->flags&IFF_POINTOPOINT)
3522 		addrconf_add_mroute(dev);
3523 }
3524 #endif
3525 
3526 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3527 static void addrconf_gre_config(struct net_device *dev)
3528 {
3529 	struct inet6_dev *idev;
3530 
3531 	ASSERT_RTNL();
3532 
3533 	idev = ipv6_find_idev(dev);
3534 	if (IS_ERR(idev)) {
3535 		pr_debug("%s: add_dev failed\n", __func__);
3536 		return;
3537 	}
3538 
3539 	if (dev->type == ARPHRD_ETHER) {
3540 		addrconf_addr_gen(idev, true);
3541 		return;
3542 	}
3543 
3544 	add_v4_addrs(idev);
3545 
3546 	if (dev->flags & IFF_POINTOPOINT)
3547 		addrconf_add_mroute(dev);
3548 }
3549 #endif
3550 
3551 static void addrconf_init_auto_addrs(struct net_device *dev)
3552 {
3553 	switch (dev->type) {
3554 #if IS_ENABLED(CONFIG_IPV6_SIT)
3555 	case ARPHRD_SIT:
3556 		addrconf_sit_config(dev);
3557 		break;
3558 #endif
3559 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3560 	case ARPHRD_IP6GRE:
3561 	case ARPHRD_IPGRE:
3562 		addrconf_gre_config(dev);
3563 		break;
3564 #endif
3565 	case ARPHRD_LOOPBACK:
3566 		init_loopback(dev);
3567 		break;
3568 
3569 	default:
3570 		addrconf_dev_config(dev);
3571 		break;
3572 	}
3573 }
3574 
3575 static int fixup_permanent_addr(struct net *net,
3576 				struct inet6_dev *idev,
3577 				struct inet6_ifaddr *ifp)
3578 {
3579 	/* !fib6_node means the host route was removed from the
3580 	 * FIB, for example, if 'lo' device is taken down. In that
3581 	 * case regenerate the host route.
3582 	 */
3583 	if (!ifp->rt || !ifp->rt->fib6_node) {
3584 		struct fib6_info *f6i, *prev;
3585 
3586 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3587 					 GFP_ATOMIC, NULL);
3588 		if (IS_ERR(f6i))
3589 			return PTR_ERR(f6i);
3590 
3591 		/* ifp->rt can be accessed outside of rtnl */
3592 		spin_lock(&ifp->lock);
3593 		prev = ifp->rt;
3594 		ifp->rt = f6i;
3595 		spin_unlock(&ifp->lock);
3596 
3597 		fib6_info_release(prev);
3598 	}
3599 
3600 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3601 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3602 				      ifp->rt_priority, idev->dev, 0, 0,
3603 				      GFP_ATOMIC);
3604 	}
3605 
3606 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3607 		addrconf_dad_start(ifp);
3608 
3609 	return 0;
3610 }
3611 
3612 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3613 {
3614 	struct inet6_ifaddr *ifp, *tmp;
3615 	struct inet6_dev *idev;
3616 
3617 	idev = __in6_dev_get(dev);
3618 	if (!idev)
3619 		return;
3620 
3621 	write_lock_bh(&idev->lock);
3622 
3623 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3624 		if ((ifp->flags & IFA_F_PERMANENT) &&
3625 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3626 			write_unlock_bh(&idev->lock);
3627 			in6_ifa_hold(ifp);
3628 			ipv6_del_addr(ifp);
3629 			write_lock_bh(&idev->lock);
3630 
3631 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3632 					     idev->dev->name, &ifp->addr);
3633 		}
3634 	}
3635 
3636 	write_unlock_bh(&idev->lock);
3637 }
3638 
3639 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3640 			   void *ptr)
3641 {
3642 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3643 	struct netdev_notifier_change_info *change_info;
3644 	struct netdev_notifier_changeupper_info *info;
3645 	struct inet6_dev *idev = __in6_dev_get(dev);
3646 	struct net *net = dev_net(dev);
3647 	int run_pending = 0;
3648 	int err;
3649 
3650 	switch (event) {
3651 	case NETDEV_REGISTER:
3652 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3653 			idev = ipv6_add_dev(dev);
3654 			if (IS_ERR(idev))
3655 				return notifier_from_errno(PTR_ERR(idev));
3656 		}
3657 		break;
3658 
3659 	case NETDEV_CHANGEMTU:
3660 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3661 		if (dev->mtu < IPV6_MIN_MTU) {
3662 			addrconf_ifdown(dev, dev != net->loopback_dev);
3663 			break;
3664 		}
3665 
3666 		if (idev) {
3667 			rt6_mtu_change(dev, dev->mtu);
3668 			WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3669 			break;
3670 		}
3671 
3672 		/* allocate new idev */
3673 		idev = ipv6_add_dev(dev);
3674 		if (IS_ERR(idev))
3675 			break;
3676 
3677 		/* device is still not ready */
3678 		if (!(idev->if_flags & IF_READY))
3679 			break;
3680 
3681 		run_pending = 1;
3682 		fallthrough;
3683 	case NETDEV_UP:
3684 	case NETDEV_CHANGE:
3685 		if (idev && idev->cnf.disable_ipv6)
3686 			break;
3687 
3688 		if (dev->priv_flags & IFF_NO_ADDRCONF) {
3689 			if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3690 			    dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3691 				ipv6_mc_up(idev);
3692 			break;
3693 		}
3694 
3695 		if (event == NETDEV_UP) {
3696 			/* restore routes for permanent addresses */
3697 			addrconf_permanent_addr(net, dev);
3698 
3699 			if (!addrconf_link_ready(dev)) {
3700 				/* device is not ready yet. */
3701 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3702 					 dev->name);
3703 				break;
3704 			}
3705 
3706 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3707 				idev = ipv6_add_dev(dev);
3708 
3709 			if (!IS_ERR_OR_NULL(idev)) {
3710 				idev->if_flags |= IF_READY;
3711 				run_pending = 1;
3712 			}
3713 		} else if (event == NETDEV_CHANGE) {
3714 			if (!addrconf_link_ready(dev)) {
3715 				/* device is still not ready. */
3716 				rt6_sync_down_dev(dev, event);
3717 				break;
3718 			}
3719 
3720 			if (!IS_ERR_OR_NULL(idev)) {
3721 				if (idev->if_flags & IF_READY) {
3722 					/* device is already configured -
3723 					 * but resend MLD reports, we might
3724 					 * have roamed and need to update
3725 					 * multicast snooping switches
3726 					 */
3727 					ipv6_mc_up(idev);
3728 					change_info = ptr;
3729 					if (change_info->flags_changed & IFF_NOARP)
3730 						addrconf_dad_run(idev, true);
3731 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3732 					break;
3733 				}
3734 				idev->if_flags |= IF_READY;
3735 			}
3736 
3737 			pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3738 				 dev->name);
3739 
3740 			run_pending = 1;
3741 		}
3742 
3743 		addrconf_init_auto_addrs(dev);
3744 
3745 		if (!IS_ERR_OR_NULL(idev)) {
3746 			if (run_pending)
3747 				addrconf_dad_run(idev, false);
3748 
3749 			/* Device has an address by now */
3750 			rt6_sync_up(dev, RTNH_F_DEAD);
3751 
3752 			/*
3753 			 * If the MTU changed during the interface down,
3754 			 * when the interface up, the changed MTU must be
3755 			 * reflected in the idev as well as routers.
3756 			 */
3757 			if (idev->cnf.mtu6 != dev->mtu &&
3758 			    dev->mtu >= IPV6_MIN_MTU) {
3759 				rt6_mtu_change(dev, dev->mtu);
3760 				WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3761 			}
3762 			WRITE_ONCE(idev->tstamp, jiffies);
3763 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3764 
3765 			/*
3766 			 * If the changed mtu during down is lower than
3767 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3768 			 */
3769 			if (dev->mtu < IPV6_MIN_MTU)
3770 				addrconf_ifdown(dev, dev != net->loopback_dev);
3771 		}
3772 		break;
3773 
3774 	case NETDEV_DOWN:
3775 	case NETDEV_UNREGISTER:
3776 		/*
3777 		 *	Remove all addresses from this interface.
3778 		 */
3779 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3780 		break;
3781 
3782 	case NETDEV_CHANGENAME:
3783 		if (idev) {
3784 			snmp6_unregister_dev(idev);
3785 			addrconf_sysctl_unregister(idev);
3786 			err = addrconf_sysctl_register(idev);
3787 			if (err)
3788 				return notifier_from_errno(err);
3789 			err = snmp6_register_dev(idev);
3790 			if (err) {
3791 				addrconf_sysctl_unregister(idev);
3792 				return notifier_from_errno(err);
3793 			}
3794 		}
3795 		break;
3796 
3797 	case NETDEV_PRE_TYPE_CHANGE:
3798 	case NETDEV_POST_TYPE_CHANGE:
3799 		if (idev)
3800 			addrconf_type_change(dev, event);
3801 		break;
3802 
3803 	case NETDEV_CHANGEUPPER:
3804 		info = ptr;
3805 
3806 		/* flush all routes if dev is linked to or unlinked from
3807 		 * an L3 master device (e.g., VRF)
3808 		 */
3809 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3810 			addrconf_ifdown(dev, false);
3811 	}
3812 
3813 	return NOTIFY_OK;
3814 }
3815 
3816 /*
3817  *	addrconf module should be notified of a device going up
3818  */
3819 static struct notifier_block ipv6_dev_notf = {
3820 	.notifier_call = addrconf_notify,
3821 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3822 };
3823 
3824 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3825 {
3826 	struct inet6_dev *idev;
3827 	ASSERT_RTNL();
3828 
3829 	idev = __in6_dev_get(dev);
3830 
3831 	if (event == NETDEV_POST_TYPE_CHANGE)
3832 		ipv6_mc_remap(idev);
3833 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3834 		ipv6_mc_unmap(idev);
3835 }
3836 
3837 static bool addr_is_local(const struct in6_addr *addr)
3838 {
3839 	return ipv6_addr_type(addr) &
3840 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3841 }
3842 
3843 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3844 {
3845 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3846 	struct net *net = dev_net(dev);
3847 	struct inet6_dev *idev;
3848 	struct inet6_ifaddr *ifa;
3849 	LIST_HEAD(tmp_addr_list);
3850 	bool keep_addr = false;
3851 	bool was_ready;
3852 	int state, i;
3853 
3854 	ASSERT_RTNL();
3855 
3856 	rt6_disable_ip(dev, event);
3857 
3858 	idev = __in6_dev_get(dev);
3859 	if (!idev)
3860 		return -ENODEV;
3861 
3862 	/*
3863 	 * Step 1: remove reference to ipv6 device from parent device.
3864 	 *	   Do not dev_put!
3865 	 */
3866 	if (unregister) {
3867 		idev->dead = 1;
3868 
3869 		/* protected by rtnl_lock */
3870 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3871 
3872 		/* Step 1.5: remove snmp6 entry */
3873 		snmp6_unregister_dev(idev);
3874 
3875 	}
3876 
3877 	/* combine the user config with event to determine if permanent
3878 	 * addresses are to be removed from address hash table
3879 	 */
3880 	if (!unregister && !idev->cnf.disable_ipv6) {
3881 		/* aggregate the system setting and interface setting */
3882 		int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3883 
3884 		if (!_keep_addr)
3885 			_keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3886 
3887 		keep_addr = (_keep_addr > 0);
3888 	}
3889 
3890 	/* Step 2: clear hash table */
3891 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3892 		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3893 
3894 		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3895 restart:
3896 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3897 			if (ifa->idev == idev) {
3898 				addrconf_del_dad_work(ifa);
3899 				/* combined flag + permanent flag decide if
3900 				 * address is retained on a down event
3901 				 */
3902 				if (!keep_addr ||
3903 				    !(ifa->flags & IFA_F_PERMANENT) ||
3904 				    addr_is_local(&ifa->addr)) {
3905 					hlist_del_init_rcu(&ifa->addr_lst);
3906 					goto restart;
3907 				}
3908 			}
3909 		}
3910 		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3911 	}
3912 
3913 	write_lock_bh(&idev->lock);
3914 
3915 	addrconf_del_rs_timer(idev);
3916 
3917 	/* Step 2: clear flags for stateless addrconf, repeated down
3918 	 *         detection
3919 	 */
3920 	was_ready = idev->if_flags & IF_READY;
3921 	if (!unregister)
3922 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3923 
3924 	/* Step 3: clear tempaddr list */
3925 	while (!list_empty(&idev->tempaddr_list)) {
3926 		ifa = list_first_entry(&idev->tempaddr_list,
3927 				       struct inet6_ifaddr, tmp_list);
3928 		list_del(&ifa->tmp_list);
3929 		write_unlock_bh(&idev->lock);
3930 		spin_lock_bh(&ifa->lock);
3931 
3932 		if (ifa->ifpub) {
3933 			in6_ifa_put(ifa->ifpub);
3934 			ifa->ifpub = NULL;
3935 		}
3936 		spin_unlock_bh(&ifa->lock);
3937 		in6_ifa_put(ifa);
3938 		write_lock_bh(&idev->lock);
3939 	}
3940 
3941 	list_for_each_entry(ifa, &idev->addr_list, if_list)
3942 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3943 	write_unlock_bh(&idev->lock);
3944 
3945 	while (!list_empty(&tmp_addr_list)) {
3946 		struct fib6_info *rt = NULL;
3947 		bool keep;
3948 
3949 		ifa = list_first_entry(&tmp_addr_list,
3950 				       struct inet6_ifaddr, if_list_aux);
3951 		list_del(&ifa->if_list_aux);
3952 
3953 		addrconf_del_dad_work(ifa);
3954 
3955 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3956 			!addr_is_local(&ifa->addr);
3957 
3958 		spin_lock_bh(&ifa->lock);
3959 
3960 		if (keep) {
3961 			/* set state to skip the notifier below */
3962 			state = INET6_IFADDR_STATE_DEAD;
3963 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3964 			if (!(ifa->flags & IFA_F_NODAD))
3965 				ifa->flags |= IFA_F_TENTATIVE;
3966 
3967 			rt = ifa->rt;
3968 			ifa->rt = NULL;
3969 		} else {
3970 			state = ifa->state;
3971 			ifa->state = INET6_IFADDR_STATE_DEAD;
3972 		}
3973 
3974 		spin_unlock_bh(&ifa->lock);
3975 
3976 		if (rt)
3977 			ip6_del_rt(net, rt, false);
3978 
3979 		if (state != INET6_IFADDR_STATE_DEAD) {
3980 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3981 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3982 		} else {
3983 			if (idev->cnf.forwarding)
3984 				addrconf_leave_anycast(ifa);
3985 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3986 		}
3987 
3988 		if (!keep) {
3989 			write_lock_bh(&idev->lock);
3990 			list_del_rcu(&ifa->if_list);
3991 			write_unlock_bh(&idev->lock);
3992 			in6_ifa_put(ifa);
3993 		}
3994 	}
3995 
3996 	/* Step 5: Discard anycast and multicast list */
3997 	if (unregister) {
3998 		ipv6_ac_destroy_dev(idev);
3999 		ipv6_mc_destroy_dev(idev);
4000 	} else if (was_ready) {
4001 		ipv6_mc_down(idev);
4002 	}
4003 
4004 	WRITE_ONCE(idev->tstamp, jiffies);
4005 	idev->ra_mtu = 0;
4006 
4007 	/* Last: Shot the device (if unregistered) */
4008 	if (unregister) {
4009 		addrconf_sysctl_unregister(idev);
4010 		neigh_parms_release(&nd_tbl, idev->nd_parms);
4011 		neigh_ifdown(&nd_tbl, dev);
4012 		in6_dev_put(idev);
4013 	}
4014 	return 0;
4015 }
4016 
4017 static void addrconf_rs_timer(struct timer_list *t)
4018 {
4019 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
4020 	struct net_device *dev = idev->dev;
4021 	struct in6_addr lladdr;
4022 	int rtr_solicits;
4023 
4024 	write_lock(&idev->lock);
4025 	if (idev->dead || !(idev->if_flags & IF_READY))
4026 		goto out;
4027 
4028 	if (!ipv6_accept_ra(idev))
4029 		goto out;
4030 
4031 	/* Announcement received after solicitation was sent */
4032 	if (idev->if_flags & IF_RA_RCVD)
4033 		goto out;
4034 
4035 	rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4036 
4037 	if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4038 		write_unlock(&idev->lock);
4039 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4040 			ndisc_send_rs(dev, &lladdr,
4041 				      &in6addr_linklocal_allrouters);
4042 		else
4043 			goto put;
4044 
4045 		write_lock(&idev->lock);
4046 		idev->rs_interval = rfc3315_s14_backoff_update(
4047 				idev->rs_interval,
4048 				READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4049 		/* The wait after the last probe can be shorter */
4050 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4051 					     READ_ONCE(idev->cnf.rtr_solicits)) ?
4052 				      READ_ONCE(idev->cnf.rtr_solicit_delay) :
4053 				      idev->rs_interval);
4054 	} else {
4055 		/*
4056 		 * Note: we do not support deprecated "all on-link"
4057 		 * assumption any longer.
4058 		 */
4059 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4060 	}
4061 
4062 out:
4063 	write_unlock(&idev->lock);
4064 put:
4065 	in6_dev_put(idev);
4066 }
4067 
4068 /*
4069  *	Duplicate Address Detection
4070  */
4071 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4072 {
4073 	struct inet6_dev *idev = ifp->idev;
4074 	unsigned long rand_num;
4075 	u64 nonce;
4076 
4077 	if (ifp->flags & IFA_F_OPTIMISTIC)
4078 		rand_num = 0;
4079 	else
4080 		rand_num = get_random_u32_below(
4081 				READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4082 
4083 	nonce = 0;
4084 	if (READ_ONCE(idev->cnf.enhanced_dad) ||
4085 	    READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4086 		do
4087 			get_random_bytes(&nonce, 6);
4088 		while (nonce == 0);
4089 	}
4090 	ifp->dad_nonce = nonce;
4091 	ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4092 	addrconf_mod_dad_work(ifp, rand_num);
4093 }
4094 
4095 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4096 {
4097 	struct inet6_dev *idev = ifp->idev;
4098 	struct net_device *dev = idev->dev;
4099 	bool bump_id, notify = false;
4100 	struct net *net;
4101 
4102 	addrconf_join_solict(dev, &ifp->addr);
4103 
4104 	read_lock_bh(&idev->lock);
4105 	spin_lock(&ifp->lock);
4106 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
4107 		goto out;
4108 
4109 	net = dev_net(dev);
4110 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4111 	    (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4112 	     READ_ONCE(idev->cnf.accept_dad) < 1) ||
4113 	    !(ifp->flags&IFA_F_TENTATIVE) ||
4114 	    ifp->flags & IFA_F_NODAD) {
4115 		bool send_na = false;
4116 
4117 		if (ifp->flags & IFA_F_TENTATIVE &&
4118 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4119 			send_na = true;
4120 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4121 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4122 		spin_unlock(&ifp->lock);
4123 		read_unlock_bh(&idev->lock);
4124 
4125 		addrconf_dad_completed(ifp, bump_id, send_na);
4126 		return;
4127 	}
4128 
4129 	if (!(idev->if_flags & IF_READY)) {
4130 		spin_unlock(&ifp->lock);
4131 		read_unlock_bh(&idev->lock);
4132 		/*
4133 		 * If the device is not ready:
4134 		 * - keep it tentative if it is a permanent address.
4135 		 * - otherwise, kill it.
4136 		 */
4137 		in6_ifa_hold(ifp);
4138 		addrconf_dad_stop(ifp, 0);
4139 		return;
4140 	}
4141 
4142 	/*
4143 	 * Optimistic nodes can start receiving
4144 	 * Frames right away
4145 	 */
4146 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4147 		ip6_ins_rt(net, ifp->rt);
4148 		if (ipv6_use_optimistic_addr(net, idev)) {
4149 			/* Because optimistic nodes can use this address,
4150 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4151 			 */
4152 			notify = true;
4153 		}
4154 	}
4155 
4156 	addrconf_dad_kick(ifp);
4157 out:
4158 	spin_unlock(&ifp->lock);
4159 	read_unlock_bh(&idev->lock);
4160 	if (notify)
4161 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4162 }
4163 
4164 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4165 {
4166 	bool begin_dad = false;
4167 
4168 	spin_lock_bh(&ifp->lock);
4169 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4170 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4171 		begin_dad = true;
4172 	}
4173 	spin_unlock_bh(&ifp->lock);
4174 
4175 	if (begin_dad)
4176 		addrconf_mod_dad_work(ifp, 0);
4177 }
4178 
4179 static void addrconf_dad_work(struct work_struct *w)
4180 {
4181 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4182 						struct inet6_ifaddr,
4183 						dad_work);
4184 	struct inet6_dev *idev = ifp->idev;
4185 	bool bump_id, disable_ipv6 = false;
4186 	struct in6_addr mcaddr;
4187 
4188 	enum {
4189 		DAD_PROCESS,
4190 		DAD_BEGIN,
4191 		DAD_ABORT,
4192 	} action = DAD_PROCESS;
4193 
4194 	rtnl_lock();
4195 
4196 	spin_lock_bh(&ifp->lock);
4197 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4198 		action = DAD_BEGIN;
4199 		ifp->state = INET6_IFADDR_STATE_DAD;
4200 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4201 		action = DAD_ABORT;
4202 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4203 
4204 		if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->accept_dad) > 1 ||
4205 		     READ_ONCE(idev->cnf.accept_dad) > 1) &&
4206 		    !idev->cnf.disable_ipv6 &&
4207 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4208 			struct in6_addr addr;
4209 
4210 			addr.s6_addr32[0] = htonl(0xfe800000);
4211 			addr.s6_addr32[1] = 0;
4212 
4213 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4214 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4215 				/* DAD failed for link-local based on MAC */
4216 				WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4217 
4218 				pr_info("%s: IPv6 being disabled!\n",
4219 					ifp->idev->dev->name);
4220 				disable_ipv6 = true;
4221 			}
4222 		}
4223 	}
4224 	spin_unlock_bh(&ifp->lock);
4225 
4226 	if (action == DAD_BEGIN) {
4227 		addrconf_dad_begin(ifp);
4228 		goto out;
4229 	} else if (action == DAD_ABORT) {
4230 		in6_ifa_hold(ifp);
4231 		addrconf_dad_stop(ifp, 1);
4232 		if (disable_ipv6)
4233 			addrconf_ifdown(idev->dev, false);
4234 		goto out;
4235 	}
4236 
4237 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4238 		goto out;
4239 
4240 	write_lock_bh(&idev->lock);
4241 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4242 		write_unlock_bh(&idev->lock);
4243 		goto out;
4244 	}
4245 
4246 	spin_lock(&ifp->lock);
4247 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4248 		spin_unlock(&ifp->lock);
4249 		write_unlock_bh(&idev->lock);
4250 		goto out;
4251 	}
4252 
4253 	if (ifp->dad_probes == 0) {
4254 		bool send_na = false;
4255 
4256 		/*
4257 		 * DAD was successful
4258 		 */
4259 
4260 		if (ifp->flags & IFA_F_TENTATIVE &&
4261 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4262 			send_na = true;
4263 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4264 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4265 		spin_unlock(&ifp->lock);
4266 		write_unlock_bh(&idev->lock);
4267 
4268 		addrconf_dad_completed(ifp, bump_id, send_na);
4269 
4270 		goto out;
4271 	}
4272 
4273 	ifp->dad_probes--;
4274 	addrconf_mod_dad_work(ifp,
4275 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4276 				  HZ/100));
4277 	spin_unlock(&ifp->lock);
4278 	write_unlock_bh(&idev->lock);
4279 
4280 	/* send a neighbour solicitation for our addr */
4281 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4282 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4283 		      ifp->dad_nonce);
4284 out:
4285 	in6_ifa_put(ifp);
4286 	rtnl_unlock();
4287 }
4288 
4289 /* ifp->idev must be at least read locked */
4290 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4291 {
4292 	struct inet6_ifaddr *ifpiter;
4293 	struct inet6_dev *idev = ifp->idev;
4294 
4295 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4296 		if (ifpiter->scope > IFA_LINK)
4297 			break;
4298 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4299 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4300 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4301 		    IFA_F_PERMANENT)
4302 			return false;
4303 	}
4304 	return true;
4305 }
4306 
4307 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4308 				   bool send_na)
4309 {
4310 	struct net_device *dev = ifp->idev->dev;
4311 	struct in6_addr lladdr;
4312 	bool send_rs, send_mld;
4313 
4314 	addrconf_del_dad_work(ifp);
4315 
4316 	/*
4317 	 *	Configure the address for reception. Now it is valid.
4318 	 */
4319 
4320 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4321 
4322 	/* If added prefix is link local and we are prepared to process
4323 	   router advertisements, start sending router solicitations.
4324 	 */
4325 
4326 	read_lock_bh(&ifp->idev->lock);
4327 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4328 	send_rs = send_mld &&
4329 		  ipv6_accept_ra(ifp->idev) &&
4330 		  READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4331 		  (dev->flags & IFF_LOOPBACK) == 0 &&
4332 		  (dev->type != ARPHRD_TUNNEL) &&
4333 		  !netif_is_team_port(dev);
4334 	read_unlock_bh(&ifp->idev->lock);
4335 
4336 	/* While dad is in progress mld report's source address is in6_addrany.
4337 	 * Resend with proper ll now.
4338 	 */
4339 	if (send_mld)
4340 		ipv6_mc_dad_complete(ifp->idev);
4341 
4342 	/* send unsolicited NA if enabled */
4343 	if (send_na &&
4344 	    (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4345 	     READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4346 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4347 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4348 			      /*solicited=*/ false, /*override=*/ true,
4349 			      /*inc_opt=*/ true);
4350 	}
4351 
4352 	if (send_rs) {
4353 		/*
4354 		 *	If a host as already performed a random delay
4355 		 *	[...] as part of DAD [...] there is no need
4356 		 *	to delay again before sending the first RS
4357 		 */
4358 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4359 			return;
4360 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4361 
4362 		write_lock_bh(&ifp->idev->lock);
4363 		spin_lock(&ifp->lock);
4364 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4365 			READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4366 		ifp->idev->rs_probes = 1;
4367 		ifp->idev->if_flags |= IF_RS_SENT;
4368 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4369 		spin_unlock(&ifp->lock);
4370 		write_unlock_bh(&ifp->idev->lock);
4371 	}
4372 
4373 	if (bump_id)
4374 		rt_genid_bump_ipv6(dev_net(dev));
4375 
4376 	/* Make sure that a new temporary address will be created
4377 	 * before this temporary address becomes deprecated.
4378 	 */
4379 	if (ifp->flags & IFA_F_TEMPORARY)
4380 		addrconf_verify_rtnl(dev_net(dev));
4381 }
4382 
4383 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4384 {
4385 	struct inet6_ifaddr *ifp;
4386 
4387 	read_lock_bh(&idev->lock);
4388 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4389 		spin_lock(&ifp->lock);
4390 		if ((ifp->flags & IFA_F_TENTATIVE &&
4391 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4392 			if (restart)
4393 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4394 			addrconf_dad_kick(ifp);
4395 		}
4396 		spin_unlock(&ifp->lock);
4397 	}
4398 	read_unlock_bh(&idev->lock);
4399 }
4400 
4401 #ifdef CONFIG_PROC_FS
4402 struct if6_iter_state {
4403 	struct seq_net_private p;
4404 	int bucket;
4405 	int offset;
4406 };
4407 
4408 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4409 {
4410 	struct if6_iter_state *state = seq->private;
4411 	struct net *net = seq_file_net(seq);
4412 	struct inet6_ifaddr *ifa = NULL;
4413 	int p = 0;
4414 
4415 	/* initial bucket if pos is 0 */
4416 	if (pos == 0) {
4417 		state->bucket = 0;
4418 		state->offset = 0;
4419 	}
4420 
4421 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4422 		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4423 					 addr_lst) {
4424 			/* sync with offset */
4425 			if (p < state->offset) {
4426 				p++;
4427 				continue;
4428 			}
4429 			return ifa;
4430 		}
4431 
4432 		/* prepare for next bucket */
4433 		state->offset = 0;
4434 		p = 0;
4435 	}
4436 	return NULL;
4437 }
4438 
4439 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4440 					 struct inet6_ifaddr *ifa)
4441 {
4442 	struct if6_iter_state *state = seq->private;
4443 	struct net *net = seq_file_net(seq);
4444 
4445 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4446 		state->offset++;
4447 		return ifa;
4448 	}
4449 
4450 	state->offset = 0;
4451 	while (++state->bucket < IN6_ADDR_HSIZE) {
4452 		hlist_for_each_entry_rcu(ifa,
4453 				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4454 			return ifa;
4455 		}
4456 	}
4457 
4458 	return NULL;
4459 }
4460 
4461 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4462 	__acquires(rcu)
4463 {
4464 	rcu_read_lock();
4465 	return if6_get_first(seq, *pos);
4466 }
4467 
4468 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4469 {
4470 	struct inet6_ifaddr *ifa;
4471 
4472 	ifa = if6_get_next(seq, v);
4473 	++*pos;
4474 	return ifa;
4475 }
4476 
4477 static void if6_seq_stop(struct seq_file *seq, void *v)
4478 	__releases(rcu)
4479 {
4480 	rcu_read_unlock();
4481 }
4482 
4483 static int if6_seq_show(struct seq_file *seq, void *v)
4484 {
4485 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4486 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4487 		   &ifp->addr,
4488 		   ifp->idev->dev->ifindex,
4489 		   ifp->prefix_len,
4490 		   ifp->scope,
4491 		   (u8) ifp->flags,
4492 		   ifp->idev->dev->name);
4493 	return 0;
4494 }
4495 
4496 static const struct seq_operations if6_seq_ops = {
4497 	.start	= if6_seq_start,
4498 	.next	= if6_seq_next,
4499 	.show	= if6_seq_show,
4500 	.stop	= if6_seq_stop,
4501 };
4502 
4503 static int __net_init if6_proc_net_init(struct net *net)
4504 {
4505 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4506 			sizeof(struct if6_iter_state)))
4507 		return -ENOMEM;
4508 	return 0;
4509 }
4510 
4511 static void __net_exit if6_proc_net_exit(struct net *net)
4512 {
4513 	remove_proc_entry("if_inet6", net->proc_net);
4514 }
4515 
4516 static struct pernet_operations if6_proc_net_ops = {
4517 	.init = if6_proc_net_init,
4518 	.exit = if6_proc_net_exit,
4519 };
4520 
4521 int __init if6_proc_init(void)
4522 {
4523 	return register_pernet_subsys(&if6_proc_net_ops);
4524 }
4525 
4526 void if6_proc_exit(void)
4527 {
4528 	unregister_pernet_subsys(&if6_proc_net_ops);
4529 }
4530 #endif	/* CONFIG_PROC_FS */
4531 
4532 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4533 /* Check if address is a home address configured on any interface. */
4534 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4535 {
4536 	unsigned int hash = inet6_addr_hash(net, addr);
4537 	struct inet6_ifaddr *ifp = NULL;
4538 	int ret = 0;
4539 
4540 	rcu_read_lock();
4541 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4542 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4543 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4544 			ret = 1;
4545 			break;
4546 		}
4547 	}
4548 	rcu_read_unlock();
4549 	return ret;
4550 }
4551 #endif
4552 
4553 /* RFC6554 has some algorithm to avoid loops in segment routing by
4554  * checking if the segments contains any of a local interface address.
4555  *
4556  * Quote:
4557  *
4558  * To detect loops in the SRH, a router MUST determine if the SRH
4559  * includes multiple addresses assigned to any interface on that router.
4560  * If such addresses appear more than once and are separated by at least
4561  * one address not assigned to that router.
4562  */
4563 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4564 			  unsigned char nsegs)
4565 {
4566 	const struct in6_addr *addr;
4567 	int i, ret = 0, found = 0;
4568 	struct inet6_ifaddr *ifp;
4569 	bool separated = false;
4570 	unsigned int hash;
4571 	bool hash_found;
4572 
4573 	rcu_read_lock();
4574 	for (i = 0; i < nsegs; i++) {
4575 		addr = &segs[i];
4576 		hash = inet6_addr_hash(net, addr);
4577 
4578 		hash_found = false;
4579 		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4580 
4581 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4582 				hash_found = true;
4583 				break;
4584 			}
4585 		}
4586 
4587 		if (hash_found) {
4588 			if (found > 1 && separated) {
4589 				ret = 1;
4590 				break;
4591 			}
4592 
4593 			separated = false;
4594 			found++;
4595 		} else {
4596 			separated = true;
4597 		}
4598 	}
4599 	rcu_read_unlock();
4600 
4601 	return ret;
4602 }
4603 
4604 /*
4605  *	Periodic address status verification
4606  */
4607 
4608 static void addrconf_verify_rtnl(struct net *net)
4609 {
4610 	unsigned long now, next, next_sec, next_sched;
4611 	struct inet6_ifaddr *ifp;
4612 	int i;
4613 
4614 	ASSERT_RTNL();
4615 
4616 	rcu_read_lock_bh();
4617 	now = jiffies;
4618 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4619 
4620 	cancel_delayed_work(&net->ipv6.addr_chk_work);
4621 
4622 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4623 restart:
4624 		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4625 			unsigned long age;
4626 
4627 			/* When setting preferred_lft to a value not zero or
4628 			 * infinity, while valid_lft is infinity
4629 			 * IFA_F_PERMANENT has a non-infinity life time.
4630 			 */
4631 			if ((ifp->flags & IFA_F_PERMANENT) &&
4632 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4633 				continue;
4634 
4635 			spin_lock(&ifp->lock);
4636 			/* We try to batch several events at once. */
4637 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4638 
4639 			if ((ifp->flags&IFA_F_TEMPORARY) &&
4640 			    !(ifp->flags&IFA_F_TENTATIVE) &&
4641 			    ifp->prefered_lft != INFINITY_LIFE_TIME &&
4642 			    !ifp->regen_count && ifp->ifpub) {
4643 				/* This is a non-regenerated temporary addr. */
4644 
4645 				unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4646 
4647 				if (age + regen_advance >= ifp->prefered_lft) {
4648 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4649 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4650 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4651 
4652 					ifp->regen_count++;
4653 					in6_ifa_hold(ifp);
4654 					in6_ifa_hold(ifpub);
4655 					spin_unlock(&ifp->lock);
4656 
4657 					spin_lock(&ifpub->lock);
4658 					ifpub->regen_count = 0;
4659 					spin_unlock(&ifpub->lock);
4660 					rcu_read_unlock_bh();
4661 					ipv6_create_tempaddr(ifpub, true);
4662 					in6_ifa_put(ifpub);
4663 					in6_ifa_put(ifp);
4664 					rcu_read_lock_bh();
4665 					goto restart;
4666 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4667 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4668 			}
4669 
4670 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4671 			    age >= ifp->valid_lft) {
4672 				spin_unlock(&ifp->lock);
4673 				in6_ifa_hold(ifp);
4674 				rcu_read_unlock_bh();
4675 				ipv6_del_addr(ifp);
4676 				rcu_read_lock_bh();
4677 				goto restart;
4678 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4679 				spin_unlock(&ifp->lock);
4680 				continue;
4681 			} else if (age >= ifp->prefered_lft) {
4682 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4683 				int deprecate = 0;
4684 
4685 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4686 					deprecate = 1;
4687 					ifp->flags |= IFA_F_DEPRECATED;
4688 				}
4689 
4690 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4691 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4692 					next = ifp->tstamp + ifp->valid_lft * HZ;
4693 
4694 				spin_unlock(&ifp->lock);
4695 
4696 				if (deprecate) {
4697 					in6_ifa_hold(ifp);
4698 
4699 					ipv6_ifa_notify(0, ifp);
4700 					in6_ifa_put(ifp);
4701 					goto restart;
4702 				}
4703 			} else {
4704 				/* ifp->prefered_lft <= ifp->valid_lft */
4705 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4706 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4707 				spin_unlock(&ifp->lock);
4708 			}
4709 		}
4710 	}
4711 
4712 	next_sec = round_jiffies_up(next);
4713 	next_sched = next;
4714 
4715 	/* If rounded timeout is accurate enough, accept it. */
4716 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4717 		next_sched = next_sec;
4718 
4719 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4720 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4721 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4722 
4723 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4724 		 now, next, next_sec, next_sched);
4725 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4726 	rcu_read_unlock_bh();
4727 }
4728 
4729 static void addrconf_verify_work(struct work_struct *w)
4730 {
4731 	struct net *net = container_of(to_delayed_work(w), struct net,
4732 				       ipv6.addr_chk_work);
4733 
4734 	rtnl_lock();
4735 	addrconf_verify_rtnl(net);
4736 	rtnl_unlock();
4737 }
4738 
4739 static void addrconf_verify(struct net *net)
4740 {
4741 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4742 }
4743 
4744 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4745 				     struct in6_addr **peer_pfx)
4746 {
4747 	struct in6_addr *pfx = NULL;
4748 
4749 	*peer_pfx = NULL;
4750 
4751 	if (addr)
4752 		pfx = nla_data(addr);
4753 
4754 	if (local) {
4755 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4756 			*peer_pfx = pfx;
4757 		pfx = nla_data(local);
4758 	}
4759 
4760 	return pfx;
4761 }
4762 
4763 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4764 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4765 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4766 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4767 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4768 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4769 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4770 	[IFA_PROTO]		= { .type = NLA_U8 },
4771 };
4772 
4773 static int
4774 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4775 		  struct netlink_ext_ack *extack)
4776 {
4777 	struct net *net = sock_net(skb->sk);
4778 	struct ifaddrmsg *ifm;
4779 	struct nlattr *tb[IFA_MAX+1];
4780 	struct in6_addr *pfx, *peer_pfx;
4781 	u32 ifa_flags;
4782 	int err;
4783 
4784 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4785 				     ifa_ipv6_policy, extack);
4786 	if (err < 0)
4787 		return err;
4788 
4789 	ifm = nlmsg_data(nlh);
4790 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4791 	if (!pfx)
4792 		return -EINVAL;
4793 
4794 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4795 
4796 	/* We ignore other flags so far. */
4797 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4798 
4799 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4800 			      ifm->ifa_prefixlen, extack);
4801 }
4802 
4803 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4804 			       unsigned long expires, u32 flags,
4805 			       bool modify_peer)
4806 {
4807 	struct fib6_table *table;
4808 	struct fib6_info *f6i;
4809 	u32 prio;
4810 
4811 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4812 					ifp->prefix_len,
4813 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4814 	if (!f6i)
4815 		return -ENOENT;
4816 
4817 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4818 	if (f6i->fib6_metric != prio) {
4819 		/* delete old one */
4820 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4821 
4822 		/* add new one */
4823 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4824 				      ifp->prefix_len,
4825 				      ifp->rt_priority, ifp->idev->dev,
4826 				      expires, flags, GFP_KERNEL);
4827 	} else {
4828 		table = f6i->fib6_table;
4829 		spin_lock_bh(&table->tb6_lock);
4830 
4831 		if (!(flags & RTF_EXPIRES)) {
4832 			fib6_clean_expires(f6i);
4833 			fib6_remove_gc_list(f6i);
4834 		} else {
4835 			fib6_set_expires(f6i, expires);
4836 			fib6_add_gc_list(f6i);
4837 		}
4838 
4839 		spin_unlock_bh(&table->tb6_lock);
4840 
4841 		fib6_info_release(f6i);
4842 	}
4843 
4844 	return 0;
4845 }
4846 
4847 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4848 			     struct ifa6_config *cfg)
4849 {
4850 	u32 flags;
4851 	clock_t expires;
4852 	unsigned long timeout;
4853 	bool was_managetempaddr;
4854 	bool had_prefixroute;
4855 	bool new_peer = false;
4856 
4857 	ASSERT_RTNL();
4858 
4859 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4860 		return -EINVAL;
4861 
4862 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4863 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4864 		return -EINVAL;
4865 
4866 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4867 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4868 
4869 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4870 	if (addrconf_finite_timeout(timeout)) {
4871 		expires = jiffies_to_clock_t(timeout * HZ);
4872 		cfg->valid_lft = timeout;
4873 		flags = RTF_EXPIRES;
4874 	} else {
4875 		expires = 0;
4876 		flags = 0;
4877 		cfg->ifa_flags |= IFA_F_PERMANENT;
4878 	}
4879 
4880 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4881 	if (addrconf_finite_timeout(timeout)) {
4882 		if (timeout == 0)
4883 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4884 		cfg->preferred_lft = timeout;
4885 	}
4886 
4887 	if (cfg->peer_pfx &&
4888 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4889 		if (!ipv6_addr_any(&ifp->peer_addr))
4890 			cleanup_prefix_route(ifp, expires, true, true);
4891 		new_peer = true;
4892 	}
4893 
4894 	spin_lock_bh(&ifp->lock);
4895 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4896 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4897 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4898 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4899 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4900 			IFA_F_NOPREFIXROUTE);
4901 	ifp->flags |= cfg->ifa_flags;
4902 	WRITE_ONCE(ifp->tstamp, jiffies);
4903 	WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4904 	WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4905 	WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4906 
4907 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4908 		WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4909 
4910 	if (new_peer)
4911 		ifp->peer_addr = *cfg->peer_pfx;
4912 
4913 	spin_unlock_bh(&ifp->lock);
4914 	if (!(ifp->flags&IFA_F_TENTATIVE))
4915 		ipv6_ifa_notify(0, ifp);
4916 
4917 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4918 		int rc = -ENOENT;
4919 
4920 		if (had_prefixroute)
4921 			rc = modify_prefix_route(ifp, expires, flags, false);
4922 
4923 		/* prefix route could have been deleted; if so restore it */
4924 		if (rc == -ENOENT) {
4925 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4926 					      ifp->rt_priority, ifp->idev->dev,
4927 					      expires, flags, GFP_KERNEL);
4928 		}
4929 
4930 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4931 			rc = modify_prefix_route(ifp, expires, flags, true);
4932 
4933 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4934 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4935 					      ifp->rt_priority, ifp->idev->dev,
4936 					      expires, flags, GFP_KERNEL);
4937 		}
4938 	} else if (had_prefixroute) {
4939 		enum cleanup_prefix_rt_t action;
4940 		unsigned long rt_expires;
4941 
4942 		write_lock_bh(&ifp->idev->lock);
4943 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4944 		write_unlock_bh(&ifp->idev->lock);
4945 
4946 		if (action != CLEANUP_PREFIX_RT_NOP) {
4947 			cleanup_prefix_route(ifp, rt_expires,
4948 				action == CLEANUP_PREFIX_RT_DEL, false);
4949 		}
4950 	}
4951 
4952 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4953 		if (was_managetempaddr &&
4954 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4955 			cfg->valid_lft = 0;
4956 			cfg->preferred_lft = 0;
4957 		}
4958 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4959 				 cfg->preferred_lft, !was_managetempaddr,
4960 				 jiffies);
4961 	}
4962 
4963 	addrconf_verify_rtnl(net);
4964 
4965 	return 0;
4966 }
4967 
4968 static int
4969 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4970 		  struct netlink_ext_ack *extack)
4971 {
4972 	struct net *net = sock_net(skb->sk);
4973 	struct ifaddrmsg *ifm;
4974 	struct nlattr *tb[IFA_MAX+1];
4975 	struct in6_addr *peer_pfx;
4976 	struct inet6_ifaddr *ifa;
4977 	struct net_device *dev;
4978 	struct inet6_dev *idev;
4979 	struct ifa6_config cfg;
4980 	int err;
4981 
4982 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4983 				     ifa_ipv6_policy, extack);
4984 	if (err < 0)
4985 		return err;
4986 
4987 	memset(&cfg, 0, sizeof(cfg));
4988 
4989 	ifm = nlmsg_data(nlh);
4990 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4991 	if (!cfg.pfx)
4992 		return -EINVAL;
4993 
4994 	cfg.peer_pfx = peer_pfx;
4995 	cfg.plen = ifm->ifa_prefixlen;
4996 	if (tb[IFA_RT_PRIORITY])
4997 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4998 
4999 	if (tb[IFA_PROTO])
5000 		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
5001 
5002 	cfg.valid_lft = INFINITY_LIFE_TIME;
5003 	cfg.preferred_lft = INFINITY_LIFE_TIME;
5004 
5005 	if (tb[IFA_CACHEINFO]) {
5006 		struct ifa_cacheinfo *ci;
5007 
5008 		ci = nla_data(tb[IFA_CACHEINFO]);
5009 		cfg.valid_lft = ci->ifa_valid;
5010 		cfg.preferred_lft = ci->ifa_prefered;
5011 	}
5012 
5013 	dev =  __dev_get_by_index(net, ifm->ifa_index);
5014 	if (!dev) {
5015 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5016 		return -ENODEV;
5017 	}
5018 
5019 	if (tb[IFA_FLAGS])
5020 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
5021 	else
5022 		cfg.ifa_flags = ifm->ifa_flags;
5023 
5024 	/* We ignore other flags so far. */
5025 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
5026 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
5027 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
5028 
5029 	idev = ipv6_find_idev(dev);
5030 	if (IS_ERR(idev))
5031 		return PTR_ERR(idev);
5032 
5033 	if (!ipv6_allow_optimistic_dad(net, idev))
5034 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5035 
5036 	if (cfg.ifa_flags & IFA_F_NODAD &&
5037 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5038 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5039 		return -EINVAL;
5040 	}
5041 
5042 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5043 	if (!ifa) {
5044 		/*
5045 		 * It would be best to check for !NLM_F_CREATE here but
5046 		 * userspace already relies on not having to provide this.
5047 		 */
5048 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
5049 	}
5050 
5051 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
5052 	    !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5053 		NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5054 		err = -EEXIST;
5055 	} else {
5056 		err = inet6_addr_modify(net, ifa, &cfg);
5057 	}
5058 
5059 	in6_ifa_put(ifa);
5060 
5061 	return err;
5062 }
5063 
5064 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5065 			  u8 scope, int ifindex)
5066 {
5067 	struct ifaddrmsg *ifm;
5068 
5069 	ifm = nlmsg_data(nlh);
5070 	ifm->ifa_family = AF_INET6;
5071 	ifm->ifa_prefixlen = prefixlen;
5072 	ifm->ifa_flags = flags;
5073 	ifm->ifa_scope = scope;
5074 	ifm->ifa_index = ifindex;
5075 }
5076 
5077 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5078 			 unsigned long tstamp, u32 preferred, u32 valid)
5079 {
5080 	struct ifa_cacheinfo ci;
5081 
5082 	ci.cstamp = cstamp_delta(cstamp);
5083 	ci.tstamp = cstamp_delta(tstamp);
5084 	ci.ifa_prefered = preferred;
5085 	ci.ifa_valid = valid;
5086 
5087 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5088 }
5089 
5090 static inline int rt_scope(int ifa_scope)
5091 {
5092 	if (ifa_scope & IFA_HOST)
5093 		return RT_SCOPE_HOST;
5094 	else if (ifa_scope & IFA_LINK)
5095 		return RT_SCOPE_LINK;
5096 	else if (ifa_scope & IFA_SITE)
5097 		return RT_SCOPE_SITE;
5098 	else
5099 		return RT_SCOPE_UNIVERSE;
5100 }
5101 
5102 static inline int inet6_ifaddr_msgsize(void)
5103 {
5104 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5105 	       + nla_total_size(16) /* IFA_LOCAL */
5106 	       + nla_total_size(16) /* IFA_ADDRESS */
5107 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
5108 	       + nla_total_size(4)  /* IFA_FLAGS */
5109 	       + nla_total_size(1)  /* IFA_PROTO */
5110 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5111 }
5112 
5113 enum addr_type_t {
5114 	UNICAST_ADDR,
5115 	MULTICAST_ADDR,
5116 	ANYCAST_ADDR,
5117 };
5118 
5119 struct inet6_fill_args {
5120 	u32 portid;
5121 	u32 seq;
5122 	int event;
5123 	unsigned int flags;
5124 	int netnsid;
5125 	int ifindex;
5126 	enum addr_type_t type;
5127 };
5128 
5129 static int inet6_fill_ifaddr(struct sk_buff *skb,
5130 			     const struct inet6_ifaddr *ifa,
5131 			     struct inet6_fill_args *args)
5132 {
5133 	struct nlmsghdr *nlh;
5134 	u32 preferred, valid;
5135 	u32 flags, priority;
5136 	u8 proto;
5137 
5138 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5139 			sizeof(struct ifaddrmsg), args->flags);
5140 	if (!nlh)
5141 		return -EMSGSIZE;
5142 
5143 	flags = READ_ONCE(ifa->flags);
5144 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5145 		      ifa->idev->dev->ifindex);
5146 
5147 	if (args->netnsid >= 0 &&
5148 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5149 		goto error;
5150 
5151 	preferred = READ_ONCE(ifa->prefered_lft);
5152 	valid = READ_ONCE(ifa->valid_lft);
5153 
5154 	if (!((flags & IFA_F_PERMANENT) &&
5155 	      (preferred == INFINITY_LIFE_TIME))) {
5156 		if (preferred != INFINITY_LIFE_TIME) {
5157 			long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5158 
5159 			if (preferred > tval)
5160 				preferred -= tval;
5161 			else
5162 				preferred = 0;
5163 			if (valid != INFINITY_LIFE_TIME) {
5164 				if (valid > tval)
5165 					valid -= tval;
5166 				else
5167 					valid = 0;
5168 			}
5169 		}
5170 	} else {
5171 		preferred = INFINITY_LIFE_TIME;
5172 		valid = INFINITY_LIFE_TIME;
5173 	}
5174 
5175 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5176 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5177 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5178 			goto error;
5179 	} else {
5180 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5181 			goto error;
5182 	}
5183 
5184 	priority = READ_ONCE(ifa->rt_priority);
5185 	if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5186 		goto error;
5187 
5188 	if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5189 			  preferred, valid) < 0)
5190 		goto error;
5191 
5192 	if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5193 		goto error;
5194 
5195 	proto = READ_ONCE(ifa->ifa_proto);
5196 	if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5197 		goto error;
5198 
5199 	nlmsg_end(skb, nlh);
5200 	return 0;
5201 
5202 error:
5203 	nlmsg_cancel(skb, nlh);
5204 	return -EMSGSIZE;
5205 }
5206 
5207 static int inet6_fill_ifmcaddr(struct sk_buff *skb,
5208 			       const struct ifmcaddr6 *ifmca,
5209 			       struct inet6_fill_args *args)
5210 {
5211 	int ifindex = ifmca->idev->dev->ifindex;
5212 	u8 scope = RT_SCOPE_UNIVERSE;
5213 	struct nlmsghdr *nlh;
5214 
5215 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5216 		scope = RT_SCOPE_SITE;
5217 
5218 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5219 			sizeof(struct ifaddrmsg), args->flags);
5220 	if (!nlh)
5221 		return -EMSGSIZE;
5222 
5223 	if (args->netnsid >= 0 &&
5224 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5225 		nlmsg_cancel(skb, nlh);
5226 		return -EMSGSIZE;
5227 	}
5228 
5229 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5230 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5231 	    put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5232 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5233 		nlmsg_cancel(skb, nlh);
5234 		return -EMSGSIZE;
5235 	}
5236 
5237 	nlmsg_end(skb, nlh);
5238 	return 0;
5239 }
5240 
5241 static int inet6_fill_ifacaddr(struct sk_buff *skb,
5242 			       const struct ifacaddr6 *ifaca,
5243 			       struct inet6_fill_args *args)
5244 {
5245 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5246 	int ifindex = dev ? dev->ifindex : 1;
5247 	u8 scope = RT_SCOPE_UNIVERSE;
5248 	struct nlmsghdr *nlh;
5249 
5250 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5251 		scope = RT_SCOPE_SITE;
5252 
5253 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5254 			sizeof(struct ifaddrmsg), args->flags);
5255 	if (!nlh)
5256 		return -EMSGSIZE;
5257 
5258 	if (args->netnsid >= 0 &&
5259 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5260 		nlmsg_cancel(skb, nlh);
5261 		return -EMSGSIZE;
5262 	}
5263 
5264 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5265 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5266 	    put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5267 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5268 		nlmsg_cancel(skb, nlh);
5269 		return -EMSGSIZE;
5270 	}
5271 
5272 	nlmsg_end(skb, nlh);
5273 	return 0;
5274 }
5275 
5276 /* called with rcu_read_lock() */
5277 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5278 			  struct netlink_callback *cb, int *s_ip_idx,
5279 			  struct inet6_fill_args *fillargs)
5280 {
5281 	const struct ifmcaddr6 *ifmca;
5282 	const struct ifacaddr6 *ifaca;
5283 	int ip_idx = 0;
5284 	int err = 0;
5285 
5286 	switch (fillargs->type) {
5287 	case UNICAST_ADDR: {
5288 		const struct inet6_ifaddr *ifa;
5289 		fillargs->event = RTM_NEWADDR;
5290 
5291 		/* unicast address incl. temp addr */
5292 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5293 			if (ip_idx < *s_ip_idx)
5294 				goto next;
5295 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5296 			if (err < 0)
5297 				break;
5298 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5299 next:
5300 			ip_idx++;
5301 		}
5302 		break;
5303 	}
5304 	case MULTICAST_ADDR:
5305 		fillargs->event = RTM_GETMULTICAST;
5306 
5307 		/* multicast address */
5308 		for (ifmca = rcu_dereference(idev->mc_list);
5309 		     ifmca;
5310 		     ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5311 			if (ip_idx < *s_ip_idx)
5312 				continue;
5313 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5314 			if (err < 0)
5315 				break;
5316 		}
5317 		break;
5318 	case ANYCAST_ADDR:
5319 		fillargs->event = RTM_GETANYCAST;
5320 		/* anycast address */
5321 		for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5322 		     ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5323 			if (ip_idx < *s_ip_idx)
5324 				continue;
5325 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5326 			if (err < 0)
5327 				break;
5328 		}
5329 		break;
5330 	default:
5331 		break;
5332 	}
5333 	*s_ip_idx = err ? ip_idx : 0;
5334 	return err;
5335 }
5336 
5337 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5338 				       struct inet6_fill_args *fillargs,
5339 				       struct net **tgt_net, struct sock *sk,
5340 				       struct netlink_callback *cb)
5341 {
5342 	struct netlink_ext_ack *extack = cb->extack;
5343 	struct nlattr *tb[IFA_MAX+1];
5344 	struct ifaddrmsg *ifm;
5345 	int err, i;
5346 
5347 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5348 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5349 		return -EINVAL;
5350 	}
5351 
5352 	ifm = nlmsg_data(nlh);
5353 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5354 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5355 		return -EINVAL;
5356 	}
5357 
5358 	fillargs->ifindex = ifm->ifa_index;
5359 	if (fillargs->ifindex) {
5360 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5361 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5362 	}
5363 
5364 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5365 					    ifa_ipv6_policy, extack);
5366 	if (err < 0)
5367 		return err;
5368 
5369 	for (i = 0; i <= IFA_MAX; ++i) {
5370 		if (!tb[i])
5371 			continue;
5372 
5373 		if (i == IFA_TARGET_NETNSID) {
5374 			struct net *net;
5375 
5376 			fillargs->netnsid = nla_get_s32(tb[i]);
5377 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5378 			if (IS_ERR(net)) {
5379 				fillargs->netnsid = -1;
5380 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5381 				return PTR_ERR(net);
5382 			}
5383 			*tgt_net = net;
5384 		} else {
5385 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5386 			return -EINVAL;
5387 		}
5388 	}
5389 
5390 	return 0;
5391 }
5392 
5393 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5394 			   enum addr_type_t type)
5395 {
5396 	struct net *tgt_net = sock_net(skb->sk);
5397 	const struct nlmsghdr *nlh = cb->nlh;
5398 	struct inet6_fill_args fillargs = {
5399 		.portid = NETLINK_CB(cb->skb).portid,
5400 		.seq = cb->nlh->nlmsg_seq,
5401 		.flags = NLM_F_MULTI,
5402 		.netnsid = -1,
5403 		.type = type,
5404 	};
5405 	struct {
5406 		unsigned long ifindex;
5407 		int ip_idx;
5408 	} *ctx = (void *)cb->ctx;
5409 	struct net_device *dev;
5410 	struct inet6_dev *idev;
5411 	int err = 0;
5412 
5413 	rcu_read_lock();
5414 	if (cb->strict_check) {
5415 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5416 						  skb->sk, cb);
5417 		if (err < 0)
5418 			goto done;
5419 
5420 		err = 0;
5421 		if (fillargs.ifindex) {
5422 			dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5423 			if (!dev) {
5424 				err = -ENODEV;
5425 				goto done;
5426 			}
5427 			idev = __in6_dev_get(dev);
5428 			if (idev)
5429 				err = in6_dump_addrs(idev, skb, cb,
5430 						     &ctx->ip_idx,
5431 						     &fillargs);
5432 			goto done;
5433 		}
5434 	}
5435 
5436 	cb->seq = inet6_base_seq(tgt_net);
5437 	for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5438 		idev = __in6_dev_get(dev);
5439 		if (!idev)
5440 			continue;
5441 		err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5442 				     &fillargs);
5443 		if (err < 0)
5444 			goto done;
5445 	}
5446 done:
5447 	rcu_read_unlock();
5448 	if (fillargs.netnsid >= 0)
5449 		put_net(tgt_net);
5450 
5451 	return err;
5452 }
5453 
5454 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5455 {
5456 	enum addr_type_t type = UNICAST_ADDR;
5457 
5458 	return inet6_dump_addr(skb, cb, type);
5459 }
5460 
5461 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5462 {
5463 	enum addr_type_t type = MULTICAST_ADDR;
5464 
5465 	return inet6_dump_addr(skb, cb, type);
5466 }
5467 
5468 
5469 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5470 {
5471 	enum addr_type_t type = ANYCAST_ADDR;
5472 
5473 	return inet6_dump_addr(skb, cb, type);
5474 }
5475 
5476 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5477 				       const struct nlmsghdr *nlh,
5478 				       struct nlattr **tb,
5479 				       struct netlink_ext_ack *extack)
5480 {
5481 	struct ifaddrmsg *ifm;
5482 	int i, err;
5483 
5484 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5485 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5486 		return -EINVAL;
5487 	}
5488 
5489 	if (!netlink_strict_get_check(skb))
5490 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5491 					      ifa_ipv6_policy, extack);
5492 
5493 	ifm = nlmsg_data(nlh);
5494 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5495 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5496 		return -EINVAL;
5497 	}
5498 
5499 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5500 					    ifa_ipv6_policy, extack);
5501 	if (err)
5502 		return err;
5503 
5504 	for (i = 0; i <= IFA_MAX; i++) {
5505 		if (!tb[i])
5506 			continue;
5507 
5508 		switch (i) {
5509 		case IFA_TARGET_NETNSID:
5510 		case IFA_ADDRESS:
5511 		case IFA_LOCAL:
5512 			break;
5513 		default:
5514 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5515 			return -EINVAL;
5516 		}
5517 	}
5518 
5519 	return 0;
5520 }
5521 
5522 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5523 			     struct netlink_ext_ack *extack)
5524 {
5525 	struct net *tgt_net = sock_net(in_skb->sk);
5526 	struct inet6_fill_args fillargs = {
5527 		.portid = NETLINK_CB(in_skb).portid,
5528 		.seq = nlh->nlmsg_seq,
5529 		.event = RTM_NEWADDR,
5530 		.flags = 0,
5531 		.netnsid = -1,
5532 	};
5533 	struct ifaddrmsg *ifm;
5534 	struct nlattr *tb[IFA_MAX+1];
5535 	struct in6_addr *addr = NULL, *peer;
5536 	struct net_device *dev = NULL;
5537 	struct inet6_ifaddr *ifa;
5538 	struct sk_buff *skb;
5539 	int err;
5540 
5541 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5542 	if (err < 0)
5543 		return err;
5544 
5545 	if (tb[IFA_TARGET_NETNSID]) {
5546 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5547 
5548 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5549 						  fillargs.netnsid);
5550 		if (IS_ERR(tgt_net))
5551 			return PTR_ERR(tgt_net);
5552 	}
5553 
5554 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5555 	if (!addr) {
5556 		err = -EINVAL;
5557 		goto errout;
5558 	}
5559 	ifm = nlmsg_data(nlh);
5560 	if (ifm->ifa_index)
5561 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5562 
5563 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5564 	if (!ifa) {
5565 		err = -EADDRNOTAVAIL;
5566 		goto errout;
5567 	}
5568 
5569 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5570 	if (!skb) {
5571 		err = -ENOBUFS;
5572 		goto errout_ifa;
5573 	}
5574 
5575 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5576 	if (err < 0) {
5577 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5578 		WARN_ON(err == -EMSGSIZE);
5579 		kfree_skb(skb);
5580 		goto errout_ifa;
5581 	}
5582 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5583 errout_ifa:
5584 	in6_ifa_put(ifa);
5585 errout:
5586 	dev_put(dev);
5587 	if (fillargs.netnsid >= 0)
5588 		put_net(tgt_net);
5589 
5590 	return err;
5591 }
5592 
5593 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5594 {
5595 	struct sk_buff *skb;
5596 	struct net *net = dev_net(ifa->idev->dev);
5597 	struct inet6_fill_args fillargs = {
5598 		.portid = 0,
5599 		.seq = 0,
5600 		.event = event,
5601 		.flags = 0,
5602 		.netnsid = -1,
5603 	};
5604 	int err = -ENOBUFS;
5605 
5606 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5607 	if (!skb)
5608 		goto errout;
5609 
5610 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5611 	if (err < 0) {
5612 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5613 		WARN_ON(err == -EMSGSIZE);
5614 		kfree_skb(skb);
5615 		goto errout;
5616 	}
5617 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5618 	return;
5619 errout:
5620 	if (err < 0)
5621 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5622 }
5623 
5624 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5625 			       __s32 *array, int bytes)
5626 {
5627 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5628 
5629 	memset(array, 0, bytes);
5630 	array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5631 	array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5632 	array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5633 	array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5634 	array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5635 	array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5636 	array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5637 	array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5638 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5639 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5640 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5641 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5642 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5643 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5644 	array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5645 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5646 		jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5647 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5648 		jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5649 	array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5650 	array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5651 	array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5652 	array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5653 	array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5654 	array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5655 	array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5656 	array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5657 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5658 		READ_ONCE(cnf->accept_ra_min_hop_limit);
5659 	array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5660 #ifdef CONFIG_IPV6_ROUTER_PREF
5661 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5662 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5663 		jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5664 #ifdef CONFIG_IPV6_ROUTE_INFO
5665 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5666 		READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5667 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5668 		READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5669 #endif
5670 #endif
5671 	array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5672 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5673 		READ_ONCE(cnf->accept_source_route);
5674 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5675 	array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5676 	array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5677 #endif
5678 #ifdef CONFIG_IPV6_MROUTE
5679 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5680 #endif
5681 	array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5682 	array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5683 	array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5684 	array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5685 	array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5686 		READ_ONCE(cnf->suppress_frag_ndisc);
5687 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5688 		READ_ONCE(cnf->accept_ra_from_local);
5689 	array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5690 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5691 		READ_ONCE(cnf->ignore_routes_with_linkdown);
5692 	/* we omit DEVCONF_STABLE_SECRET for now */
5693 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5694 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5695 		READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5696 	array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5697 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5698 	array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5699 #ifdef CONFIG_IPV6_SEG6_HMAC
5700 	array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5701 #endif
5702 	array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5703 	array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5704 	array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5705 	array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5706 	array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5707 	array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5708 	array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5709 	array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5710 	array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5711 		READ_ONCE(cnf->ndisc_evict_nocarrier);
5712 	array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5713 		READ_ONCE(cnf->accept_untracked_na);
5714 	array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5715 }
5716 
5717 static inline size_t inet6_ifla6_size(void)
5718 {
5719 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5720 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5721 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5722 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5723 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5724 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5725 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5726 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5727 	     + 0;
5728 }
5729 
5730 static inline size_t inet6_if_nlmsg_size(void)
5731 {
5732 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5733 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5734 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5735 	       + nla_total_size(4) /* IFLA_MTU */
5736 	       + nla_total_size(4) /* IFLA_LINK */
5737 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5738 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5739 }
5740 
5741 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5742 					int bytes)
5743 {
5744 	int i;
5745 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5746 	BUG_ON(pad < 0);
5747 
5748 	/* Use put_unaligned() because stats may not be aligned for u64. */
5749 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5750 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5751 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5752 
5753 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5754 }
5755 
5756 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5757 					int bytes, size_t syncpoff)
5758 {
5759 	int i, c;
5760 	u64 buff[IPSTATS_MIB_MAX];
5761 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5762 
5763 	BUG_ON(pad < 0);
5764 
5765 	memset(buff, 0, sizeof(buff));
5766 	buff[0] = IPSTATS_MIB_MAX;
5767 
5768 	for_each_possible_cpu(c) {
5769 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5770 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5771 	}
5772 
5773 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5774 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5775 }
5776 
5777 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5778 			     int bytes)
5779 {
5780 	switch (attrtype) {
5781 	case IFLA_INET6_STATS:
5782 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5783 				     offsetof(struct ipstats_mib, syncp));
5784 		break;
5785 	case IFLA_INET6_ICMP6STATS:
5786 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5787 		break;
5788 	}
5789 }
5790 
5791 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5792 				  u32 ext_filter_mask)
5793 {
5794 	struct ifla_cacheinfo ci;
5795 	struct nlattr *nla;
5796 	u32 ra_mtu;
5797 
5798 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5799 		goto nla_put_failure;
5800 	ci.max_reasm_len = IPV6_MAXPLEN;
5801 	ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5802 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5803 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5804 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5805 		goto nla_put_failure;
5806 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5807 	if (!nla)
5808 		goto nla_put_failure;
5809 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5810 
5811 	/* XXX - MC not implemented */
5812 
5813 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5814 		return 0;
5815 
5816 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5817 	if (!nla)
5818 		goto nla_put_failure;
5819 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5820 
5821 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5822 	if (!nla)
5823 		goto nla_put_failure;
5824 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5825 
5826 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5827 	if (!nla)
5828 		goto nla_put_failure;
5829 	read_lock_bh(&idev->lock);
5830 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5831 	read_unlock_bh(&idev->lock);
5832 
5833 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5834 		       READ_ONCE(idev->cnf.addr_gen_mode)))
5835 		goto nla_put_failure;
5836 
5837 	ra_mtu = READ_ONCE(idev->ra_mtu);
5838 	if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5839 		goto nla_put_failure;
5840 
5841 	return 0;
5842 
5843 nla_put_failure:
5844 	return -EMSGSIZE;
5845 }
5846 
5847 static size_t inet6_get_link_af_size(const struct net_device *dev,
5848 				     u32 ext_filter_mask)
5849 {
5850 	if (!__in6_dev_get(dev))
5851 		return 0;
5852 
5853 	return inet6_ifla6_size();
5854 }
5855 
5856 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5857 			      u32 ext_filter_mask)
5858 {
5859 	struct inet6_dev *idev = __in6_dev_get(dev);
5860 
5861 	if (!idev)
5862 		return -ENODATA;
5863 
5864 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5865 		return -EMSGSIZE;
5866 
5867 	return 0;
5868 }
5869 
5870 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5871 			     struct netlink_ext_ack *extack)
5872 {
5873 	struct inet6_ifaddr *ifp;
5874 	struct net_device *dev = idev->dev;
5875 	bool clear_token, update_rs = false;
5876 	struct in6_addr ll_addr;
5877 
5878 	ASSERT_RTNL();
5879 
5880 	if (!token)
5881 		return -EINVAL;
5882 
5883 	if (dev->flags & IFF_LOOPBACK) {
5884 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5885 		return -EINVAL;
5886 	}
5887 
5888 	if (dev->flags & IFF_NOARP) {
5889 		NL_SET_ERR_MSG_MOD(extack,
5890 				   "Device does not do neighbour discovery");
5891 		return -EINVAL;
5892 	}
5893 
5894 	if (!ipv6_accept_ra(idev)) {
5895 		NL_SET_ERR_MSG_MOD(extack,
5896 				   "Router advertisement is disabled on device");
5897 		return -EINVAL;
5898 	}
5899 
5900 	if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5901 		NL_SET_ERR_MSG(extack,
5902 			       "Router solicitation is disabled on device");
5903 		return -EINVAL;
5904 	}
5905 
5906 	write_lock_bh(&idev->lock);
5907 
5908 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5909 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5910 
5911 	write_unlock_bh(&idev->lock);
5912 
5913 	clear_token = ipv6_addr_any(token);
5914 	if (clear_token)
5915 		goto update_lft;
5916 
5917 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5918 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5919 			     IFA_F_OPTIMISTIC)) {
5920 		/* If we're not ready, then normal ifup will take care
5921 		 * of this. Otherwise, we need to request our rs here.
5922 		 */
5923 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5924 		update_rs = true;
5925 	}
5926 
5927 update_lft:
5928 	write_lock_bh(&idev->lock);
5929 
5930 	if (update_rs) {
5931 		idev->if_flags |= IF_RS_SENT;
5932 		idev->rs_interval = rfc3315_s14_backoff_init(
5933 			READ_ONCE(idev->cnf.rtr_solicit_interval));
5934 		idev->rs_probes = 1;
5935 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5936 	}
5937 
5938 	/* Well, that's kinda nasty ... */
5939 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5940 		spin_lock(&ifp->lock);
5941 		if (ifp->tokenized) {
5942 			ifp->valid_lft = 0;
5943 			ifp->prefered_lft = 0;
5944 		}
5945 		spin_unlock(&ifp->lock);
5946 	}
5947 
5948 	write_unlock_bh(&idev->lock);
5949 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5950 	addrconf_verify_rtnl(dev_net(dev));
5951 	return 0;
5952 }
5953 
5954 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5955 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5956 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5957 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5958 					    .reject_message =
5959 						"IFLA_INET6_RA_MTU can not be set" },
5960 };
5961 
5962 static int check_addr_gen_mode(int mode)
5963 {
5964 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5965 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5966 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5967 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5968 		return -EINVAL;
5969 	return 1;
5970 }
5971 
5972 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5973 				int mode)
5974 {
5975 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5976 	    !idev->cnf.stable_secret.initialized &&
5977 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5978 		return -EINVAL;
5979 	return 1;
5980 }
5981 
5982 static int inet6_validate_link_af(const struct net_device *dev,
5983 				  const struct nlattr *nla,
5984 				  struct netlink_ext_ack *extack)
5985 {
5986 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5987 	struct inet6_dev *idev = NULL;
5988 	int err;
5989 
5990 	if (dev) {
5991 		idev = __in6_dev_get(dev);
5992 		if (!idev)
5993 			return -EAFNOSUPPORT;
5994 	}
5995 
5996 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5997 					  inet6_af_policy, extack);
5998 	if (err)
5999 		return err;
6000 
6001 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6002 		return -EINVAL;
6003 
6004 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6005 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6006 
6007 		if (check_addr_gen_mode(mode) < 0)
6008 			return -EINVAL;
6009 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6010 			return -EINVAL;
6011 	}
6012 
6013 	return 0;
6014 }
6015 
6016 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6017 			     struct netlink_ext_ack *extack)
6018 {
6019 	struct inet6_dev *idev = __in6_dev_get(dev);
6020 	struct nlattr *tb[IFLA_INET6_MAX + 1];
6021 	int err;
6022 
6023 	if (!idev)
6024 		return -EAFNOSUPPORT;
6025 
6026 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6027 		return -EINVAL;
6028 
6029 	if (tb[IFLA_INET6_TOKEN]) {
6030 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6031 					extack);
6032 		if (err)
6033 			return err;
6034 	}
6035 
6036 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6037 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6038 
6039 		WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6040 	}
6041 
6042 	return 0;
6043 }
6044 
6045 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6046 			     u32 portid, u32 seq, int event, unsigned int flags)
6047 {
6048 	struct net_device *dev = idev->dev;
6049 	struct ifinfomsg *hdr;
6050 	struct nlmsghdr *nlh;
6051 	int ifindex, iflink;
6052 	void *protoinfo;
6053 
6054 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6055 	if (!nlh)
6056 		return -EMSGSIZE;
6057 
6058 	hdr = nlmsg_data(nlh);
6059 	hdr->ifi_family = AF_INET6;
6060 	hdr->__ifi_pad = 0;
6061 	hdr->ifi_type = dev->type;
6062 	ifindex = READ_ONCE(dev->ifindex);
6063 	hdr->ifi_index = ifindex;
6064 	hdr->ifi_flags = dev_get_flags(dev);
6065 	hdr->ifi_change = 0;
6066 
6067 	iflink = dev_get_iflink(dev);
6068 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6069 	    (dev->addr_len &&
6070 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6071 	    nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6072 	    (ifindex != iflink &&
6073 	     nla_put_u32(skb, IFLA_LINK, iflink)) ||
6074 	    nla_put_u8(skb, IFLA_OPERSTATE,
6075 		       netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6076 		goto nla_put_failure;
6077 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6078 	if (!protoinfo)
6079 		goto nla_put_failure;
6080 
6081 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6082 		goto nla_put_failure;
6083 
6084 	nla_nest_end(skb, protoinfo);
6085 	nlmsg_end(skb, nlh);
6086 	return 0;
6087 
6088 nla_put_failure:
6089 	nlmsg_cancel(skb, nlh);
6090 	return -EMSGSIZE;
6091 }
6092 
6093 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6094 				   struct netlink_ext_ack *extack)
6095 {
6096 	struct ifinfomsg *ifm;
6097 
6098 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6099 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6100 		return -EINVAL;
6101 	}
6102 
6103 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6104 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6105 		return -EINVAL;
6106 	}
6107 
6108 	ifm = nlmsg_data(nlh);
6109 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6110 	    ifm->ifi_change || ifm->ifi_index) {
6111 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6112 		return -EINVAL;
6113 	}
6114 
6115 	return 0;
6116 }
6117 
6118 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6119 {
6120 	struct net *net = sock_net(skb->sk);
6121 	struct {
6122 		unsigned long ifindex;
6123 	} *ctx = (void *)cb->ctx;
6124 	struct net_device *dev;
6125 	struct inet6_dev *idev;
6126 	int err;
6127 
6128 	/* only requests using strict checking can pass data to
6129 	 * influence the dump
6130 	 */
6131 	if (cb->strict_check) {
6132 		err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6133 
6134 		if (err < 0)
6135 			return err;
6136 	}
6137 
6138 	err = 0;
6139 	rcu_read_lock();
6140 	for_each_netdev_dump(net, dev, ctx->ifindex) {
6141 		idev = __in6_dev_get(dev);
6142 		if (!idev)
6143 			continue;
6144 		err = inet6_fill_ifinfo(skb, idev,
6145 					NETLINK_CB(cb->skb).portid,
6146 					cb->nlh->nlmsg_seq,
6147 					RTM_NEWLINK, NLM_F_MULTI);
6148 		if (err < 0)
6149 			break;
6150 	}
6151 	rcu_read_unlock();
6152 
6153 	return err;
6154 }
6155 
6156 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6157 {
6158 	struct sk_buff *skb;
6159 	struct net *net = dev_net(idev->dev);
6160 	int err = -ENOBUFS;
6161 
6162 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6163 	if (!skb)
6164 		goto errout;
6165 
6166 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6167 	if (err < 0) {
6168 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6169 		WARN_ON(err == -EMSGSIZE);
6170 		kfree_skb(skb);
6171 		goto errout;
6172 	}
6173 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6174 	return;
6175 errout:
6176 	if (err < 0)
6177 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6178 }
6179 
6180 static inline size_t inet6_prefix_nlmsg_size(void)
6181 {
6182 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6183 	       + nla_total_size(sizeof(struct in6_addr))
6184 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6185 }
6186 
6187 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6188 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6189 			     int event, unsigned int flags)
6190 {
6191 	struct prefixmsg *pmsg;
6192 	struct nlmsghdr *nlh;
6193 	struct prefix_cacheinfo	ci;
6194 
6195 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6196 	if (!nlh)
6197 		return -EMSGSIZE;
6198 
6199 	pmsg = nlmsg_data(nlh);
6200 	pmsg->prefix_family = AF_INET6;
6201 	pmsg->prefix_pad1 = 0;
6202 	pmsg->prefix_pad2 = 0;
6203 	pmsg->prefix_ifindex = idev->dev->ifindex;
6204 	pmsg->prefix_len = pinfo->prefix_len;
6205 	pmsg->prefix_type = pinfo->type;
6206 	pmsg->prefix_pad3 = 0;
6207 	pmsg->prefix_flags = pinfo->flags;
6208 
6209 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6210 		goto nla_put_failure;
6211 	ci.preferred_time = ntohl(pinfo->prefered);
6212 	ci.valid_time = ntohl(pinfo->valid);
6213 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6214 		goto nla_put_failure;
6215 	nlmsg_end(skb, nlh);
6216 	return 0;
6217 
6218 nla_put_failure:
6219 	nlmsg_cancel(skb, nlh);
6220 	return -EMSGSIZE;
6221 }
6222 
6223 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6224 			 struct prefix_info *pinfo)
6225 {
6226 	struct sk_buff *skb;
6227 	struct net *net = dev_net(idev->dev);
6228 	int err = -ENOBUFS;
6229 
6230 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6231 	if (!skb)
6232 		goto errout;
6233 
6234 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6235 	if (err < 0) {
6236 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6237 		WARN_ON(err == -EMSGSIZE);
6238 		kfree_skb(skb);
6239 		goto errout;
6240 	}
6241 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6242 	return;
6243 errout:
6244 	if (err < 0)
6245 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6246 }
6247 
6248 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6249 {
6250 	struct net *net = dev_net(ifp->idev->dev);
6251 
6252 	if (event)
6253 		ASSERT_RTNL();
6254 
6255 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6256 
6257 	switch (event) {
6258 	case RTM_NEWADDR:
6259 		/*
6260 		 * If the address was optimistic we inserted the route at the
6261 		 * start of our DAD process, so we don't need to do it again.
6262 		 * If the device was taken down in the middle of the DAD
6263 		 * cycle there is a race where we could get here without a
6264 		 * host route, so nothing to insert. That will be fixed when
6265 		 * the device is brought up.
6266 		 */
6267 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6268 			ip6_ins_rt(net, ifp->rt);
6269 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6270 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6271 				&ifp->addr, ifp->idev->dev->name);
6272 		}
6273 
6274 		if (ifp->idev->cnf.forwarding)
6275 			addrconf_join_anycast(ifp);
6276 		if (!ipv6_addr_any(&ifp->peer_addr))
6277 			addrconf_prefix_route(&ifp->peer_addr, 128,
6278 					      ifp->rt_priority, ifp->idev->dev,
6279 					      0, 0, GFP_ATOMIC);
6280 		break;
6281 	case RTM_DELADDR:
6282 		if (ifp->idev->cnf.forwarding)
6283 			addrconf_leave_anycast(ifp);
6284 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6285 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6286 			struct fib6_info *rt;
6287 
6288 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6289 						       ifp->idev->dev, 0, 0,
6290 						       false);
6291 			if (rt)
6292 				ip6_del_rt(net, rt, false);
6293 		}
6294 		if (ifp->rt) {
6295 			ip6_del_rt(net, ifp->rt, false);
6296 			ifp->rt = NULL;
6297 		}
6298 		rt_genid_bump_ipv6(net);
6299 		break;
6300 	}
6301 	atomic_inc(&net->ipv6.dev_addr_genid);
6302 }
6303 
6304 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6305 {
6306 	if (likely(ifp->idev->dead == 0))
6307 		__ipv6_ifa_notify(event, ifp);
6308 }
6309 
6310 #ifdef CONFIG_SYSCTL
6311 
6312 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6313 		void *buffer, size_t *lenp, loff_t *ppos)
6314 {
6315 	int *valp = ctl->data;
6316 	int val = *valp;
6317 	loff_t pos = *ppos;
6318 	struct ctl_table lctl;
6319 	int ret;
6320 
6321 	/*
6322 	 * ctl->data points to idev->cnf.forwarding, we should
6323 	 * not modify it until we get the rtnl lock.
6324 	 */
6325 	lctl = *ctl;
6326 	lctl.data = &val;
6327 
6328 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6329 
6330 	if (write)
6331 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6332 	if (ret)
6333 		*ppos = pos;
6334 	return ret;
6335 }
6336 
6337 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6338 		void *buffer, size_t *lenp, loff_t *ppos)
6339 {
6340 	struct inet6_dev *idev = ctl->extra1;
6341 	int min_mtu = IPV6_MIN_MTU;
6342 	struct ctl_table lctl;
6343 
6344 	lctl = *ctl;
6345 	lctl.extra1 = &min_mtu;
6346 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6347 
6348 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6349 }
6350 
6351 static void dev_disable_change(struct inet6_dev *idev)
6352 {
6353 	struct netdev_notifier_info info;
6354 
6355 	if (!idev || !idev->dev)
6356 		return;
6357 
6358 	netdev_notifier_info_init(&info, idev->dev);
6359 	if (idev->cnf.disable_ipv6)
6360 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6361 	else
6362 		addrconf_notify(NULL, NETDEV_UP, &info);
6363 }
6364 
6365 static void addrconf_disable_change(struct net *net, __s32 newf)
6366 {
6367 	struct net_device *dev;
6368 	struct inet6_dev *idev;
6369 
6370 	for_each_netdev(net, dev) {
6371 		idev = __in6_dev_get(dev);
6372 		if (idev) {
6373 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6374 
6375 			WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6376 			if (changed)
6377 				dev_disable_change(idev);
6378 		}
6379 	}
6380 }
6381 
6382 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6383 {
6384 	struct net *net = (struct net *)table->extra2;
6385 	int old;
6386 
6387 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6388 		WRITE_ONCE(*p, newf);
6389 		return 0;
6390 	}
6391 
6392 	if (!rtnl_trylock())
6393 		return restart_syscall();
6394 
6395 	old = *p;
6396 	WRITE_ONCE(*p, newf);
6397 
6398 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6399 		WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6400 		addrconf_disable_change(net, newf);
6401 	} else if ((!newf) ^ (!old))
6402 		dev_disable_change((struct inet6_dev *)table->extra1);
6403 
6404 	rtnl_unlock();
6405 	return 0;
6406 }
6407 
6408 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6409 		void *buffer, size_t *lenp, loff_t *ppos)
6410 {
6411 	int *valp = ctl->data;
6412 	int val = *valp;
6413 	loff_t pos = *ppos;
6414 	struct ctl_table lctl;
6415 	int ret;
6416 
6417 	/*
6418 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6419 	 * not modify it until we get the rtnl lock.
6420 	 */
6421 	lctl = *ctl;
6422 	lctl.data = &val;
6423 
6424 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6425 
6426 	if (write)
6427 		ret = addrconf_disable_ipv6(ctl, valp, val);
6428 	if (ret)
6429 		*ppos = pos;
6430 	return ret;
6431 }
6432 
6433 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6434 		void *buffer, size_t *lenp, loff_t *ppos)
6435 {
6436 	int *valp = ctl->data;
6437 	int ret;
6438 	int old, new;
6439 
6440 	old = *valp;
6441 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6442 	new = *valp;
6443 
6444 	if (write && old != new) {
6445 		struct net *net = ctl->extra2;
6446 
6447 		if (!rtnl_trylock())
6448 			return restart_syscall();
6449 
6450 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6451 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6452 						     NETCONFA_PROXY_NEIGH,
6453 						     NETCONFA_IFINDEX_DEFAULT,
6454 						     net->ipv6.devconf_dflt);
6455 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6456 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6457 						     NETCONFA_PROXY_NEIGH,
6458 						     NETCONFA_IFINDEX_ALL,
6459 						     net->ipv6.devconf_all);
6460 		else {
6461 			struct inet6_dev *idev = ctl->extra1;
6462 
6463 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6464 						     NETCONFA_PROXY_NEIGH,
6465 						     idev->dev->ifindex,
6466 						     &idev->cnf);
6467 		}
6468 		rtnl_unlock();
6469 	}
6470 
6471 	return ret;
6472 }
6473 
6474 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6475 					 void *buffer, size_t *lenp,
6476 					 loff_t *ppos)
6477 {
6478 	int ret = 0;
6479 	u32 new_val;
6480 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6481 	struct net *net = (struct net *)ctl->extra2;
6482 	struct ctl_table tmp = {
6483 		.data = &new_val,
6484 		.maxlen = sizeof(new_val),
6485 		.mode = ctl->mode,
6486 	};
6487 
6488 	if (!rtnl_trylock())
6489 		return restart_syscall();
6490 
6491 	new_val = *((u32 *)ctl->data);
6492 
6493 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6494 	if (ret != 0)
6495 		goto out;
6496 
6497 	if (write) {
6498 		if (check_addr_gen_mode(new_val) < 0) {
6499 			ret = -EINVAL;
6500 			goto out;
6501 		}
6502 
6503 		if (idev) {
6504 			if (check_stable_privacy(idev, net, new_val) < 0) {
6505 				ret = -EINVAL;
6506 				goto out;
6507 			}
6508 
6509 			if (idev->cnf.addr_gen_mode != new_val) {
6510 				WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6511 				addrconf_init_auto_addrs(idev->dev);
6512 			}
6513 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6514 			struct net_device *dev;
6515 
6516 			WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6517 			for_each_netdev(net, dev) {
6518 				idev = __in6_dev_get(dev);
6519 				if (idev &&
6520 				    idev->cnf.addr_gen_mode != new_val) {
6521 					WRITE_ONCE(idev->cnf.addr_gen_mode,
6522 						  new_val);
6523 					addrconf_init_auto_addrs(idev->dev);
6524 				}
6525 			}
6526 		}
6527 
6528 		WRITE_ONCE(*((u32 *)ctl->data), new_val);
6529 	}
6530 
6531 out:
6532 	rtnl_unlock();
6533 
6534 	return ret;
6535 }
6536 
6537 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6538 					 void *buffer, size_t *lenp,
6539 					 loff_t *ppos)
6540 {
6541 	int err;
6542 	struct in6_addr addr;
6543 	char str[IPV6_MAX_STRLEN];
6544 	struct ctl_table lctl = *ctl;
6545 	struct net *net = ctl->extra2;
6546 	struct ipv6_stable_secret *secret = ctl->data;
6547 
6548 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6549 		return -EIO;
6550 
6551 	lctl.maxlen = IPV6_MAX_STRLEN;
6552 	lctl.data = str;
6553 
6554 	if (!rtnl_trylock())
6555 		return restart_syscall();
6556 
6557 	if (!write && !secret->initialized) {
6558 		err = -EIO;
6559 		goto out;
6560 	}
6561 
6562 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6563 	if (err >= sizeof(str)) {
6564 		err = -EIO;
6565 		goto out;
6566 	}
6567 
6568 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6569 	if (err || !write)
6570 		goto out;
6571 
6572 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6573 		err = -EIO;
6574 		goto out;
6575 	}
6576 
6577 	secret->initialized = true;
6578 	secret->secret = addr;
6579 
6580 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6581 		struct net_device *dev;
6582 
6583 		for_each_netdev(net, dev) {
6584 			struct inet6_dev *idev = __in6_dev_get(dev);
6585 
6586 			if (idev) {
6587 				WRITE_ONCE(idev->cnf.addr_gen_mode,
6588 					   IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6589 			}
6590 		}
6591 	} else {
6592 		struct inet6_dev *idev = ctl->extra1;
6593 
6594 		WRITE_ONCE(idev->cnf.addr_gen_mode,
6595 			   IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6596 	}
6597 
6598 out:
6599 	rtnl_unlock();
6600 
6601 	return err;
6602 }
6603 
6604 static
6605 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6606 						int write, void *buffer,
6607 						size_t *lenp,
6608 						loff_t *ppos)
6609 {
6610 	int *valp = ctl->data;
6611 	int val = *valp;
6612 	loff_t pos = *ppos;
6613 	struct ctl_table lctl;
6614 	int ret;
6615 
6616 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6617 	 * we should not modify it until we get the rtnl lock.
6618 	 */
6619 	lctl = *ctl;
6620 	lctl.data = &val;
6621 
6622 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6623 
6624 	if (write)
6625 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6626 	if (ret)
6627 		*ppos = pos;
6628 	return ret;
6629 }
6630 
6631 static
6632 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6633 {
6634 	if (rt) {
6635 		if (action)
6636 			rt->dst.flags |= DST_NOPOLICY;
6637 		else
6638 			rt->dst.flags &= ~DST_NOPOLICY;
6639 	}
6640 }
6641 
6642 static
6643 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6644 {
6645 	struct inet6_ifaddr *ifa;
6646 
6647 	read_lock_bh(&idev->lock);
6648 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6649 		spin_lock(&ifa->lock);
6650 		if (ifa->rt) {
6651 			/* host routes only use builtin fib6_nh */
6652 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6653 			int cpu;
6654 
6655 			rcu_read_lock();
6656 			ifa->rt->dst_nopolicy = val ? true : false;
6657 			if (nh->rt6i_pcpu) {
6658 				for_each_possible_cpu(cpu) {
6659 					struct rt6_info **rtp;
6660 
6661 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6662 					addrconf_set_nopolicy(*rtp, val);
6663 				}
6664 			}
6665 			rcu_read_unlock();
6666 		}
6667 		spin_unlock(&ifa->lock);
6668 	}
6669 	read_unlock_bh(&idev->lock);
6670 }
6671 
6672 static
6673 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6674 {
6675 	struct net *net = (struct net *)ctl->extra2;
6676 	struct inet6_dev *idev;
6677 
6678 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6679 		WRITE_ONCE(*valp, val);
6680 		return 0;
6681 	}
6682 
6683 	if (!rtnl_trylock())
6684 		return restart_syscall();
6685 
6686 	WRITE_ONCE(*valp, val);
6687 
6688 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6689 		struct net_device *dev;
6690 
6691 		for_each_netdev(net, dev) {
6692 			idev = __in6_dev_get(dev);
6693 			if (idev)
6694 				addrconf_disable_policy_idev(idev, val);
6695 		}
6696 	} else {
6697 		idev = (struct inet6_dev *)ctl->extra1;
6698 		addrconf_disable_policy_idev(idev, val);
6699 	}
6700 
6701 	rtnl_unlock();
6702 	return 0;
6703 }
6704 
6705 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6706 				   void *buffer, size_t *lenp, loff_t *ppos)
6707 {
6708 	int *valp = ctl->data;
6709 	int val = *valp;
6710 	loff_t pos = *ppos;
6711 	struct ctl_table lctl;
6712 	int ret;
6713 
6714 	lctl = *ctl;
6715 	lctl.data = &val;
6716 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6717 
6718 	if (write && (*valp != val))
6719 		ret = addrconf_disable_policy(ctl, valp, val);
6720 
6721 	if (ret)
6722 		*ppos = pos;
6723 
6724 	return ret;
6725 }
6726 
6727 static int minus_one = -1;
6728 static const int two_five_five = 255;
6729 static u32 ioam6_if_id_max = U16_MAX;
6730 
6731 static const struct ctl_table addrconf_sysctl[] = {
6732 	{
6733 		.procname	= "forwarding",
6734 		.data		= &ipv6_devconf.forwarding,
6735 		.maxlen		= sizeof(int),
6736 		.mode		= 0644,
6737 		.proc_handler	= addrconf_sysctl_forward,
6738 	},
6739 	{
6740 		.procname	= "hop_limit",
6741 		.data		= &ipv6_devconf.hop_limit,
6742 		.maxlen		= sizeof(int),
6743 		.mode		= 0644,
6744 		.proc_handler	= proc_dointvec_minmax,
6745 		.extra1		= (void *)SYSCTL_ONE,
6746 		.extra2		= (void *)&two_five_five,
6747 	},
6748 	{
6749 		.procname	= "mtu",
6750 		.data		= &ipv6_devconf.mtu6,
6751 		.maxlen		= sizeof(int),
6752 		.mode		= 0644,
6753 		.proc_handler	= addrconf_sysctl_mtu,
6754 	},
6755 	{
6756 		.procname	= "accept_ra",
6757 		.data		= &ipv6_devconf.accept_ra,
6758 		.maxlen		= sizeof(int),
6759 		.mode		= 0644,
6760 		.proc_handler	= proc_dointvec,
6761 	},
6762 	{
6763 		.procname	= "accept_redirects",
6764 		.data		= &ipv6_devconf.accept_redirects,
6765 		.maxlen		= sizeof(int),
6766 		.mode		= 0644,
6767 		.proc_handler	= proc_dointvec,
6768 	},
6769 	{
6770 		.procname	= "autoconf",
6771 		.data		= &ipv6_devconf.autoconf,
6772 		.maxlen		= sizeof(int),
6773 		.mode		= 0644,
6774 		.proc_handler	= proc_dointvec,
6775 	},
6776 	{
6777 		.procname	= "dad_transmits",
6778 		.data		= &ipv6_devconf.dad_transmits,
6779 		.maxlen		= sizeof(int),
6780 		.mode		= 0644,
6781 		.proc_handler	= proc_dointvec,
6782 	},
6783 	{
6784 		.procname	= "router_solicitations",
6785 		.data		= &ipv6_devconf.rtr_solicits,
6786 		.maxlen		= sizeof(int),
6787 		.mode		= 0644,
6788 		.proc_handler	= proc_dointvec_minmax,
6789 		.extra1		= &minus_one,
6790 	},
6791 	{
6792 		.procname	= "router_solicitation_interval",
6793 		.data		= &ipv6_devconf.rtr_solicit_interval,
6794 		.maxlen		= sizeof(int),
6795 		.mode		= 0644,
6796 		.proc_handler	= proc_dointvec_jiffies,
6797 	},
6798 	{
6799 		.procname	= "router_solicitation_max_interval",
6800 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6801 		.maxlen		= sizeof(int),
6802 		.mode		= 0644,
6803 		.proc_handler	= proc_dointvec_jiffies,
6804 	},
6805 	{
6806 		.procname	= "router_solicitation_delay",
6807 		.data		= &ipv6_devconf.rtr_solicit_delay,
6808 		.maxlen		= sizeof(int),
6809 		.mode		= 0644,
6810 		.proc_handler	= proc_dointvec_jiffies,
6811 	},
6812 	{
6813 		.procname	= "force_mld_version",
6814 		.data		= &ipv6_devconf.force_mld_version,
6815 		.maxlen		= sizeof(int),
6816 		.mode		= 0644,
6817 		.proc_handler	= proc_dointvec,
6818 	},
6819 	{
6820 		.procname	= "mldv1_unsolicited_report_interval",
6821 		.data		=
6822 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6823 		.maxlen		= sizeof(int),
6824 		.mode		= 0644,
6825 		.proc_handler	= proc_dointvec_ms_jiffies,
6826 	},
6827 	{
6828 		.procname	= "mldv2_unsolicited_report_interval",
6829 		.data		=
6830 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6831 		.maxlen		= sizeof(int),
6832 		.mode		= 0644,
6833 		.proc_handler	= proc_dointvec_ms_jiffies,
6834 	},
6835 	{
6836 		.procname	= "use_tempaddr",
6837 		.data		= &ipv6_devconf.use_tempaddr,
6838 		.maxlen		= sizeof(int),
6839 		.mode		= 0644,
6840 		.proc_handler	= proc_dointvec,
6841 	},
6842 	{
6843 		.procname	= "temp_valid_lft",
6844 		.data		= &ipv6_devconf.temp_valid_lft,
6845 		.maxlen		= sizeof(int),
6846 		.mode		= 0644,
6847 		.proc_handler	= proc_dointvec,
6848 	},
6849 	{
6850 		.procname	= "temp_prefered_lft",
6851 		.data		= &ipv6_devconf.temp_prefered_lft,
6852 		.maxlen		= sizeof(int),
6853 		.mode		= 0644,
6854 		.proc_handler	= proc_dointvec,
6855 	},
6856 	{
6857 		.procname       = "regen_min_advance",
6858 		.data           = &ipv6_devconf.regen_min_advance,
6859 		.maxlen         = sizeof(int),
6860 		.mode           = 0644,
6861 		.proc_handler   = proc_dointvec,
6862 	},
6863 	{
6864 		.procname	= "regen_max_retry",
6865 		.data		= &ipv6_devconf.regen_max_retry,
6866 		.maxlen		= sizeof(int),
6867 		.mode		= 0644,
6868 		.proc_handler	= proc_dointvec,
6869 	},
6870 	{
6871 		.procname	= "max_desync_factor",
6872 		.data		= &ipv6_devconf.max_desync_factor,
6873 		.maxlen		= sizeof(int),
6874 		.mode		= 0644,
6875 		.proc_handler	= proc_dointvec,
6876 	},
6877 	{
6878 		.procname	= "max_addresses",
6879 		.data		= &ipv6_devconf.max_addresses,
6880 		.maxlen		= sizeof(int),
6881 		.mode		= 0644,
6882 		.proc_handler	= proc_dointvec,
6883 	},
6884 	{
6885 		.procname	= "accept_ra_defrtr",
6886 		.data		= &ipv6_devconf.accept_ra_defrtr,
6887 		.maxlen		= sizeof(int),
6888 		.mode		= 0644,
6889 		.proc_handler	= proc_dointvec,
6890 	},
6891 	{
6892 		.procname	= "ra_defrtr_metric",
6893 		.data		= &ipv6_devconf.ra_defrtr_metric,
6894 		.maxlen		= sizeof(u32),
6895 		.mode		= 0644,
6896 		.proc_handler	= proc_douintvec_minmax,
6897 		.extra1		= (void *)SYSCTL_ONE,
6898 	},
6899 	{
6900 		.procname	= "accept_ra_min_hop_limit",
6901 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6902 		.maxlen		= sizeof(int),
6903 		.mode		= 0644,
6904 		.proc_handler	= proc_dointvec,
6905 	},
6906 	{
6907 		.procname	= "accept_ra_min_lft",
6908 		.data		= &ipv6_devconf.accept_ra_min_lft,
6909 		.maxlen		= sizeof(int),
6910 		.mode		= 0644,
6911 		.proc_handler	= proc_dointvec,
6912 	},
6913 	{
6914 		.procname	= "accept_ra_pinfo",
6915 		.data		= &ipv6_devconf.accept_ra_pinfo,
6916 		.maxlen		= sizeof(int),
6917 		.mode		= 0644,
6918 		.proc_handler	= proc_dointvec,
6919 	},
6920 	{
6921 		.procname	= "ra_honor_pio_life",
6922 		.data		= &ipv6_devconf.ra_honor_pio_life,
6923 		.maxlen		= sizeof(u8),
6924 		.mode		= 0644,
6925 		.proc_handler	= proc_dou8vec_minmax,
6926 		.extra1		= SYSCTL_ZERO,
6927 		.extra2		= SYSCTL_ONE,
6928 	},
6929 #ifdef CONFIG_IPV6_ROUTER_PREF
6930 	{
6931 		.procname	= "accept_ra_rtr_pref",
6932 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6933 		.maxlen		= sizeof(int),
6934 		.mode		= 0644,
6935 		.proc_handler	= proc_dointvec,
6936 	},
6937 	{
6938 		.procname	= "router_probe_interval",
6939 		.data		= &ipv6_devconf.rtr_probe_interval,
6940 		.maxlen		= sizeof(int),
6941 		.mode		= 0644,
6942 		.proc_handler	= proc_dointvec_jiffies,
6943 	},
6944 #ifdef CONFIG_IPV6_ROUTE_INFO
6945 	{
6946 		.procname	= "accept_ra_rt_info_min_plen",
6947 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6948 		.maxlen		= sizeof(int),
6949 		.mode		= 0644,
6950 		.proc_handler	= proc_dointvec,
6951 	},
6952 	{
6953 		.procname	= "accept_ra_rt_info_max_plen",
6954 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6955 		.maxlen		= sizeof(int),
6956 		.mode		= 0644,
6957 		.proc_handler	= proc_dointvec,
6958 	},
6959 #endif
6960 #endif
6961 	{
6962 		.procname	= "proxy_ndp",
6963 		.data		= &ipv6_devconf.proxy_ndp,
6964 		.maxlen		= sizeof(int),
6965 		.mode		= 0644,
6966 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6967 	},
6968 	{
6969 		.procname	= "accept_source_route",
6970 		.data		= &ipv6_devconf.accept_source_route,
6971 		.maxlen		= sizeof(int),
6972 		.mode		= 0644,
6973 		.proc_handler	= proc_dointvec,
6974 	},
6975 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6976 	{
6977 		.procname	= "optimistic_dad",
6978 		.data		= &ipv6_devconf.optimistic_dad,
6979 		.maxlen		= sizeof(int),
6980 		.mode		= 0644,
6981 		.proc_handler   = proc_dointvec,
6982 	},
6983 	{
6984 		.procname	= "use_optimistic",
6985 		.data		= &ipv6_devconf.use_optimistic,
6986 		.maxlen		= sizeof(int),
6987 		.mode		= 0644,
6988 		.proc_handler	= proc_dointvec,
6989 	},
6990 #endif
6991 #ifdef CONFIG_IPV6_MROUTE
6992 	{
6993 		.procname	= "mc_forwarding",
6994 		.data		= &ipv6_devconf.mc_forwarding,
6995 		.maxlen		= sizeof(int),
6996 		.mode		= 0444,
6997 		.proc_handler	= proc_dointvec,
6998 	},
6999 #endif
7000 	{
7001 		.procname	= "disable_ipv6",
7002 		.data		= &ipv6_devconf.disable_ipv6,
7003 		.maxlen		= sizeof(int),
7004 		.mode		= 0644,
7005 		.proc_handler	= addrconf_sysctl_disable,
7006 	},
7007 	{
7008 		.procname	= "accept_dad",
7009 		.data		= &ipv6_devconf.accept_dad,
7010 		.maxlen		= sizeof(int),
7011 		.mode		= 0644,
7012 		.proc_handler	= proc_dointvec,
7013 	},
7014 	{
7015 		.procname	= "force_tllao",
7016 		.data		= &ipv6_devconf.force_tllao,
7017 		.maxlen		= sizeof(int),
7018 		.mode		= 0644,
7019 		.proc_handler	= proc_dointvec
7020 	},
7021 	{
7022 		.procname	= "ndisc_notify",
7023 		.data		= &ipv6_devconf.ndisc_notify,
7024 		.maxlen		= sizeof(int),
7025 		.mode		= 0644,
7026 		.proc_handler	= proc_dointvec
7027 	},
7028 	{
7029 		.procname	= "suppress_frag_ndisc",
7030 		.data		= &ipv6_devconf.suppress_frag_ndisc,
7031 		.maxlen		= sizeof(int),
7032 		.mode		= 0644,
7033 		.proc_handler	= proc_dointvec
7034 	},
7035 	{
7036 		.procname	= "accept_ra_from_local",
7037 		.data		= &ipv6_devconf.accept_ra_from_local,
7038 		.maxlen		= sizeof(int),
7039 		.mode		= 0644,
7040 		.proc_handler	= proc_dointvec,
7041 	},
7042 	{
7043 		.procname	= "accept_ra_mtu",
7044 		.data		= &ipv6_devconf.accept_ra_mtu,
7045 		.maxlen		= sizeof(int),
7046 		.mode		= 0644,
7047 		.proc_handler	= proc_dointvec,
7048 	},
7049 	{
7050 		.procname	= "stable_secret",
7051 		.data		= &ipv6_devconf.stable_secret,
7052 		.maxlen		= IPV6_MAX_STRLEN,
7053 		.mode		= 0600,
7054 		.proc_handler	= addrconf_sysctl_stable_secret,
7055 	},
7056 	{
7057 		.procname	= "use_oif_addrs_only",
7058 		.data		= &ipv6_devconf.use_oif_addrs_only,
7059 		.maxlen		= sizeof(int),
7060 		.mode		= 0644,
7061 		.proc_handler	= proc_dointvec,
7062 	},
7063 	{
7064 		.procname	= "ignore_routes_with_linkdown",
7065 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
7066 		.maxlen		= sizeof(int),
7067 		.mode		= 0644,
7068 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
7069 	},
7070 	{
7071 		.procname	= "drop_unicast_in_l2_multicast",
7072 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
7073 		.maxlen		= sizeof(int),
7074 		.mode		= 0644,
7075 		.proc_handler	= proc_dointvec,
7076 	},
7077 	{
7078 		.procname	= "drop_unsolicited_na",
7079 		.data		= &ipv6_devconf.drop_unsolicited_na,
7080 		.maxlen		= sizeof(int),
7081 		.mode		= 0644,
7082 		.proc_handler	= proc_dointvec,
7083 	},
7084 	{
7085 		.procname	= "keep_addr_on_down",
7086 		.data		= &ipv6_devconf.keep_addr_on_down,
7087 		.maxlen		= sizeof(int),
7088 		.mode		= 0644,
7089 		.proc_handler	= proc_dointvec,
7090 
7091 	},
7092 	{
7093 		.procname	= "seg6_enabled",
7094 		.data		= &ipv6_devconf.seg6_enabled,
7095 		.maxlen		= sizeof(int),
7096 		.mode		= 0644,
7097 		.proc_handler	= proc_dointvec,
7098 	},
7099 #ifdef CONFIG_IPV6_SEG6_HMAC
7100 	{
7101 		.procname	= "seg6_require_hmac",
7102 		.data		= &ipv6_devconf.seg6_require_hmac,
7103 		.maxlen		= sizeof(int),
7104 		.mode		= 0644,
7105 		.proc_handler	= proc_dointvec,
7106 	},
7107 #endif
7108 	{
7109 		.procname       = "enhanced_dad",
7110 		.data           = &ipv6_devconf.enhanced_dad,
7111 		.maxlen         = sizeof(int),
7112 		.mode           = 0644,
7113 		.proc_handler   = proc_dointvec,
7114 	},
7115 	{
7116 		.procname	= "addr_gen_mode",
7117 		.data		= &ipv6_devconf.addr_gen_mode,
7118 		.maxlen		= sizeof(int),
7119 		.mode		= 0644,
7120 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
7121 	},
7122 	{
7123 		.procname       = "disable_policy",
7124 		.data           = &ipv6_devconf.disable_policy,
7125 		.maxlen         = sizeof(int),
7126 		.mode           = 0644,
7127 		.proc_handler   = addrconf_sysctl_disable_policy,
7128 	},
7129 	{
7130 		.procname	= "ndisc_tclass",
7131 		.data		= &ipv6_devconf.ndisc_tclass,
7132 		.maxlen		= sizeof(int),
7133 		.mode		= 0644,
7134 		.proc_handler	= proc_dointvec_minmax,
7135 		.extra1		= (void *)SYSCTL_ZERO,
7136 		.extra2		= (void *)&two_five_five,
7137 	},
7138 	{
7139 		.procname	= "rpl_seg_enabled",
7140 		.data		= &ipv6_devconf.rpl_seg_enabled,
7141 		.maxlen		= sizeof(int),
7142 		.mode		= 0644,
7143 		.proc_handler	= proc_dointvec,
7144 	},
7145 	{
7146 		.procname	= "ioam6_enabled",
7147 		.data		= &ipv6_devconf.ioam6_enabled,
7148 		.maxlen		= sizeof(u8),
7149 		.mode		= 0644,
7150 		.proc_handler	= proc_dou8vec_minmax,
7151 		.extra1		= (void *)SYSCTL_ZERO,
7152 		.extra2		= (void *)SYSCTL_ONE,
7153 	},
7154 	{
7155 		.procname	= "ioam6_id",
7156 		.data		= &ipv6_devconf.ioam6_id,
7157 		.maxlen		= sizeof(u32),
7158 		.mode		= 0644,
7159 		.proc_handler	= proc_douintvec_minmax,
7160 		.extra1		= (void *)SYSCTL_ZERO,
7161 		.extra2		= (void *)&ioam6_if_id_max,
7162 	},
7163 	{
7164 		.procname	= "ioam6_id_wide",
7165 		.data		= &ipv6_devconf.ioam6_id_wide,
7166 		.maxlen		= sizeof(u32),
7167 		.mode		= 0644,
7168 		.proc_handler	= proc_douintvec,
7169 	},
7170 	{
7171 		.procname	= "ndisc_evict_nocarrier",
7172 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
7173 		.maxlen		= sizeof(u8),
7174 		.mode		= 0644,
7175 		.proc_handler	= proc_dou8vec_minmax,
7176 		.extra1		= (void *)SYSCTL_ZERO,
7177 		.extra2		= (void *)SYSCTL_ONE,
7178 	},
7179 	{
7180 		.procname	= "accept_untracked_na",
7181 		.data		= &ipv6_devconf.accept_untracked_na,
7182 		.maxlen		= sizeof(int),
7183 		.mode		= 0644,
7184 		.proc_handler	= proc_dointvec_minmax,
7185 		.extra1		= SYSCTL_ZERO,
7186 		.extra2		= SYSCTL_TWO,
7187 	},
7188 };
7189 
7190 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7191 		struct inet6_dev *idev, struct ipv6_devconf *p)
7192 {
7193 	size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7194 	int i, ifindex;
7195 	struct ctl_table *table;
7196 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7197 
7198 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7199 	if (!table)
7200 		goto out;
7201 
7202 	for (i = 0; i < table_size; i++) {
7203 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7204 		/* If one of these is already set, then it is not safe to
7205 		 * overwrite either of them: this makes proc_dointvec_minmax
7206 		 * usable.
7207 		 */
7208 		if (!table[i].extra1 && !table[i].extra2) {
7209 			table[i].extra1 = idev; /* embedded; no ref */
7210 			table[i].extra2 = net;
7211 		}
7212 	}
7213 
7214 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7215 
7216 	p->sysctl_header = register_net_sysctl_sz(net, path, table,
7217 						  table_size);
7218 	if (!p->sysctl_header)
7219 		goto free;
7220 
7221 	if (!strcmp(dev_name, "all"))
7222 		ifindex = NETCONFA_IFINDEX_ALL;
7223 	else if (!strcmp(dev_name, "default"))
7224 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7225 	else
7226 		ifindex = idev->dev->ifindex;
7227 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7228 				     ifindex, p);
7229 	return 0;
7230 
7231 free:
7232 	kfree(table);
7233 out:
7234 	return -ENOBUFS;
7235 }
7236 
7237 static void __addrconf_sysctl_unregister(struct net *net,
7238 					 struct ipv6_devconf *p, int ifindex)
7239 {
7240 	const struct ctl_table *table;
7241 
7242 	if (!p->sysctl_header)
7243 		return;
7244 
7245 	table = p->sysctl_header->ctl_table_arg;
7246 	unregister_net_sysctl_table(p->sysctl_header);
7247 	p->sysctl_header = NULL;
7248 	kfree(table);
7249 
7250 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7251 }
7252 
7253 static int addrconf_sysctl_register(struct inet6_dev *idev)
7254 {
7255 	int err;
7256 
7257 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7258 		return -EINVAL;
7259 
7260 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7261 				    &ndisc_ifinfo_sysctl_change);
7262 	if (err)
7263 		return err;
7264 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7265 					 idev, &idev->cnf);
7266 	if (err)
7267 		neigh_sysctl_unregister(idev->nd_parms);
7268 
7269 	return err;
7270 }
7271 
7272 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7273 {
7274 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7275 				     idev->dev->ifindex);
7276 	neigh_sysctl_unregister(idev->nd_parms);
7277 }
7278 
7279 
7280 #endif
7281 
7282 static int __net_init addrconf_init_net(struct net *net)
7283 {
7284 	int err = -ENOMEM;
7285 	struct ipv6_devconf *all, *dflt;
7286 
7287 	spin_lock_init(&net->ipv6.addrconf_hash_lock);
7288 	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7289 	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7290 					   sizeof(struct hlist_head),
7291 					   GFP_KERNEL);
7292 	if (!net->ipv6.inet6_addr_lst)
7293 		goto err_alloc_addr;
7294 
7295 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7296 	if (!all)
7297 		goto err_alloc_all;
7298 
7299 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7300 	if (!dflt)
7301 		goto err_alloc_dflt;
7302 
7303 	if (!net_eq(net, &init_net)) {
7304 		switch (net_inherit_devconf()) {
7305 		case 1:  /* copy from init_net */
7306 			memcpy(all, init_net.ipv6.devconf_all,
7307 			       sizeof(ipv6_devconf));
7308 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7309 			       sizeof(ipv6_devconf_dflt));
7310 			break;
7311 		case 3: /* copy from the current netns */
7312 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7313 			       sizeof(ipv6_devconf));
7314 			memcpy(dflt,
7315 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7316 			       sizeof(ipv6_devconf_dflt));
7317 			break;
7318 		case 0:
7319 		case 2:
7320 			/* use compiled values */
7321 			break;
7322 		}
7323 	}
7324 
7325 	/* these will be inherited by all namespaces */
7326 	dflt->autoconf = ipv6_defaults.autoconf;
7327 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7328 
7329 	dflt->stable_secret.initialized = false;
7330 	all->stable_secret.initialized = false;
7331 
7332 	net->ipv6.devconf_all = all;
7333 	net->ipv6.devconf_dflt = dflt;
7334 
7335 #ifdef CONFIG_SYSCTL
7336 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7337 	if (err < 0)
7338 		goto err_reg_all;
7339 
7340 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7341 	if (err < 0)
7342 		goto err_reg_dflt;
7343 #endif
7344 	return 0;
7345 
7346 #ifdef CONFIG_SYSCTL
7347 err_reg_dflt:
7348 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7349 err_reg_all:
7350 	kfree(dflt);
7351 	net->ipv6.devconf_dflt = NULL;
7352 #endif
7353 err_alloc_dflt:
7354 	kfree(all);
7355 	net->ipv6.devconf_all = NULL;
7356 err_alloc_all:
7357 	kfree(net->ipv6.inet6_addr_lst);
7358 err_alloc_addr:
7359 	return err;
7360 }
7361 
7362 static void __net_exit addrconf_exit_net(struct net *net)
7363 {
7364 	int i;
7365 
7366 #ifdef CONFIG_SYSCTL
7367 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7368 				     NETCONFA_IFINDEX_DEFAULT);
7369 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7370 				     NETCONFA_IFINDEX_ALL);
7371 #endif
7372 	kfree(net->ipv6.devconf_dflt);
7373 	net->ipv6.devconf_dflt = NULL;
7374 	kfree(net->ipv6.devconf_all);
7375 	net->ipv6.devconf_all = NULL;
7376 
7377 	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7378 	/*
7379 	 *	Check hash table, then free it.
7380 	 */
7381 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7382 		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7383 
7384 	kfree(net->ipv6.inet6_addr_lst);
7385 	net->ipv6.inet6_addr_lst = NULL;
7386 }
7387 
7388 static struct pernet_operations addrconf_ops = {
7389 	.init = addrconf_init_net,
7390 	.exit = addrconf_exit_net,
7391 };
7392 
7393 static struct rtnl_af_ops inet6_ops __read_mostly = {
7394 	.family		  = AF_INET6,
7395 	.fill_link_af	  = inet6_fill_link_af,
7396 	.get_link_af_size = inet6_get_link_af_size,
7397 	.validate_link_af = inet6_validate_link_af,
7398 	.set_link_af	  = inet6_set_link_af,
7399 };
7400 
7401 /*
7402  *	Init / cleanup code
7403  */
7404 
7405 int __init addrconf_init(void)
7406 {
7407 	struct inet6_dev *idev;
7408 	int err;
7409 
7410 	err = ipv6_addr_label_init();
7411 	if (err < 0) {
7412 		pr_crit("%s: cannot initialize default policy table: %d\n",
7413 			__func__, err);
7414 		goto out;
7415 	}
7416 
7417 	err = register_pernet_subsys(&addrconf_ops);
7418 	if (err < 0)
7419 		goto out_addrlabel;
7420 
7421 	/* All works using addrconf_wq need to lock rtnl. */
7422 	addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7423 	if (!addrconf_wq) {
7424 		err = -ENOMEM;
7425 		goto out_nowq;
7426 	}
7427 
7428 	rtnl_lock();
7429 	idev = ipv6_add_dev(blackhole_netdev);
7430 	rtnl_unlock();
7431 	if (IS_ERR(idev)) {
7432 		err = PTR_ERR(idev);
7433 		goto errlo;
7434 	}
7435 
7436 	ip6_route_init_special_entries();
7437 
7438 	register_netdevice_notifier(&ipv6_dev_notf);
7439 
7440 	addrconf_verify(&init_net);
7441 
7442 	rtnl_af_register(&inet6_ops);
7443 
7444 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7445 				   NULL, inet6_dump_ifinfo, RTNL_FLAG_DUMP_UNLOCKED);
7446 	if (err < 0)
7447 		goto errout;
7448 
7449 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7450 				   inet6_rtm_newaddr, NULL, 0);
7451 	if (err < 0)
7452 		goto errout;
7453 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7454 				   inet6_rtm_deladdr, NULL, 0);
7455 	if (err < 0)
7456 		goto errout;
7457 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7458 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7459 				   RTNL_FLAG_DOIT_UNLOCKED |
7460 				   RTNL_FLAG_DUMP_UNLOCKED);
7461 	if (err < 0)
7462 		goto errout;
7463 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7464 				   NULL, inet6_dump_ifmcaddr,
7465 				   RTNL_FLAG_DUMP_UNLOCKED);
7466 	if (err < 0)
7467 		goto errout;
7468 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7469 				   NULL, inet6_dump_ifacaddr,
7470 				   RTNL_FLAG_DUMP_UNLOCKED);
7471 	if (err < 0)
7472 		goto errout;
7473 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7474 				   inet6_netconf_get_devconf,
7475 				   inet6_netconf_dump_devconf,
7476 				   RTNL_FLAG_DOIT_UNLOCKED |
7477 				   RTNL_FLAG_DUMP_UNLOCKED);
7478 	if (err < 0)
7479 		goto errout;
7480 	err = ipv6_addr_label_rtnl_register();
7481 	if (err < 0)
7482 		goto errout;
7483 
7484 	return 0;
7485 errout:
7486 	rtnl_unregister_all(PF_INET6);
7487 	rtnl_af_unregister(&inet6_ops);
7488 	unregister_netdevice_notifier(&ipv6_dev_notf);
7489 errlo:
7490 	destroy_workqueue(addrconf_wq);
7491 out_nowq:
7492 	unregister_pernet_subsys(&addrconf_ops);
7493 out_addrlabel:
7494 	ipv6_addr_label_cleanup();
7495 out:
7496 	return err;
7497 }
7498 
7499 void addrconf_cleanup(void)
7500 {
7501 	struct net_device *dev;
7502 
7503 	unregister_netdevice_notifier(&ipv6_dev_notf);
7504 	unregister_pernet_subsys(&addrconf_ops);
7505 	ipv6_addr_label_cleanup();
7506 
7507 	rtnl_af_unregister(&inet6_ops);
7508 
7509 	rtnl_lock();
7510 
7511 	/* clean dev list */
7512 	for_each_netdev(&init_net, dev) {
7513 		if (__in6_dev_get(dev) == NULL)
7514 			continue;
7515 		addrconf_ifdown(dev, true);
7516 	}
7517 	addrconf_ifdown(init_net.loopback_dev, true);
7518 
7519 	rtnl_unlock();
7520 
7521 	destroy_workqueue(addrconf_wq);
7522 }
7523